Пример #1
0
class FloatingInstance(object):
    def __init__(self):
        self.prob = Problem()
        self.params = {}
        self.optimizerSet = False
        self.desvarSet = False
        self.optimizer = None

        # Environmental parameters
        self.params['water_depth'] = 200.0
        #self.params['air_density'] = 1.198
        self.params['base.windLoads.rho'] = 1.198
        #self.params['air_viscosity'] = 1.81e-5
        self.params['base.windLoads.mu'] = 1.81e-5
        self.params['water_density'] = 1025.0
        #self.params['water_viscosity'] = 8.9e-4
        self.params['base.waveLoads.mu'] = 8.9e-4
        #self.params['wave_height'] = 10.8
        self.params['hmax'] = 10.8
        #self.params['wave_period'] = 9.8
        self.params['T'] = 9.8
        self.params['Uc'] = 0.0
        #self.params['wind_reference_speed'] = 11.0
        self.params['Uref'] = 11.0
        #self.params['wind_reference_height'] = 90.0
        self.params['zref'] = 119.0
        #self.params['alpha'] = 0.11
        self.params['shearExp'] = 0.11
        #self.params['morison_mass_coefficient'] = 2.0
        self.params['cm'] = 2.0
        self.params['z0'] = 0.0
        self.params['yaw'] = 0.0
        self.params['beta'] = 0.0
        self.params['cd_usr'] = np.inf

        # Mooring parameters
        self.params['mooring_max_offset'] = 0.1 * self.params[
            'water_depth']  # Assumption
        self.params['mooring_max_heel'] = 10.0
        self.params['number_of_mooring_lines'] = 3
        self.params['mooring_type'] = 'chain'
        self.params['anchor_type'] = 'suctionpile'
        self.params['mooring_cost_rate'] = 1.1
        self.params['drag_embedment_extra_length'] = 300.0
        self.params['number_of_mooring_lines'] = 3

        # Steel properties
        self.params['material_density'] = 7850.0
        self.params['E'] = 200e9
        self.params['G'] = 79.3e9
        self.params['nu'] = 0.26
        self.params['yield_stress'] = 3.45e8
        self.params['loading'] = 'hydrostatic'

        # Design constraints
        self.params['min_taper_ratio'] = 0.4
        self.params['min_diameter_thickness_ratio'] = 120.0

        # Safety factors
        self.params['gamma_f'] = 1.35
        self.params['gamma_b'] = 1.1
        self.params['gamma_m'] = 1.1
        self.params['gamma_n'] = 1.0
        self.params['gamma_fatigue'] = 1.755

        # Typically static- set defaults
        self.params['permanent_ballast_density'] = 4492.0
        self.params['bulkhead_mass_factor'] = 1.0
        self.params['ring_mass_factor'] = 1.0
        self.params['shell_mass_factor'] = 1.0
        self.params['column_mass_factor'] = 1.05
        self.params['outfitting_mass_fraction'] = 0.06
        self.params['ballast_cost_rate'] = 100.0
        self.params['tapered_col_cost_rate'] = 4720.0
        self.params['outfitting_cost_rate'] = 6980.0
        self.params['cross_attachment_pontoons_int'] = 1
        self.params['lower_attachment_pontoons_int'] = 1
        self.params['upper_attachment_pontoons_int'] = 1
        self.params['lower_ring_pontoons_int'] = 1
        self.params['upper_ring_pontoons_int'] = 1
        self.params['outer_cross_pontoons_int'] = 1  #False
        self.params['pontoon_cost_rate'] = 6.250

        # OC4 Tower
        self.params['hub_height'] = 87.6
        self.params['tower_outer_diameter'] = np.linspace(
            6.5, 3.87, NSECTIONS + 1)
        self.params['tower_section_height'] = vecOption(
            77.6 / NSECTIONS, NSECTIONS)
        self.params['tower_wall_thickness'] = np.linspace(
            0.027, 0.019, NSECTIONS + 1)
        self.params['tower_buckling_length'] = 30.0
        self.params['tower_outfitting_factor'] = 1.07
        self.params['rna_mass'] = 350e3  #285598.8
        self.params['rna_I'] = np.array([
            1.14930678e+08, 2.20354030e+07, 1.87597425e+07, 0.0,
            5.03710467e+05, 0.0
        ])
        self.params['rna_cg'] = np.array([-1.13197635, 0.0, 0.50875268])
        # Max thrust
        self.params['rna_force'] = np.array([1284744.196, 0.0, -112400.5527])
        self.params['rna_moment'] = np.array(
            [3963732.762, 896380.8464, -346781.6819])
        # Max wind speed
        #self.params['rna_force']  = np.array([188038.8045, 0,  -16451.2637])
        #self.params['rna_moment'] = np.array([0.0, 131196.8431,  0.0])

        # Typically design (start at OC4 semi)
        self.params['radius_to_auxiliary_column'] = 28.867513459481287
        self.params['number_of_auxiliary_columns'] = 3
        self.params['base_freeboard'] = 10.0
        self.params['auxiliary_freeboard'] = 12.0
        self.params['fairlead'] = 14.0
        self.params[
            'fairlead_offset_from_shell'] = 40.868 - 28.867513459481287 - 6.0
        self.params['base_outer_diameter'] = 6.5
        self.params['base_wall_thickness'] = 0.03
        self.params['auxiliary_wall_thickness'] = 0.06
        self.params['base_permanent_ballast_height'] = 1.0
        self.params['base_stiffener_web_height'] = 0.1
        self.params['base_stiffener_web_thickness'] = 0.04
        self.params['base_stiffener_flange_width'] = 0.1
        self.params['base_stiffener_flange_thickness'] = 0.02
        self.params['base_stiffener_spacing'] = 0.4
        self.params['auxiliary_permanent_ballast_height'] = 0.1
        self.params['auxiliary_stiffener_web_height'] = 0.1
        self.params['auxiliary_stiffener_web_thickness'] = 0.04
        self.params['auxiliary_stiffener_flange_width'] = 0.1
        self.params['auxiliary_stiffener_flange_thickness'] = 0.02
        self.params['auxiliary_stiffener_spacing'] = 0.4
        self.params['pontoon_outer_diameter'] = 2 * 1.6
        self.params['pontoon_wall_thickness'] = 0.0175
        self.params['connection_ratio_max'] = 0.25
        self.params['base_pontoon_attach_lower'] = -20.0
        self.params['base_pontoon_attach_upper'] = 10.0

        self.set_length_base(30.0)
        self.set_length_aux(32.0)

        self.params['auxiliary_section_height'] = np.array(
            [6.0, 0.1, 7.9, 8.0, 10.0])
        self.params['auxiliary_outer_diameter'] = 2 * np.array(
            [12.0, 12.0, 6.0, 6.0, 6.0, 6.0])

        self.params['scope_ratio'] = 835.5 / (self.params['water_depth'] -
                                              self.params['fairlead'])
        self.params['anchor_radius'] = 837.6
        self.params['mooring_diameter'] = 0.0766

    def set_length_base(self, inval):
        self.params['base_section_height'] = vecOption(inval / NSECTIONS,
                                                       NSECTIONS)

    def set_length_aux(self, inval):
        self.params['auxiliary_section_height'] = vecOption(
            inval / NSECTIONS, NSECTIONS)

    def check_vectors(self):
        self.params['tower_outer_diameter'] = vecOption(
            self.params['tower_outer_diameter'], NSECTIONS + 1)
        self.params['tower_wall_thickness'] = vecOption(
            self.params['tower_wall_thickness'], NSECTIONS + 1)
        self.params['tower_section_height'] = vecOption(
            self.params['tower_section_height'], NSECTIONS + 1)
        self.params['base_outer_diameter'] = vecOption(
            self.params['base_outer_diameter'], NSECTIONS + 1)
        self.params['base_wall_thickness'] = vecOption(
            self.params['base_wall_thickness'], NSECTIONS + 1)
        self.params['base_stiffener_web_height'] = vecOption(
            self.params['base_stiffener_web_height'], NSECTIONS)
        self.params['base_stiffener_web_thickness'] = vecOption(
            self.params['base_stiffener_web_thickness'], NSECTIONS)
        self.params['base_stiffener_flange_width'] = vecOption(
            self.params['base_stiffener_flange_width'], NSECTIONS)
        self.params['base_stiffener_flange_thickness'] = vecOption(
            self.params['base_stiffener_flange_thickness'], NSECTIONS)
        self.params['base_stiffener_spacing'] = vecOption(
            self.params['base_stiffener_spacing'], NSECTIONS)
        self.params['base_bulkhead_thickness'] = vecOption(
            self.params['base_bulkhead_thickness'], NSECTIONS + 1)
        #self.params['base_bulkhead_thickness'][:2] = 0.05

        self.params['auxiliary_outer_diameter'] = vecOption(
            self.params['auxiliary_outer_diameter'], NSECTIONS + 1)
        self.params['auxiliary_wall_thickness'] = vecOption(
            self.params['auxiliary_wall_thickness'], NSECTIONS + 1)
        self.params['auxiliary_stiffener_web_height'] = vecOption(
            self.params['auxiliary_stiffener_web_height'], NSECTIONS)
        self.params['auxiliary_stiffener_web_thickness'] = vecOption(
            self.params['auxiliary_stiffener_web_thickness'], NSECTIONS)
        self.params['auxiliary_stiffener_flange_width'] = vecOption(
            self.params['auxiliary_stiffener_flange_width'], NSECTIONS)
        self.params['auxiliary_stiffener_flange_thickness'] = vecOption(
            self.params['auxiliary_stiffener_flange_thickness'], NSECTIONS)
        self.params['auxiliary_stiffener_spacing'] = vecOption(
            self.params['auxiliary_stiffener_spacing'], NSECTIONS)
        self.params['auxiliary_bulkhead_thickness'] = vecOption(
            self.params['auxiliary_bulkhead_thickness'], NSECTIONS + 1)
        #self.params['auxiliary_bulkhead_thickness'][:2] = 0.05

    def save(self, fname):
        assert type(fname) == type(''), 'Input filename must be a string'
        with open(fname, 'wb') as fp:
            pickle.dump(self.params, fp)

    def load(self, fname):
        assert type(fname) == type(''), 'Input filename must be a string'
        with open(fname, 'rb') as fp:
            self.params = pickle.load(fp)

    def get_assembly(self):
        raise NotImplementedError("Subclasses should implement this!")

    def add_design_variable(self, varStr, lowVal, highVal):
        if not self.optimizerSet:
            raise RuntimeError(
                'Must set the optimizer to set the driver first')

        assert type(varStr) == type(''), 'Input variable must be a string'
        assert isinstance(lowVal, (float, int, np.float32, np.float64,
                                   np.int32, np.int64)), 'Invalid lower bound'
        assert isinstance(highVal, (float, int, np.float32, np.float64,
                                    np.int32, np.int64)), 'Invalid upper bound'

        if self.optimizer in ['COBLYA']:
            iscale = min(1.0 / lowVal, highVal / 1.0)
            self.prob.driver.add_desvar(varStr,
                                        lower=lowVal,
                                        upper=highVal,
                                        scaler=iscale)
        else:
            self.prob.driver.add_desvar(varStr, lower=lowVal, upper=highVal)

        self.desvarSet = True

    def set_optimizer(self, optStr):
        assert type(optStr) == type(''), 'Input optimizer must be a string'
        self.optimizer = optStr.upper()

        # Establish the optimization driver
        if self.optimizer in ['SOGA']:
            self.prob.driver = SOGADriverParallel()
        elif self.optimizer in ['COBYLA', 'SLSQP']:
            self.prob.driver = ScipyOptimizer()
        elif self.optimizer in ['CONMIN', 'PSQP', 'SNOPT', 'NSGA2', 'ALPSO']:
            self.prob.driver = pyOptSparseDriver()
        else:
            raise ValueError('Unknown or unworking optimizer. ' + validStr)

        self.optimizerSet = True

        # Set default options
        self.prob.driver.options['optimizer'] = self.optimizer
        if self.optimizer == 'CONMIN':
            self.prob.driver.opt_settings['ITMAX'] = 1000
        elif self.optimizer in ['PSQP']:
            self.prob.driver.opt_settings['MIT'] = 1000
        elif self.optimizer in ['SOGA']:
            self.prob.driver.options['population'] = 200
            self.prob.driver.options['generations'] = 500
        elif self.optimizer in ['NSGA2']:
            self.prob.driver.opt_settings['PopSize'] = 200
            self.prob.driver.opt_settings['maxGen'] = 500
        elif self.optimizer in ['SNOPT']:
            self.prob.driver.opt_settings['Major iterations limit'] = 500
            self.prob.driver.opt_settings['Minor iterations limit'] = 250
            #self.prob.driver.opt_settings['Major optimality tolerance'] = 1e-5
            #self.prob.driver.opt_settings['Major feasibility tolerance'] = 1e-6
            #self.prob.driver.opt_settings['Minor feasibility tolerance'] = 1e-6
            #self.prob.driver.opt_settings['Function precision'] = 1e-12
            #self.prob.driver.opt_settings['Linesearch tolerance'] = 0.4
            #self.prob.driver.opt_settings['LU singularity tolerance'] = 1e30
        elif self.optimizer in ['COBYLA', 'SLSQP']:
            self.prob.driver.options['tol'] = 1e-6
            self.prob.driver.options['maxiter'] = 1000

    def set_options(self, indict):
        if not self.optimizerSet:
            raise RuntimeError(
                'Must set the optimizer to set the driver first')
        assert isinstance(
            indict,
            dict), 'Options must be passed as a string:value dictionary'

        for k in indict.keys():
            if self.optimizer in ['SOGA', 'COBYLA', 'SLSQP']:
                self.prob.driver.options[k] = indict[k]
            elif self.optimizer in [
                    'CONMIN', 'PSQP', 'SNOPT', 'NSGA2', 'ALPSO'
            ]:
                if k in ['title', 'print_results', 'gradient method']:
                    self.prob.driver.options[k] = indict[k]
                else:
                    self.prob.driver.opt_settings[k] = indict[k]

    def get_constraints(self):
        raise NotImplementedError("Subclasses should implement this!")

    def add_objective(self):
        raise NotImplementedError("Subclasses should implement this!")

    def set_inputs(self):
        # Load all variables from local params dictionary
        localnames = self.params.keys()
        for ivar in localnames:
            try:
                self.prob[ivar] = self.params[ivar]
            except KeyError:
                print('Cannot set: ', ivar)
                continue
            except AttributeError as e:
                print('Vector issues?: ', ivar)
                print(e)
                raise e
            except ValueError as e:
                print('Badding setting of: ', ivar)
                print(e)
                raise e

        # Check that everything got set correctly
        namesAssembly = self.prob.root._unknowns_dict.keys()
        for ivar in namesAssembly:
            if self.prob.root._unknowns_dict[ivar].has_key(
                    '_canset_'
            ) and self.prob.root._unknowns_dict[ivar]['_canset_']:
                selfvar = ivar.split('.')[-1]
                if not selfvar in localnames:
                    print('WARNING:', selfvar, 'has not been set!')
        '''
        namesAssembly = self.prob.root._unknowns_dict.keys()
        for ivar in namesAssembly:
            if self.prob.root._unknowns_dict[ivar].has_key('_canset_') and self.prob.root._unknowns_dict[ivar]['_canset_']:
                selfvar = ivar.split('.')[-1]

                selfval = self.prob[selfvar]
                if ( (type(selfval) == type(0.0)) or (type(selfval) == np.float64) or (type(selfval) == np.float32) ) and selfval == 0.0:
                    print(selfvar, 'is zero! Did you set it?')
                if ( (type(selfval) == type(0)) or (type(selfval) == np.int64) or (type(selfval) == np.int32) ) and selfval == 0:
                    print(selfvar, 'is zero! Did you set it?')
                elif type(selfval) == type(np.array([])) and not np.any(selfval):
                    print(selfvar, 'is zero! Did you set it?')

                selfval = getattr(self, selfvar, None)
                if selfval is None:
                    print('Variable not found:', ivar, selfvar, self.prob[ivar])
                else:
                    self.prob[ivar] = selfval
        #raise NotImplementedError("Subclasses should implement this!")
        '''

    def store_results(self):
        localnames = self.params.keys()
        optDict = self.prob.driver.get_desvars()
        for ivar in optDict.keys():
            ival = optDict[ivar]
            if type(ival) == type(np.array([])) and len(ival) == 1:
                ival = ival[0]
            selfvar = ivar.split('.')[-1]
            self.prob[ivar] = ival
            if selfvar in localnames:
                self.params[ivar] = ival
                ivalstr = 'np.array( ' + str(ival.tolist()) + ' )' if type(
                    ival) == type(np.array([])) else str(ival)
                print(ivar, '=', ivalstr)

    def init_problem(self, optFlag=False):
        self.prob.root = self.get_assembly()

        if optFlag:
            if not self.optimizerSet:
                raise RuntimeError(
                    'Must set the optimizer to set the driver first')
            if not self.desvarSet:
                raise RuntimeError(
                    'Must set design variables before running optimization')

            constlist = self.get_constraints()
            for con in constlist:
                self.prob.driver.add_constraint(con[0],
                                                lower=con[1],
                                                upper=con[2],
                                                equals=con[3])

            self.add_objective()

        # Recorder
        #recorder = DumpRecorder('floatingOptimization.dat')
        #recorder.options['record_params'] = True
        #recorder.options['record_metadata'] = False
        #recorder.options['record_derivs'] = False
        #self.prob.driver.add_recorder(recorder)

        # Note this command must be done after the constraints, design variables, and objective have been set,
        # but before the initial conditions are specified (unless we use the default initial conditions )
        # After setting the intial conditions, running setup() again will revert them back to default values
        self.prob.setup()

        self.set_inputs()

        # Checks
        self.prob.check_setup()
        #self.prob.check_total_derivatives()

    def run(self):
        if not self.optimizerSet:
            print('WARNING: executing once because optimizer is not set')
            self.evaluate()
        else:
            self.init_problem(optFlag=True)
            self.prob.run()
            self.store_results()
            print(self.prob.driver.get_constraints())
            #print(self.prob.driver.get_desvars())
            print(self.prob.driver.get_objectives())

    def evaluate(self):
        self.init_problem(optFlag=False)
        self.prob.run_once()

    def visualize(self, fname=None):
        raise NotImplementedError("Subclasses should implement this!")

    def init_figure(self):
        mysky = np.array([135, 206, 250]) / 255.0
        mysky = tuple(mysky.tolist())
        #fig = plt.figure()
        #ax = fig.add_subplot(111, projection='3d')
        #fig = mlab.figure(bgcolor=(1,)*3, size=(1600,1100))
        #fig = mlab.figure(bgcolor=mysky, size=(1600,1100))
        fig = mlab.figure(bgcolor=(0, ) * 3, size=(1600, 1100))
        return fig

    def draw_ocean(self, fig=None):
        if fig is None: fig = self.init_figure()
        npts = 100

        #mybrown = np.array([244, 170, 66]) / 255.0
        #mybrown = tuple(mybrown.tolist())
        mywater = np.array([95, 158, 160
                            ]) / 255.0  #(0.0, 0.0, 0.8) [143, 188, 143]
        mywater = tuple(mywater.tolist())
        alpha = 0.3

        # Waterplane box
        x = y = 50 * np.linspace(-1, 1, npts)
        X, Y = np.meshgrid(x, y)
        Z = np.sin(100 * X * Y)  #np.zeros(X.shape)
        #ax.plot_surface(X, Y, Z, alpha=alpha, color=mywater)
        mlab.mesh(X, Y, Z, opacity=alpha, color=mywater, figure=fig)

        # Sea floor
        Z = -self.params['water_depth'] * np.ones(X.shape)
        #ax.plot_surface(10*X, 10*Y, Z, alpha=1.0, color=mybrown)
        #mlab.mesh(10*X,10*Y,Z, opacity=1.0, color=mybrown, figure=fig)

        # Sides
        #x = 500 * np.linspace(-1, 1, npts)
        #z = self.params['water_depth'] * np.linspace(-1, 0, npts)
        #X,Z = np.meshgrid(x,z)
        #Y = x.max()*np.ones(Z.shape)
        ##ax.plot_surface(X, Y, Z, alpha=alpha, color=mywater)
        #mlab.mesh(X,Y,Z, opacity=alpha, color=mywater, figure=fig)
        #mlab.mesh(X,-Y,Z, opacity=alpha, color=mywater, figure=fig)
        #mlab.mesh(Y,X,Z, opacity=alpha, color=mywater, figure=fig)
        ##mlab.mesh(-Y,X,Z, opacity=alpha, color=mywater, figure=fig)

    def draw_mooring(self, fig, mooring):
        mybrown = np.array([244, 170, 66]) / 255.0
        mybrown = tuple(mybrown.tolist())
        npts = 100

        # Sea floor
        r = np.linspace(0, self.params['anchor_radius'], npts)
        th = np.linspace(0, 2 * np.pi, npts)
        R, TH = np.meshgrid(r, th)
        X = R * np.cos(TH)
        Y = R * np.sin(TH)
        Z = -self.params['water_depth'] * np.ones(X.shape)
        #ax.plot_surface(X, Y, Z, alpha=1.0, color=mybrown)
        mlab.mesh(X, Y, Z, opacity=1.0, color=mybrown, figure=fig)

        cmoor = (0, 1, 0)
        for k in xrange(int(self.params['number_of_mooring_lines'])):
            #ax.plot(mooring[k,:,0], mooring[k,:,1], mooring[k,:,2], 'k', lw=2)
            mlab.plot3d(mooring[k, :, 0],
                        mooring[k, :, 1],
                        mooring[k, :, 2],
                        color=cmoor,
                        tube_radius=0.5 * self.params['mooring_diameter'],
                        figure=fig)

    def draw_pontoons(self, fig, truss, R, freeboard):
        nE = truss.shape[0]
        c = (0.5, 0, 0)
        for k in xrange(nE):
            if np.any(truss[k, 2, :] > freeboard): continue
            mlab.plot3d(truss[k, 0, :],
                        truss[k, 1, :],
                        truss[k, 2, :],
                        color=c,
                        tube_radius=R,
                        figure=fig)

    def draw_column(self,
                    fig,
                    centerline,
                    freeboard,
                    h_section,
                    r_nodes,
                    spacingVec=None,
                    ckIn=None):
        npts = 20

        z_nodes = np.flipud(freeboard -
                            np.r_[0.0, np.cumsum(np.flipud(h_section))])

        th = np.linspace(0, 2 * np.pi, npts)
        for k in xrange(NSECTIONS):
            rk = np.linspace(r_nodes[k], r_nodes[k + 1], npts)
            z = np.linspace(z_nodes[k], z_nodes[k + 1], npts)
            R, TH = np.meshgrid(rk, th)
            Z, _ = np.meshgrid(z, th)
            X = R * np.cos(TH) + centerline[0]
            Y = R * np.sin(TH) + centerline[1]

            # Draw parameters
            if ckIn is None:
                ck = (0.6, ) * 3 if np.mod(k, 2) == 0 else (0.4, ) * 3
            else:
                ck = ckIn
            #ax.plot_surface(X, Y, Z, alpha=0.5, color=ck)
            mlab.mesh(X, Y, Z, opacity=0.9, color=ck, figure=fig)

            if spacingVec is None: continue

            z = z_nodes[k] + spacingVec[k]
            while z < z_nodes[k + 1]:
                rk = np.interp(z, z_nodes[k:], r_nodes[k:])
                #print(z, z_nodes[k], z_nodes[k+1], rk, r_nodes[k], r_nodes[k+1])
                #ax.plot(rk*np.cos(th), rk*np.sin(th), z*np.ones(th.shape), 'r', lw=0.25)
                mlab.plot3d(rk * np.cos(th) + centerline[0],
                            rk * np.sin(th) + centerline[1],
                            z * np.ones(th.shape),
                            color=(0.5, 0, 0),
                            figure=fig)
                z += spacingVec[k]
                '''
                # Web
                r   = np.linspace(rk - self.params['stiffener_web_height'][k], rk, npts)
                R, TH = np.meshgrid(r, th)
                Z, _  = np.meshgrid(z, th)
                X = R*np.cos(TH)
                Y = R*np.sin(TH)
                ax.plot_surface(X, Y, Z, alpha=0.7, color='r')

                # Flange
                r = r[0]
                h = np.linspace(0, self.params['stiffener_flange_width'][k], npts)
                zflange = z + h - 0.5*self.params['stiffener_flange_width'][k]
                R, TH = np.meshgrid(r, th)
                Z, _  = np.meshgrid(zflange, th)
                X = R*np.cos(TH)
                Y = R*np.sin(TH)
                ax.plot_surface(X, Y, Z, alpha=0.7, color='r')
                '''

    def set_figure(self, fig, fname=None):
        #ax.set_aspect('equal')
        #set_axes_equal(ax)
        #ax.autoscale_view(tight=True)
        #ax.set_xlim([-125, 125])
        #ax.set_ylim([-125, 125])
        #ax.set_zlim([-220, 30])
        #plt.axis('off')
        #plt.show()
        mlab.move([-517.16728532, -87.0711504, 5.60826224],
                  [1.35691603e+01, -2.84217094e-14, -1.06547500e+02])
        mlab.view(-170.68320804213343, 78.220729198686854, 549.40101471336777,
                  [1.35691603e+01, 0.0, -1.06547500e+02])
        if not fname is None: mlab.savefig(fname, figure=fig)
        mlab.show(stop=True)
Пример #2
0
class FloatingInstance(object):
    def __init__(self):
        self.prob = Problem()
        self.params = {}
        self.design_variables = []
        self.constraints = []
        self.optimizer = None
        self.objective = None

        # Environmental parameters
        self.params['water_depth']               = 200.0
        #self.params['air_density']              = 1.198
        self.params['main.windLoads.rho']        = 1.198
        #self.params['air_viscosity']            = 1.81e-5
        self.params['main.windLoads.mu']         = 1.81e-5
        self.params['water_density']             = 1025.0
        #self.params['water_viscosity']          = 8.9e-4
        self.params['main.waveLoads.mu']         = 8.9e-4
        #self.params['wave_height']              = 10.8
        self.params['Hs']                      = 10.8
        #self.params['wave_period']              = 9.8
        self.params['T']                         = 9.8
        self.params['Uc']                        = 0.0
        #self.params['wind_reference_speed']     = 11.0
        self.params['Uref']                      = 11.0
        #self.params['wind_reference_height']    = 90.0
        self.params['zref']                      = 119.0
        #self.params['alpha']                    = 0.11
        self.params['shearExp']                  = 0.11
        #self.params['morison_mass_coefficient'] = 2.0
        self.params['cm']                        = 2.0
        self.params['z0']                        = 0.0
        self.params['yaw']                       = 0.0
        self.params['beta']                      = 0.0
        self.params['cd_usr']                    = np.inf
        self.params['z_offset']                  = 0.0
        
        # Encironmental constaints
        self.params['wave_period_range_low']                = 2.0
        self.params['wave_period_range_high']               = 20.0
        
        # Mooring parameters
        self.params['max_offset']                   = 0.1*self.params['water_depth'] # Assumption        
        self.params['operational_heel']             = 6.0
        self.params['max_survival_heel']                    = 15.0
        self.params['mooring_type']                         = 'chain'
        self.params['anchor_type']                          = 'suctionpile'
        self.params['mooring_cost_rate']                    = 1.1
        self.params['number_of_mooring_connections']        = 3
        self.params['mooring_lines_per_connection']         = 1

        # Steel properties
        self.params['material_density']                     = 7850.0
        self.params['E']                                    = 200e9
        self.params['G']                                    = 79.3e9
        self.params['nu']                                   = 0.26
        self.params['yield_stress']                         = 3.45e8
        self.params['loading']                              = 'hydrostatic'

        # Design constraints
        self.params['max_draft']                      = 150.0
        self.params['max_taper_ratio']                      = 0.4
        self.params['min_diameter_thickness_ratio']         = 120.0

        # Safety factors
        self.params['gamma_f']                              = 1.35
        self.params['gamma_b']                              = 1.1
        self.params['gamma_m']                              = 1.1
        self.params['gamma_n']                              = 1.0
        self.params['gamma_fatigue']                        = 1.755
        
        # Typically static- set defaults
        self.params['permanent_ballast_density']            = 4492.0
        self.params['bulkhead_mass_factor']                 = 1.0
        self.params['ring_mass_factor']                     = 1.0
        self.params['shell_mass_factor']                    = 1.0
        self.params['column_mass_factor']                   = 1.05
        self.params['outfitting_mass_fraction']             = 0.06
        self.params['ballast_cost_rate']                    = 0.1
        self.params['material_cost_rate']                   = 1.1
        self.params['labor_cost_rate']                      = 1.0
        self.params['painting_cost_rate']                   = 14.4
        self.params['outfitting_cost_rate']                 = 1.5 * self.params['material_cost_rate']
        self.params['cross_attachment_pontoons_int']        = 1
        self.params['lower_attachment_pontoons_int']        = 1
        self.params['upper_attachment_pontoons_int']        = 1
        self.params['lower_ring_pontoons_int']              = 1
        self.params['upper_ring_pontoons_int']              = 1
        self.params['outer_cross_pontoons_int']             = 1 #False

        # OC4 Tower
        self.params['hub_height']                           = 90.0
        self.params['tower_outer_diameter']                 = np.linspace(6.5, 3.87, TOWSEC+1)
        self.params['tower_section_height']                 = vecOption(77.6/TOWSEC, TOWSEC)
        self.params['tower_wall_thickness']                 = np.linspace(0.027, 0.019, TOWSEC)
        self.params['tower_buckling_length']                = 30.0
        self.params['tower_outfitting_factor']              = 1.07
        self.params['hub_cm']                               = np.array([-5.01910, 0.0, 1.96256])
        self.params['tip_position']                         = np.array([-0.9315, 0.0, 27.5])
        self.params['rna_mass']                             = 350e3 #285598.8
        self.params['rna_I']                                = np.array([1.14930678e+08, 2.20354030e+07, 1.87597425e+07, 0.0, 5.03710467e+05, 0.0])
        self.params['rna_cg']                               = np.array([-1.13197635, 0.0, 0.50875268])
        # Max thrust
        self.params['rna_force']                            = np.array([1284744.196, 0.0,  -112400.5527])
        self.params['rna_moment']                           = np.array([3963732.762, 896380.8464,  -346781.6819])
        # Max wind speed
        #self.params['rna_force']                           = np.array([188038.8045, 0,  -16451.2637])
        #self.params['rna_moment']                          = np.array([0.0, 131196.8431,  0.0])
        self.params['main_bulkhead_thickness']              = 0.05*np.array([1, 1, 0, 0, 1]) # Locations/thickness of internal bulkheads at section interfaces [m]
        self.params['offset_bulkhead_thickness']         = 0.05*np.array([1, 1, 0, 0, 1]) # Locations/thickness of internal bulkheads at section interfaces [m]
        self.params['Rhub']                              = 1.125
        self.params['downwind'] = False
        
        # Typically design (start at OC4 semi)
        self.params['radius_to_offset_column']           = 28.867513459481287
        self.params['number_of_offset_columns']          = 3
        self.params['main_freeboard']                       = 10.0
        self.params['offset_freeboard']                  = 12.0
        self.params['fairlead_location']                    = 0.1875 # Want 14 m
        self.params['fairlead_offset_from_shell']           = 40.868-28.867513459481287-6.0
        self.params['main_outer_diameter']                  = 6.5
        self.params['main_wall_thickness']                  = 0.03
        self.params['offset_wall_thickness']             = 0.06
        self.params['main_permanent_ballast_height']        = 1.0
        self.params['main_stiffener_web_height']            = 0.1
        self.params['main_stiffener_web_thickness']         = 0.04
        self.params['main_stiffener_flange_width']          = 0.1
        self.params['main_stiffener_flange_thickness']      = 0.02
        self.params['main_stiffener_spacing']               = 0.4
        self.params['offset_permanent_ballast_height']   = 0.1
        self.params['offset_stiffener_web_height']       = 0.1
        self.params['offset_stiffener_web_thickness']    = 0.04
        self.params['offset_stiffener_flange_width']     = 0.1
        self.params['offset_stiffener_flange_thickness'] = 0.02
        self.params['offset_stiffener_spacing']          = 0.4
        self.params['fairlead_support_outer_diameter']      = 2*1.6
        self.params['fairlead_support_wall_thickness']      = 0.0175
        self.params['pontoon_outer_diameter']               = 2*1.6
        self.params['pontoon_wall_thickness']               = 0.0175
        self.params['connection_ratio_max']                 = 0.25
        self.params['main_pontoon_attach_lower']            = 0.1
        self.params['main_pontoon_attach_upper']            = 1.0
        self.params['main_buoyancy_tank_diameter']      = 0.0
        self.params['main_buoyancy_tank_height']        = 0.0
        self.params['main_buoyancy_tank_location']      = 0.0
        self.params['offset_buoyancy_tank_diameter']      = 0.0
        self.params['offset_buoyancy_tank_height']        = 0.0
        self.params['offset_buoyancy_tank_location']      = 0.0
        
        self.set_length_main( 30.0 )
        self.set_length_aux( 32.0 )

        self.params['offset_outer_diameter']             = 2*6.0
        self.params['offset_buoyancy_tank_diameter']       = 24.0
        self.params['offset_buoyancy_tank_height']         = 5.0

        self.params['mooring_line_length']                  = 835.5
        self.params['anchor_radius']                        = 837.6
        self.params['mooring_diameter']                     = 0.0766

    def set_length_main(self, inval):
        self.params['main_section_height'] =  vecOption(inval/NSECTIONS, NSECTIONS)
        
    def set_length_aux(self, inval):
        self.params['offset_section_height'] =  vecOption(inval/NSECTIONS, NSECTIONS)
        
    def check_vectors(self):
        self.params['tower_outer_diameter']            = vecOption(self.params['tower_outer_diameter'], TOWSEC+1)
        self.params['tower_wall_thickness']            = vecOption(self.params['tower_wall_thickness'], TOWSEC)
        self.params['tower_section_height']            = vecOption(self.params['tower_section_height'], TOWSEC)
        self.params['main_outer_diameter']             = vecOption(self.params['main_outer_diameter'], NSECTIONS+1)
        self.params['main_wall_thickness']             = vecOption(self.params['main_wall_thickness'], NSECTIONS)
        self.params['main_stiffener_web_height']       = vecOption(self.params['main_stiffener_web_height'], NSECTIONS)
        self.params['main_stiffener_web_thickness']    = vecOption(self.params['main_stiffener_web_thickness'], NSECTIONS)
        self.params['main_stiffener_flange_width']     = vecOption(self.params['main_stiffener_flange_width'], NSECTIONS)
        self.params['main_stiffener_flange_thickness'] = vecOption(self.params['main_stiffener_flange_thickness'], NSECTIONS)
        self.params['main_stiffener_spacing']          = vecOption(self.params['main_stiffener_spacing'], NSECTIONS)
        self.params['main_bulkhead_thickness']         = vecOption(self.params['main_bulkhead_thickness'], NSECTIONS+1)
        #self.params['main_bulkhead_thickness'][:2] = 0.05
        
        self.params['offset_outer_diameter']             = vecOption(self.params['offset_outer_diameter'], NSECTIONS+1)
        self.params['offset_wall_thickness']             = vecOption(self.params['offset_wall_thickness'], NSECTIONS)
        self.params['offset_stiffener_web_height']       = vecOption(self.params['offset_stiffener_web_height'], NSECTIONS)
        self.params['offset_stiffener_web_thickness']    = vecOption(self.params['offset_stiffener_web_thickness'], NSECTIONS)
        self.params['offset_stiffener_flange_width']     = vecOption(self.params['offset_stiffener_flange_width'], NSECTIONS)
        self.params['offset_stiffener_flange_thickness'] = vecOption(self.params['offset_stiffener_flange_thickness'], NSECTIONS)
        self.params['offset_stiffener_spacing']          = vecOption(self.params['offset_stiffener_spacing'], NSECTIONS)
        self.params['offset_bulkhead_thickness']         = vecOption(self.params['offset_bulkhead_thickness'], NSECTIONS+1)
        #self.params['offset_bulkhead_thickness'][:2] = 0.05


    def set_reference(self, instr):
        if instr.upper() in Five_strings:

            self.params['hub_height']              = 90.0
            self.params['tower_outer_diameter']    = np.linspace(6.5, 3.87, TOWSEC+1)
            self.params['tower_section_height']    = vecOption(77.6/NSECTIONS, TOWSEC)
            self.params['tower_wall_thickness']    = np.linspace(0.027, 0.019, TOWSEC)
            self.params['Rhub']                 = 1.125
            self.params['main_freeboard'] = 10.0
            self.params['hub_cm']   = np.array([-5.01910, 0.0, 1.96256])

            if self.params.has_key('rna_mass'):
                self.params['rna_mass'] = 350e3 #285598.8
                self.params['rna_I'] = np.array([1.14930678e+08, 2.20354030e+07, 1.87597425e+07, 0.0, 5.03710467e+05, 0.0])
                self.params['rna_cg'] = np.array([-1.13197635, 0.0, 0.50875268])
                # Max thrust
                self.params['rna_force']      = np.array([ 1.03086517e+06, 0.0, -3.53463639e+06])
                self.params['rna_moment']     = np.array([9817509.35136043, 566675.9644231, -858920.77230378])
                # Max wind speed
                #self.params['rna_force']     = np.array([188038.8045, 0,  -16451.2637])
                #self.params['rna_moment']    = np.array([0.0, 131196.8431,  0.0])
                self.params['tip_position'] = np.array([-0.9315, 0.0, 27.5])
                
            
        elif instr.upper() in Ten_strings:

            #z Do Di Ixx= Iyy J
            #[m] [m] [m] [m4] [m4]
            dtuTowerData = StringIO("""
            145.63 5.50 5.44 2.03 4.07
            134.55 5.79 5.73 2.76 5.52
            124.04 6.07 6.00 3.63 7.26
            113.54 6.35 6.26 4.67 9.35
            103.03 6.63 6.53 5.90 11.80
            92.53 6.91 6.80 7.33 14.67
            82.02 7.19 7.07 8.99 17.98
            71.52 7.46 7.34 10.9 21.80
            61.01 7.74 7.61 13.08 26.15
            50.51 8.02 7.88 15.55 31.09
            40.00 8.30 8.16 15.55 31.09
            40.00 9.00 8.70 34.84 69.68
            38.00 9.00 8.70 35.74 71.48
            36.00 9.00 8.70 36.66 73.32
            34.00 9.00 8.70 37.59 75.18
            32.00 9.00 8.69 38.54 77.08
            30.00 9.00 8.69 39.51 79.01
            28.00 9.00 8.69 40.49 80.97
            26.00 9.00 8.69 41.48 82.97
            24.00 9.00 8.69 42.5 85.00
            22.00 9.00 8.69 43.53 87.05
            20.00 9.00 8.69 44.05 88.10
            16.00 9.00 8.80 28.09 56.18
            12.00 9.00 8.80 28.09 56.18
            8.00 9.00 8.80 28.09 56.18
            4.00 9.00 8.80 28.09 56.18
            0.00 9.00 8.80 28.09 56.18
            -8.400 9.00 8.80 28.09 56.18
            -16.800 9.00 8.80 28.09 56.18
            -25.20 9.00 8.80 28.09 56.18
            -33.60 9.00 8.80 28.09 56.18
            -42.60 9.00 8.80 28.09 56.18
            """)

            
            self.params['hub_height'] = 149.0
            self.params['main_freeboard'] = 30.0
            self.params['main_section_height'] += 3.0
            towerData = np.loadtxt(dtuTowerData)
            towerData = towerData[(towerData[:,0] >= 30.0),:]
            towerData = np.vstack((towerData[0,:], towerData))
            towerData[0,0] = self.params['hub_height']
            towerData[(towerData[:,0] == 40.0),0] += np.array([0.01, 0.0])
            trans_idx = np.where(towerData[:,1] == towerData[-1,1])[0]
            idx = [0, np.int_(trans_idx[0]/2), trans_idx[0], trans_idx[-1]]
            self.params['tower_section_height'] = np.diff( np.flipud( towerData[idx,0] ) )
            self.params['tower_outer_diameter'] = np.flipud( towerData[idx, 1] )
            t_temp = np.flipud( towerData[idx, 1] - towerData[idx, 2] )
            self.params['tower_wall_thickness'] = 0.5*(t_temp[:-1] + t_temp[1:])

            self.params['hub_cm']   = np.array([-7.1, 0.0, 2.75])

            if self.params.has_key('rna_mass'):
                self.params['rna_mass'] = 672300.5303006992
                self.params['rna_I'] = np.array([1.67424279e+08, 1.05417282e+08, 8.52142680e+07, 0.0, 1.53820004e+06, 0.0])
                self.params['rna_cg'] = np.array([-0.60484688, 0.0, 2.5177933 ])
                self.params['rna_force']   = np.array([ 2.11271060e+06, 0.0, -6.81485569e+06])
                self.params['rna_moment']  = np.array([29259007.24076359, 245729.82542406, -3075245.58067333])
                self.params['Rhub']     = 2.3
                self.params['tip_position'] = np.array([-3.551, 0.0, 21.504])
            
        else:
            raise ValueError('Inputs must be either NREL5MW or DTU10MW')

    def save(self, fname):
        assert type(fname) == type(''), 'Input filename must be a string'
        with open(fname,'wb') as fp:
            pickle.dump((self.params, self.design_variables, self.constraints, self.objective, self.optimizer), fp)
            
    def load(self, fname):
        assert type(fname) == type(''), 'Input filename must be a string'
        with open(fname,'rb') as fp:
            newobj = pickle.load(fp)

        if isinstance(newobj, tuple):
            newparams, newdesvar, newcons, newobj, newopt = newobj
            for k in newparams.keys():
                self.params[k] = newparams[k]

            if not newopt is None:
                self.set_optimizer(newopt)
            
            for k in newdesvar:
                if k[0].startswith('sm.'): k[0] = k[0][3:]
                self.add_design_variable(k[0], k[1], k[2])
            for k in newcons:
                if k[0].startswith('sm.'): k[0] = k[0][3:]
                self.add_constraint(k[0], k[1], k[2], k[3])

            if not newobj is None:
                self.add_objective(newobj[0], newobj[1])
            
        # Compatibility with older files
        elif isinstance(newobj, dict):
            for k in newobj.keys():
                self.params[k] = newobj[k]
        
    def get_assembly(self):
        return FloatingSE()

    
    def add_design_variable(self, varStr, lowVal, highVal):
        if self.optimizer is None:
            raise RuntimeError('Must set the optimizer to set the driver first')
        
        assert type(varStr) == type(''), 'Input variable must be a string'
        assert isinstance(lowVal, (float, int, np.float32, np.float64, np.int32, np.int64)), 'Invalid lower bound'
        assert isinstance(highVal, (float, int, np.float32, np.float64, np.int32, np.int64)), 'Invalid upper bound'
        
        if self.optimizer in ['COBLYA']:
            iscale=min(1.0 / lowVal, highVal / 1.0)
            self.prob.driver.add_desvar(varStr, lower=lowVal, upper=highVal, scaler=iscale)
        else:
            self.prob.driver.add_desvar(varStr, lower=lowVal, upper=highVal)

        # Internal record keeping
        self.design_variables.append([varStr, lowVal, highVal])

                                     
    def add_constraint(self, varStr, lowVal, highVal, eqVal):
        if self.optimizer is None:
            raise RuntimeError('Must set the optimizer to set the driver first')
        
        assert type(varStr) == type(''), 'Input variable must be a string'
        if not lowVal is None:
            assert isinstance(lowVal, (float, int, np.float32, np.float64, np.int32, np.int64)), 'Invalid lower bound'
        if not highVal is None:
            assert isinstance(highVal, (float, int, np.float32, np.float64, np.int32, np.int64)), 'Invalid upper bound'
        if not eqVal is None:
            assert isinstance(eqVal, (float, int, np.float32, np.float64, np.int32, np.int64)), 'Invalid equality value'

        self.prob.driver.add_constraint(varStr, lower=lowVal, upper=highVal, equals=eqVal)

        # Internal record keeping
        self.constraints.append([varStr, lowVal, highVal, eqVal])
        
            
    def set_optimizer(self, optStr):
        assert type(optStr) == type(''), 'Input optimizer must be a string'
        self.optimizer = optStr.upper()

        # Reset all design variables and constraints
        self.design_variables = []
        self.constraints = []
        self.objective = None
        
        # Establish the optimization driver
        if self.optimizer in ['SOGA','SOPSO','NM','SUBPLEX']:
            from openmdao.api import HeuristicDriverParallel
            self.prob.driver = HeuristicDriverParallel()
        #elif self.optimizer in ['SUBPLEX']:
            #from openmdao.api import HeuristicDriver
            #self.prob.driver = HeuristicDriver()
        elif self.optimizer in ['COBYLA','SLSQP']:
            self.prob.driver = ScipyOptimizer()
        elif self.optimizer in ['CONMIN', 'PSQP','SNOPT','NSGA2','ALPSO']:
            from openmdao.api import pyOptSparseDriver
            self.prob.driver = pyOptSparseDriver()
        else:
            raise ValueError('Unknown or unworking optimizer. '+self.optimizer)

        # Set default options
        self.prob.driver.options['optimizer'] = self.optimizer
        if self.optimizer == 'CONMIN':
            self.prob.driver.opt_settings['ITMAX'] = 1000
        elif self.optimizer in ['PSQP']:
            self.prob.driver.opt_settings['MIT'] = 1000
        elif self.optimizer in ['SOGA','SOPSO','NM','SUBPLEX']:
            self.prob.driver.options['population'] = 50
            self.prob.driver.options['generations'] = 500
        elif self.optimizer in ['NSGA2']:
            self.prob.driver.opt_settings['PopSize'] = 200
            self.prob.driver.opt_settings['maxGen'] = 500
        elif self.optimizer in ['SNOPT']:
            self.prob.driver.opt_settings['Iterations limit'] = 500
            self.prob.driver.opt_settings['Major optimality tolerance'] = 1e-4
            self.prob.driver.opt_settings['Major feasibility tolerance'] = 1e-6
            self.prob.driver.opt_settings['Minor feasibility tolerance'] = 1e-6
            self.prob.driver.opt_settings['Function precision'] = 1e-8
        elif self.optimizer in ['COBYLA','SLSQP']:
            self.prob.driver.options['tol'] = 1e-6
            self.prob.driver.options['maxiter'] = 1000


    def set_options(self, indict):
        if self.optimizer is None:
            raise RuntimeError('Must set the optimizer to set the driver first')
        assert isinstance(indict, dict), 'Options must be passed as a string:value dictionary'
        
        for k in indict.keys():
            if self.optimizer in ['SOGA','SOPSO','NM','SUBPLEX','COBYLA','SLSQP']:
                self.prob.driver.options[k] = indict[k]
            elif self.optimizer in ['CONMIN', 'PSQP','SNOPT','NSGA2','ALPSO']:
                if k in ['title','print_results','gradient method']:
                    self.prob.driver.options[k] = indict[k]
                else:
                    self.prob.driver.opt_settings[k] = indict[k]
            
    def get_constraints(self):

        conlist = [
            # Try to get tower height to match desired hub height
            ['tow.height_constraint', -1e-2, 1e-2, None],
            ['tow.tip_deflection_margin', 1.0, None, None],
            
            # Ensure that draft is greater than 0 (spar length>0) and that less than water depth
            ['main.draft_margin', None, 1.0, None],
            ['off.draft_margin', None, 1.0, None],
            ['main.wave_height_freeboard_ratio', None, 1.0, None],
            ['off.wave_height_freeboard_ratio', None, 1.0, None],
            ['wave_height_fairlead_ratio', None, 1.0, None],
            
            ['main_offset_spacing', 1.0, None, None],
            
            # Ensure that the radius doesn't change dramatically over a section
            ['main.manufacturability', 0.0, None, None],
            ['main.weldability', None, 0.0, None],
            ['off.manufacturability', 0.0, None, None],
            ['off.weldability', None, 0.0, None],
            ['tow.manufacturability', 0.0, None, None],
            ['tow.weldability', None, 0.0, None],
            
            # Ensure that the spar top matches the tower main
            ['tower_transition_buffer', 0.0, None, None],
            ['nacelle_transition_buffer', 0.0, None, None],

            # Make sure semisub columns don't get submerged
            ['offset_freeboard_heel_margin', 0.0, None, None],
            ['offset_draft_heel_margin', 0.0, None, None],
            
            # Ensure max mooring line tension is less than X% of MBL: 60% for intact mooring, 80% for damanged
            ['axial_unity', 0.0, 1.0, None],
            
            # Ensure there is sufficient mooring line length, MAP doesn't give an error about this
            ['mooring_length_max', None, 1.0, None],
            
            # API Bulletin 2U constraints
            ['main.flange_spacing_ratio', None, 1.0, None],
            ['main.stiffener_radius_ratio', None, 0.5, None],
            ['main.flange_compactness', 1.0, None, None],
            ['main.web_compactness', 1.0, None, None],
            ['main.axial_local_api', None, 1.0, None],
            ['main.axial_general_api', None, 1.0, None],
            ['main.external_local_api', None, 1.0, None],
            ['main.external_general_api', None, 1.0, None],
            
            ['off.flange_spacing_ratio', None, 1.0, None],
            ['off.stiffener_radius_ratio', None, 0.5, None],
            ['off.flange_compactness', 1.0, None, None],
            ['off.web_compactness', 1.0, None, None],
            ['off.axial_local_api', None, 1.0, None],
            ['off.axial_general_api', None, 1.0, None],
            ['off.external_local_api', None, 1.0, None],
            ['off.external_general_api', None, 1.0, None],
            
            # Pontoon tube radii
            #['main_connection_ratio', 0.0, None, None],
            #['offset_connection_ratio', 0.0, None, None],
            
            # Pontoon stress safety factor
            ['pontoon_stress', None, 1.0, None],
            ['pontoon_wave_height_depth_margin', 1.0, None, None],

            # Frame3DD estimates of stress and buckling
            ['tower_stress', None, 1.0, None],
            ['tower_shell_buckling', None, 1.0, None],
            ['tower_global_buckling', None, 1.0, None],

            ['main_stress', None, 1.0, None],
            ['main_shell_buckling', None, 1.0, None],
            ['main_global_buckling', None, 1.0, None],

            ['offset_stress', None, 1.0, None],
            ['offset_shell_buckling', None, 1.0, None],
            ['offset_global_buckling', None, 1.0, None],
            
            # Achieving non-zero variable ballast height means the semi can be balanced with margin as conditions change
            ['variable_ballast_height_ratio', 0.0, 1.0, None],
            ['variable_ballast_mass', 0.0, None, None],
            
            # Metacentric height should be positive for static stability
            ['metacentric_height', 1.0, None, None],
            
            # Center of buoyancy should be above CG (difference should be positive, None],
            #['buoyancy_to_gravity', 0.1, None, None],
            
            # Surge restoring force should be greater than wave-wind forces (ratio < 1, None],
            ['offset_force_ratio', None, 1.0, None],
            
            # Heel angle should be less than 6deg for ordinary operation, less than 10 for extreme conditions
            ['heel_moment_ratio', None, 1.0, None],

            # Wave forcing period should be different than natural periods and structural modes
            ['period_margin_low', None, 1.0, None],
            ['period_margin_high', 1.0, None, None],
            ['modal_margin_low', None, 1.0, None],
            ['modal_margin_high', 1.0, None, None]
        ]
        #raise NotImplementedError("Subclasses should implement this!")
        return self.constraints if len(self.constraints) > 0 else conlist


    def constraint_report(self, tol=2e-2, printFlag=True):
        passStr   = 'yes'
        noStr     = 'NO'
        activeStr = 'ACTIVE'
        eps = 1e-12
        
        deslist = self.design_variables
        for k in deslist:
            myval   = self.params[k[0]]
            normLin = (myval - k[1]) / (k[2] - k[1])
            normLog = (np.log(myval+eps) - np.log(k[1]+eps)) / (np.log(k[2]+eps) - np.log(k[1]+eps))
            margin  = np.maximum(np.minimum(normLin, 1.0-normLin), np.minimum(normLog, 1.0-normLog))
            lowStr  = str(k[1])+' <\t'
            highStr = '< '+str(k[2])+'\t'
            conStr  = activeStr if np.any(margin <= tol) else ''
            valStr  = str(self.params[k[0]])
            if printFlag: print(conStr, lowStr, k[0], highStr, valStr)
        
        #print('Status\tLow\tName\tHigh\tEq\tValue')
        conlist = self.constraints if len(self.constraints) > 0 else self.get_constraints()
        allPass = True
        for k in conlist:
            lowStr   = ''
            highStr  = ''
            eqStr    = ''
            passFlag = True
            myval    = self.prob[k[0]]
            margin   = 0.0 if type(myval) == type(0.0) else np.inf * np.ones(myval.shape)
            if not k[1] is None:
                lowStr   = str(k[1])+' <\t'
                passFlag = passFlag and np.all(myval >= k[1])
                normVal  = myval - k[1]
                if k[1] != 0.0: normVal /= k[1]
                margin = np.minimum(margin, np.abs(normVal))
                
            if not k[2] is None:
                highStr  = '< '+str(k[2])+'\t'
                passFlag = passFlag and np.all(myval <= k[2])
                normVal  = k[2] - myval
                if k[2] != 0.0: normVal /= k[2]
                margin = np.minimum(margin, np.abs(normVal))

            if not k[3] is None:
                highStr = '= '+str(k[3])+'\t'
                passFlag = passFlag and np.all(myval == k[3])
                normVal  = np.abs(k[3] - myval)
                if k[3] != 0.0: normVal /= k[3]
                margin = np.minimum(margin, np.abs(normVal))

            valStr = str(myval)
            if not passFlag:
                conStr = noStr+'-'+activeStr
            elif np.any(margin <= tol):
                conStr = activeStr
            else:
                conStr = passStr
                valStr = ''
            if printFlag: print(conStr, '\t', lowStr, k[0], '\t', highStr, eqStr, '\t', valStr)

            # Record summary statistic
            allPass = allPass and passFlag
            
        return allPass
    
            
    def add_objective(self, varname='total_cost', scale=1e-9):
        if (len(self.prob.driver._objs) == 0):
            self.prob.driver.add_objective(varname, scaler=scale)
            
        self.objective = (varname, scale)

    def set_inputs(self):
        # Load all variables from local params dictionary
        localnames = self.params.keys()

        for ivar in localnames:
            try:
                self.prob[ivar] = self.params[ivar]
            except KeyError:
                print('Cannot set: ', ivar, '=', self.params[ivar])
                continue
            except AttributeError as e:
                print('Vector issues?: ', ivar)
                print(e)
                raise e
            except ValueError as e:
                print('Badding setting of: ', ivar, '=', self.params[ivar])
                print(e)
                raise e

        # Check that everything got set correctly
        namesAssembly = self.prob.root._unknowns_dict.keys()
        for ivar in namesAssembly:
            if self.prob.root._unknowns_dict[ivar].has_key('_canset_') and self.prob.root._unknowns_dict[ivar]['_canset_']:
                selfvar = ivar.split('.')[-1]
                if not selfvar in localnames:
                    print('WARNING:', selfvar, 'has not been set!')
                    

        '''
        namesAssembly = self.prob.root._unknowns_dict.keys()
        for ivar in namesAssembly:
            if self.prob.root._unknowns_dict[ivar].has_key('_canset_') and self.prob.root._unknowns_dict[ivar]['_canset_']:
                selfvar = ivar.split('.')[-1]

                selfval = self.prob[selfvar]
                if ( (type(selfval) == type(0.0)) or (type(selfval) == np.float64) or (type(selfval) == np.float32) ) and selfval == 0.0:
                    print(selfvar, 'is zero! Did you set it?')
                if ( (type(selfval) == type(0)) or (type(selfval) == np.int64) or (type(selfval) == np.int32) ) and selfval == 0:
                    print(selfvar, 'is zero! Did you set it?')
                elif type(selfval) == type(np.array([])) and not np.any(selfval):
                    print(selfvar, 'is zero! Did you set it?')

                selfval = getattr(self, selfvar, None)
                if selfval is None:
                    print('Variable not found:', ivar, selfvar, self.prob[ivar])
                else:
                    self.prob[ivar] = selfval
        #raise NotImplementedError("Subclasses should implement this!")
        '''

    def store_results(self):
        localnames = self.params.keys()
        optDict = self.prob.driver.get_desvars()
        for ivar in optDict.keys():
            ival = optDict[ivar]
            if type(ival) == type(np.array([])) and len(ival) == 1: ival=ival[0]
            selfvar = ivar.split('.')[-1]
            self.prob[ivar] = ival
            if selfvar in localnames:
                self.params[ivar] = ival
                ivalstr = 'np.array( '+str(ival.tolist())+' )' if type(ival)==type(np.array([])) else str(ival)
                print(ivar, '=', ivalstr)

                
    def init_problem(self, optFlag=False):
        self.prob.root = self.get_assembly()
        
        if optFlag:
            if self.optimizer is None:
                raise RuntimeError('Must set the optimizer to set the driver first')
            if len(self.design_variables) == 0:
                raise RuntimeError('Must set design variables before running optimization')
            if len(self.constraints) == 0:
                print('Warning: no constraints set')
            if self.objective is None:
                self.add_objective()

        # Recorder
        #recorder = DumpRecorder('floatingOptimization.dat')
        #recorder.options['record_params'] = True
        #recorder.options['record_metadata'] = False
        #recorder.options['record_derivs'] = False
        #self.prob.driver.add_recorder(recorder)
        
        # Note this command must be done after the constraints, design variables, and objective have been set,
        # but before the initial conditions are specified (unless we use the default initial conditions )
        # After setting the intial conditions, running setup() again will revert them back to default values
        self.prob.setup()

        self.set_inputs()

        # Checks
        self.prob.check_setup()
        #self.prob.check_total_derivatives()

    def run(self):
        if self.optimizer is None:
            print('WARNING: executing once because optimizer is not set')
            self.evaluate()
        else:
            self.init_problem(optFlag=True)
            self.prob.run()
            self.store_results()
            print(self.prob.driver.get_constraints())
            #print(self.prob.driver.get_desvars())
            print(self.prob.driver.get_objectives())
        
    def evaluate(self):
        self.init_problem(optFlag=False)
        self.prob.run_once()


    def visualize(self, fname=None):
        raise NotImplementedError("Subclasses should implement this!")

    def init_figure(self):
        from mayavi import mlab
        mysky   = np.array([135, 206, 250]) / 255.0
        mysky   = tuple(mysky.tolist())
        #fig = plt.figure()
        #ax = fig.add_subplot(111, projection='3d')
        #fig = mlab.figure(bgcolor=(1,)*3, size=(1600,1100))
        #fig = mlab.figure(bgcolor=mysky, size=(1600,1100))
        fig = mlab.figure(bgcolor=(0,)*3, size=(1600,1100))
        return fig

    def draw_ocean(self, fig=None):
        from mayavi import mlab
        if fig is None: fig=self.init_figure()
        npts = 100
        
        #mybrown = np.array([244, 170, 66]) / 255.0
        #mybrown = tuple(mybrown.tolist())
        mywater = np.array([95, 158, 160 ]) / 255.0 #(0.0, 0.0, 0.8) [143, 188, 143]
        mywater = tuple(mywater.tolist())
        alpha   = 0.3

        # Waterplane box
        x = y = 100 * np.linspace(-1, 1, npts)
        X,Y = np.meshgrid(x,y)
        Z   = np.sin(100*X*Y) #np.zeros(X.shape)
        #ax.plot_surface(X, Y, Z, alpha=alpha, color=mywater)
        mlab.mesh(X, Y, Z, opacity=alpha, color=mywater, figure=fig)
        
        # Sea floor
        Z = -self.params['water_depth'] * np.ones(X.shape)
        #ax.plot_surface(10*X, 10*Y, Z, alpha=1.0, color=mybrown)
        #mlab.mesh(10*X,10*Y,Z, opacity=1.0, color=mybrown, figure=fig)

        # Sides
        #x = 500 * np.linspace(-1, 1, npts)
        #z = self.params['water_depth'] * np.linspace(-1, 0, npts)
        #X,Z = np.meshgrid(x,z)
        #Y = x.max()*np.ones(Z.shape)
        ##ax.plot_surface(X, Y, Z, alpha=alpha, color=mywater)
        #mlab.mesh(X,Y,Z, opacity=alpha, color=mywater, figure=fig)
        #mlab.mesh(X,-Y,Z, opacity=alpha, color=mywater, figure=fig)
        #mlab.mesh(Y,X,Z, opacity=alpha, color=mywater, figure=fig)
        ##mlab.mesh(-Y,X,Z, opacity=alpha, color=mywater, figure=fig)

    def draw_mooring(self, fig, mooring):
        from mayavi import mlab
        mybrown = np.array([244, 170, 66]) / 255.0
        mybrown = tuple(mybrown.tolist())
        npts    = 100
        
        # Sea floor
        r  = np.linspace(0, self.params['anchor_radius'], npts)
        th = np.linspace(0, 2*np.pi, npts)
        R, TH = np.meshgrid(r, th)
        X = R*np.cos(TH)
        Y = R*np.sin(TH)
        Z = -self.params['water_depth'] * np.ones(X.shape)
        #ax.plot_surface(X, Y, Z, alpha=1.0, color=mybrown)
        mlab.mesh(X,Y,Z, opacity=1.0, color=mybrown, figure=fig)
        cmoor = (0,0.8,0)
        nlines = int( self.params['number_of_mooring_connections'] * self.params['mooring_lines_per_connection'] )
        for k in xrange(nlines):
            #ax.plot(mooring[k,:,0], mooring[k,:,1], mooring[k,:,2], 'k', lw=2)
            mlab.plot3d(mooring[k,:,0], mooring[k,:,1], mooring[k,:,2], color=cmoor, tube_radius=0.5*self.params['mooring_diameter'], figure=fig)

            
    def draw_pontoons(self, fig, truss, R, freeboard):
        from mayavi import mlab
        nE = truss.shape[0]
        c = (0.5,0,0)
        for k in xrange(nE):
            if np.any(truss[k,2,:] > freeboard): continue
            mlab.plot3d(truss[k,0,:], truss[k,1,:], truss[k,2,:], color=c, tube_radius=R, figure=fig)

            
    def draw_column(self, fig, centerline, freeboard, h_section, r_nodes, spacingVec=None, ckIn=None):
        from mayavi import mlab
        npts = 20

        nsection = h_section.size
        z_nodes = np.flipud( freeboard - np.r_[0.0, np.cumsum(np.flipud(h_section))] )

        th = np.linspace(0, 2*np.pi, npts)
        for k in xrange(nsection):
            rk = np.linspace(r_nodes[k], r_nodes[k+1], npts)
            z  = np.linspace(z_nodes[k], z_nodes[k+1], npts)
            R, TH = np.meshgrid(rk, th)
            Z, _  = np.meshgrid(z, th)
            X = R*np.cos(TH) + centerline[0]
            Y = R*np.sin(TH) + centerline[1]

            # Draw parameters
            if ckIn is None:
                ck = (0.6,)*3 if np.mod(k,2) == 0 else (0.4,)*3
            else:
                ck = ckIn
            #ax.plot_surface(X, Y, Z, alpha=0.5, color=ck)
            mlab.mesh(X, Y, Z, opacity=0.7, color=ck, figure=fig)

            if spacingVec is None: continue
            
            z = z_nodes[k] + spacingVec[k]
            while z < z_nodes[k+1]:
                rk = np.interp(z, z_nodes[k:], r_nodes[k:])
                #ax.plot(rk*np.cos(th), rk*np.sin(th), z*np.ones(th.shape), 'r', lw=0.25)
                mlab.plot3d(rk*np.cos(th) + centerline[0], rk*np.sin(th) + centerline[1], z*np.ones(th.shape), color=(0.5,0,0), figure=fig)
                z += spacingVec[k]
                
                '''
                # Web
                r   = np.linspace(rk - self.params['stiffener_web_height'][k], rk, npts)
                R, TH = np.meshgrid(r, th)
                Z, _  = np.meshgrid(z, th)
                X = R*np.cos(TH)
                Y = R*np.sin(TH)
                ax.plot_surface(X, Y, Z, alpha=0.7, color='r')

                # Flange
                r = r[0]
                h = np.linspace(0, self.params['stiffener_flange_width'][k], npts)
                zflange = z + h - 0.5*self.params['stiffener_flange_width'][k]
                R, TH = np.meshgrid(r, th)
                Z, _  = np.meshgrid(zflange, th)
                X = R*np.cos(TH)
                Y = R*np.sin(TH)
                ax.plot_surface(X, Y, Z, alpha=0.7, color='r')
                '''

    def draw_ballast(self, fig, centerline, freeboard, h_section, r_nodes, h_perm, h_water):
        from mayavi import mlab
        npts = 40
        th = np.linspace(0, 2*np.pi, npts)
        z_nodes = np.flipud( freeboard - np.r_[0.0, np.cumsum(np.flipud(h_section))] )

        # Permanent ballast
        z_perm = z_nodes[0] + np.linspace(0, h_perm, npts)
        r_perm = np.interp(z_perm, z_nodes, r_nodes)
        R, TH = np.meshgrid(r_perm, th)
        Z, _  = np.meshgrid(z_perm, th)
        X = R*np.cos(TH) + centerline[0]
        Y = R*np.sin(TH) + centerline[1]
        ck = np.array([122, 85, 33]) / 255.0
        ck = tuple(ck.tolist())
        mlab.mesh(X, Y, Z, color=ck, figure=fig)

        # Water ballast
        z_water = z_perm[-1] + np.linspace(0, h_water, npts)
        r_water = np.interp(z_water, z_nodes, r_nodes)
        R, TH = np.meshgrid(r_water, th)
        Z, _  = np.meshgrid(z_water, th)
        X = R*np.cos(TH) + centerline[0]
        Y = R*np.sin(TH) + centerline[1]
        ck = (0.0, 0.1, 0.8) # Dark blue
        mlab.mesh(X, Y, Z, color=ck, figure=fig)

    def draw_buoyancy_tank(self, fig, centerline, freeboard, h_section, loc, r_box, h_box):
        from mayavi import mlab
        npts = 20
        z_nodes = np.flipud( freeboard - np.r_[0.0, np.cumsum(np.flipud(h_section))] )
        z_lower = loc*(z_nodes[-1] - z_nodes[0]) + z_nodes[0]

        # Lower and Upper surfaces
        r  = np.linspace(0, r_box, npts)
        th = np.linspace(0, 2*np.pi, npts)
        R, TH = np.meshgrid(r, th)
        X = R*np.cos(TH) + centerline[0]
        Y = R*np.sin(TH) + centerline[1]
        Z = z_lower * np.ones(X.shape)
        ck = (0.9,)*3
        mlab.mesh(X,Y,Z, opacity=0.5, color=ck, figure=fig)
        Z += h_box
        mlab.mesh(X,Y,Z, opacity=0.5, color=ck, figure=fig)

        # Cylinder part
        z  = z_lower + np.linspace(0, h_box, npts)
        Z, TH  = np.meshgrid(z, th)
        R = r_box * np.ones(Z.shape)
        X = R*np.cos(TH) + centerline[0]
        Y = R*np.sin(TH) + centerline[1]
        mlab.mesh(X, Y, Z, opacity=0.5, color=ck, figure=fig)
        
    def set_figure(self, fig, fname=None):
        from mayavi import mlab
        #ax.set_aspect('equal')
        #set_axes_equal(ax)
        #ax.autoscale_view(tight=True)
        #ax.set_xlim([-125, 125])
        #ax.set_ylim([-125, 125])
        #ax.set_zlim([-220, 30])
        #plt.axis('off')
        #plt.show()
        #mlab.move([-517.16728532, -87.0711504, 5.60826224], [1.35691603e+01, -2.84217094e-14, -1.06547500e+02])
        #mlab.view(-170.68320804213343, 78.220729198686854, 549.40101471336777, [1.35691603e+01,  0.0, -1.06547500e+02])
        if not fname is None: mlab.savefig(fname, figure=fig)
        mlab.show(stop=True)
Пример #3
0
                            counter = counter + 1

print cfg_master[caseids[i]]

# Add parallel group to omdao problem, pass in master config file
# strangely Turbsim needs to come second to be run first
if dlc_cfg['RunFAST'] == True:
    root.add('ParallelFASTCases', FST8AeroElasticSolver(cfg_master, caseids))
if dlc_cfg['RunTurbsim'] == True:
    root.add('ParallelTurbsimCases',
             TurbSim_Solver(cfg_master_wind, caseids_wind))

top.setup()

top.check_setup()
#view_model(top)

top.run()

top.cleanup()  #Good practice, especially when using recorder

# ===================== Write Output files into txt =====================

postdir = os.path.join(postdir, 'DLC%s') % dlcstring
if not os.path.exists(postdir):
    os.makedirs(postdir)
print postdir
outfiles_file = os.path.join(postdir, 'Outputfiles.txt')
f = open(outfiles_file, 'w')
f.write(
Пример #4
0
class FloatingInstance(object):
    def __init__(self):
        self.prob = Problem()
        self.params = {}
        self.design_variables = []
        self.constraints = []
        self.optimizer = None
        self.objective = None

        # Environmental parameters
        self.params['water_depth'] = 200.0
        #self.params['air_density']              = 1.198
        self.params['main.windLoads.rho'] = 1.198
        #self.params['air_viscosity']            = 1.81e-5
        self.params['main.windLoads.mu'] = 1.81e-5
        self.params['water_density'] = 1025.0
        #self.params['water_viscosity']          = 8.9e-4
        self.params['main.waveLoads.mu'] = 8.9e-4
        #self.params['wave_height']              = 10.8
        self.params['Hs'] = 10.8
        #self.params['wave_period']              = 9.8
        self.params['T'] = 9.8
        self.params['Uc'] = 0.0
        #self.params['wind_reference_speed']     = 11.0
        self.params['Uref'] = 11.0
        #self.params['wind_reference_height']    = 90.0
        self.params['zref'] = 119.0
        #self.params['alpha']                    = 0.11
        self.params['shearExp'] = 0.11
        #self.params['morison_mass_coefficient'] = 2.0
        self.params['cm'] = 2.0
        self.params['z0'] = 0.0
        self.params['yaw'] = 0.0
        self.params['beta'] = 0.0
        self.params['cd_usr'] = np.inf
        self.params['z_offset'] = 0.0

        # Encironmental constaints
        self.params['wave_period_range_low'] = 2.0
        self.params['wave_period_range_high'] = 20.0

        # Mooring parameters
        self.params['max_offset'] = 0.1 * self.params[
            'water_depth']  # Assumption
        self.params['operational_heel'] = 6.0
        self.params['max_survival_heel'] = 15.0
        self.params['mooring_type'] = 'chain'
        self.params['anchor_type'] = 'suctionpile'
        self.params['mooring_cost_rate'] = 1.1
        self.params['number_of_mooring_connections'] = 3
        self.params['mooring_lines_per_connection'] = 1

        # Steel properties
        self.params['material_density'] = 7850.0
        self.params['E'] = 200e9
        self.params['G'] = 79.3e9
        self.params['nu'] = 0.26
        self.params['yield_stress'] = 3.45e8
        self.params['loading'] = 'hydrostatic'

        # Design constraints
        self.params['max_draft'] = 150.0
        self.params['max_taper_ratio'] = 0.4
        self.params['min_diameter_thickness_ratio'] = 120.0

        # Safety factors
        self.params['gamma_f'] = 1.35
        self.params['gamma_b'] = 1.1
        self.params['gamma_m'] = 1.1
        self.params['gamma_n'] = 1.0
        self.params['gamma_fatigue'] = 1.755

        # Typically static- set defaults
        self.params['permanent_ballast_density'] = 4492.0
        self.params['bulkhead_mass_factor'] = 1.0
        self.params['ring_mass_factor'] = 1.0
        self.params['shell_mass_factor'] = 1.0
        self.params['column_mass_factor'] = 1.05
        self.params['outfitting_mass_fraction'] = 0.06
        self.params['ballast_cost_rate'] = 0.1
        self.params['material_cost_rate'] = 1.1
        self.params['labor_cost_rate'] = 1.0
        self.params['painting_cost_rate'] = 14.4
        self.params[
            'outfitting_cost_rate'] = 1.5 * self.params['material_cost_rate']
        self.params['cross_attachment_pontoons_int'] = 1
        self.params['lower_attachment_pontoons_int'] = 1
        self.params['upper_attachment_pontoons_int'] = 1
        self.params['lower_ring_pontoons_int'] = 1
        self.params['upper_ring_pontoons_int'] = 1
        self.params['outer_cross_pontoons_int'] = 1  #False

        # OC4 Tower
        self.params['hub_height'] = 90.0
        self.params['tower_outer_diameter'] = np.linspace(
            6.5, 3.87, TOWSEC + 1)
        self.params['tower_section_height'] = vecOption(77.6 / TOWSEC, TOWSEC)
        self.params['tower_wall_thickness'] = np.linspace(0.027, 0.019, TOWSEC)
        self.params['tower_buckling_length'] = 30.0
        self.params['tower_outfitting_factor'] = 1.07
        self.params['hub_cm'] = np.array([-5.01910, 0.0, 1.96256])
        self.params['tip_position'] = np.array([-0.9315, 0.0, 27.5])
        self.params['rna_mass'] = 350e3  #285598.8
        self.params['rna_I'] = np.array([
            1.14930678e+08, 2.20354030e+07, 1.87597425e+07, 0.0,
            5.03710467e+05, 0.0
        ])
        self.params['rna_cg'] = np.array([-1.13197635, 0.0, 0.50875268])
        # Max thrust
        self.params['rna_force'] = np.array([1284744.196, 0.0, -112400.5527])
        self.params['rna_moment'] = np.array(
            [3963732.762, 896380.8464, -346781.6819])
        # Max wind speed
        #self.params['rna_force']                           = np.array([188038.8045, 0,  -16451.2637])
        #self.params['rna_moment']                          = np.array([0.0, 131196.8431,  0.0])
        self.params['main_bulkhead_thickness'] = 0.05 * np.array(
            [1, 1, 0, 0, 1]
        )  # Locations/thickness of internal bulkheads at section interfaces [m]
        self.params['offset_bulkhead_thickness'] = 0.05 * np.array(
            [1, 1, 0, 0, 1]
        )  # Locations/thickness of internal bulkheads at section interfaces [m]
        self.params['Rhub'] = 1.125
        self.params['downwind'] = False

        # Typically design (start at OC4 semi)
        self.params['radius_to_offset_column'] = 28.867513459481287
        self.params['number_of_offset_columns'] = 3
        self.params['main_freeboard'] = 10.0
        self.params['offset_freeboard'] = 12.0
        self.params['fairlead_location'] = 0.1875  # Want 14 m
        self.params[
            'fairlead_offset_from_shell'] = 40.868 - 28.867513459481287 - 6.0
        self.params['main_outer_diameter'] = 6.5
        self.params['main_wall_thickness'] = 0.03
        self.params['offset_wall_thickness'] = 0.06
        self.params['main_permanent_ballast_height'] = 1.0
        self.params['main_stiffener_web_height'] = 0.1
        self.params['main_stiffener_web_thickness'] = 0.04
        self.params['main_stiffener_flange_width'] = 0.1
        self.params['main_stiffener_flange_thickness'] = 0.02
        self.params['main_stiffener_spacing'] = 0.4
        self.params['offset_permanent_ballast_height'] = 0.1
        self.params['offset_stiffener_web_height'] = 0.1
        self.params['offset_stiffener_web_thickness'] = 0.04
        self.params['offset_stiffener_flange_width'] = 0.1
        self.params['offset_stiffener_flange_thickness'] = 0.02
        self.params['offset_stiffener_spacing'] = 0.4
        self.params['fairlead_support_outer_diameter'] = 2 * 1.6
        self.params['fairlead_support_wall_thickness'] = 0.0175
        self.params['pontoon_outer_diameter'] = 2 * 1.6
        self.params['pontoon_wall_thickness'] = 0.0175
        self.params['connection_ratio_max'] = 0.25
        self.params['main_pontoon_attach_lower'] = 0.1
        self.params['main_pontoon_attach_upper'] = 1.0
        self.params['main_buoyancy_tank_diameter'] = 0.0
        self.params['main_buoyancy_tank_height'] = 0.0
        self.params['main_buoyancy_tank_location'] = 0.0
        self.params['offset_buoyancy_tank_diameter'] = 0.0
        self.params['offset_buoyancy_tank_height'] = 0.0
        self.params['offset_buoyancy_tank_location'] = 0.0

        self.set_length_main(30.0)
        self.set_length_aux(32.0)

        self.params['offset_outer_diameter'] = 2 * 6.0
        self.params['offset_buoyancy_tank_diameter'] = 24.0
        self.params['offset_buoyancy_tank_height'] = 5.0

        self.params['mooring_line_length'] = 835.5
        self.params['anchor_radius'] = 837.6
        self.params['mooring_diameter'] = 0.0766

    def set_length_main(self, inval):
        self.params['main_section_height'] = vecOption(inval / NSECTIONS,
                                                       NSECTIONS)

    def set_length_aux(self, inval):
        self.params['offset_section_height'] = vecOption(
            inval / NSECTIONS, NSECTIONS)

    def check_vectors(self):
        self.params['tower_outer_diameter'] = vecOption(
            self.params['tower_outer_diameter'], TOWSEC + 1)
        self.params['tower_wall_thickness'] = vecOption(
            self.params['tower_wall_thickness'], TOWSEC)
        self.params['tower_section_height'] = vecOption(
            self.params['tower_section_height'], TOWSEC)
        self.params['main_outer_diameter'] = vecOption(
            self.params['main_outer_diameter'], NSECTIONS + 1)
        self.params['main_wall_thickness'] = vecOption(
            self.params['main_wall_thickness'], NSECTIONS)
        self.params['main_stiffener_web_height'] = vecOption(
            self.params['main_stiffener_web_height'], NSECTIONS)
        self.params['main_stiffener_web_thickness'] = vecOption(
            self.params['main_stiffener_web_thickness'], NSECTIONS)
        self.params['main_stiffener_flange_width'] = vecOption(
            self.params['main_stiffener_flange_width'], NSECTIONS)
        self.params['main_stiffener_flange_thickness'] = vecOption(
            self.params['main_stiffener_flange_thickness'], NSECTIONS)
        self.params['main_stiffener_spacing'] = vecOption(
            self.params['main_stiffener_spacing'], NSECTIONS)
        self.params['main_bulkhead_thickness'] = vecOption(
            self.params['main_bulkhead_thickness'], NSECTIONS + 1)
        #self.params['main_bulkhead_thickness'][:2] = 0.05

        self.params['offset_outer_diameter'] = vecOption(
            self.params['offset_outer_diameter'], NSECTIONS + 1)
        self.params['offset_wall_thickness'] = vecOption(
            self.params['offset_wall_thickness'], NSECTIONS)
        self.params['offset_stiffener_web_height'] = vecOption(
            self.params['offset_stiffener_web_height'], NSECTIONS)
        self.params['offset_stiffener_web_thickness'] = vecOption(
            self.params['offset_stiffener_web_thickness'], NSECTIONS)
        self.params['offset_stiffener_flange_width'] = vecOption(
            self.params['offset_stiffener_flange_width'], NSECTIONS)
        self.params['offset_stiffener_flange_thickness'] = vecOption(
            self.params['offset_stiffener_flange_thickness'], NSECTIONS)
        self.params['offset_stiffener_spacing'] = vecOption(
            self.params['offset_stiffener_spacing'], NSECTIONS)
        self.params['offset_bulkhead_thickness'] = vecOption(
            self.params['offset_bulkhead_thickness'], NSECTIONS + 1)
        #self.params['offset_bulkhead_thickness'][:2] = 0.05

    def set_reference(self, instr):
        if instr.upper() in Five_strings:

            self.params['hub_height'] = 90.0
            self.params['tower_outer_diameter'] = np.linspace(
                6.5, 3.87, TOWSEC + 1)
            self.params['tower_section_height'] = vecOption(
                77.6 / NSECTIONS, TOWSEC)
            self.params['tower_wall_thickness'] = np.linspace(
                0.027, 0.019, TOWSEC)
            self.params['Rhub'] = 1.125
            self.params['main_freeboard'] = 10.0
            self.params['hub_cm'] = np.array([-5.01910, 0.0, 1.96256])

            if self.params.has_key('rna_mass'):
                self.params['rna_mass'] = 350e3  #285598.8
                self.params['rna_I'] = np.array([
                    1.14930678e+08, 2.20354030e+07, 1.87597425e+07, 0.0,
                    5.03710467e+05, 0.0
                ])
                self.params['rna_cg'] = np.array(
                    [-1.13197635, 0.0, 0.50875268])
                # Max thrust
                self.params['rna_force'] = np.array(
                    [1.03086517e+06, 0.0, -3.53463639e+06])
                self.params['rna_moment'] = np.array(
                    [9817509.35136043, 566675.9644231, -858920.77230378])
                # Max wind speed
                #self.params['rna_force']     = np.array([188038.8045, 0,  -16451.2637])
                #self.params['rna_moment']    = np.array([0.0, 131196.8431,  0.0])
                self.params['tip_position'] = np.array([-0.9315, 0.0, 27.5])

        elif instr.upper() in Ten_strings:

            #z Do Di Ixx= Iyy J
            #[m] [m] [m] [m4] [m4]
            dtuTowerData = StringIO("""
            145.63 5.50 5.44 2.03 4.07
            134.55 5.79 5.73 2.76 5.52
            124.04 6.07 6.00 3.63 7.26
            113.54 6.35 6.26 4.67 9.35
            103.03 6.63 6.53 5.90 11.80
            92.53 6.91 6.80 7.33 14.67
            82.02 7.19 7.07 8.99 17.98
            71.52 7.46 7.34 10.9 21.80
            61.01 7.74 7.61 13.08 26.15
            50.51 8.02 7.88 15.55 31.09
            40.00 8.30 8.16 15.55 31.09
            40.00 9.00 8.70 34.84 69.68
            38.00 9.00 8.70 35.74 71.48
            36.00 9.00 8.70 36.66 73.32
            34.00 9.00 8.70 37.59 75.18
            32.00 9.00 8.69 38.54 77.08
            30.00 9.00 8.69 39.51 79.01
            28.00 9.00 8.69 40.49 80.97
            26.00 9.00 8.69 41.48 82.97
            24.00 9.00 8.69 42.5 85.00
            22.00 9.00 8.69 43.53 87.05
            20.00 9.00 8.69 44.05 88.10
            16.00 9.00 8.80 28.09 56.18
            12.00 9.00 8.80 28.09 56.18
            8.00 9.00 8.80 28.09 56.18
            4.00 9.00 8.80 28.09 56.18
            0.00 9.00 8.80 28.09 56.18
            -8.400 9.00 8.80 28.09 56.18
            -16.800 9.00 8.80 28.09 56.18
            -25.20 9.00 8.80 28.09 56.18
            -33.60 9.00 8.80 28.09 56.18
            -42.60 9.00 8.80 28.09 56.18
            """)

            self.params['hub_height'] = 149.0
            self.params['main_freeboard'] = 30.0
            self.params['main_section_height'] += 3.0
            towerData = np.loadtxt(dtuTowerData)
            towerData = towerData[(towerData[:, 0] >= 30.0), :]
            towerData = np.vstack((towerData[0, :], towerData))
            towerData[0, 0] = self.params['hub_height']
            towerData[(towerData[:, 0] == 40.0), 0] += np.array([0.01, 0.0])
            trans_idx = np.where(towerData[:, 1] == towerData[-1, 1])[0]
            idx = [0, np.int_(trans_idx[0] / 2), trans_idx[0], trans_idx[-1]]
            self.params['tower_section_height'] = np.diff(
                np.flipud(towerData[idx, 0]))
            self.params['tower_outer_diameter'] = np.flipud(towerData[idx, 1])
            t_temp = np.flipud(towerData[idx, 1] - towerData[idx, 2])
            self.params['tower_wall_thickness'] = 0.5 * (t_temp[:-1] +
                                                         t_temp[1:])

            self.params['hub_cm'] = np.array([-7.1, 0.0, 2.75])

            if self.params.has_key('rna_mass'):
                self.params['rna_mass'] = 672300.5303006992
                self.params['rna_I'] = np.array([
                    1.67424279e+08, 1.05417282e+08, 8.52142680e+07, 0.0,
                    1.53820004e+06, 0.0
                ])
                self.params['rna_cg'] = np.array([-0.60484688, 0.0, 2.5177933])
                self.params['rna_force'] = np.array(
                    [2.11271060e+06, 0.0, -6.81485569e+06])
                self.params['rna_moment'] = np.array(
                    [29259007.24076359, 245729.82542406, -3075245.58067333])
                self.params['Rhub'] = 2.3
                self.params['tip_position'] = np.array([-3.551, 0.0, 21.504])

        else:
            raise ValueError('Inputs must be either NREL5MW or DTU10MW')

    def save(self, fname):
        assert type(fname) == type(''), 'Input filename must be a string'
        with open(fname, 'wb') as fp:
            pickle.dump((self.params, self.design_variables, self.constraints,
                         self.objective, self.optimizer), fp)

    def load(self, fname):
        assert type(fname) == type(''), 'Input filename must be a string'
        with open(fname, 'rb') as fp:
            newobj = pickle.load(fp)

        if isinstance(newobj, tuple):
            newparams, newdesvar, newcons, newobj, newopt = newobj
            for k in newparams.keys():
                self.params[k] = newparams[k]

            if not newopt is None:
                self.set_optimizer(newopt)

            for k in newdesvar:
                if k[0].startswith('sm.'): k[0] = k[0][3:]
                self.add_design_variable(k[0], k[1], k[2])
            for k in newcons:
                if k[0].startswith('sm.'): k[0] = k[0][3:]
                self.add_constraint(k[0], k[1], k[2], k[3])

            if not newobj is None:
                self.add_objective(newobj[0], newobj[1])

        # Compatibility with older files
        elif isinstance(newobj, dict):
            for k in newobj.keys():
                self.params[k] = newobj[k]

    def get_assembly(self):
        return FloatingSE()

    def add_design_variable(self, varStr, lowVal, highVal):
        if self.optimizer is None:
            raise RuntimeError(
                'Must set the optimizer to set the driver first')

        assert type(varStr) == type(''), 'Input variable must be a string'
        assert isinstance(lowVal, (float, int, np.float32, np.float64,
                                   np.int32, np.int64)), 'Invalid lower bound'
        assert isinstance(highVal, (float, int, np.float32, np.float64,
                                    np.int32, np.int64)), 'Invalid upper bound'

        if self.optimizer in ['COBLYA']:
            iscale = min(1.0 / lowVal, highVal / 1.0)
            self.prob.driver.add_desvar(varStr,
                                        lower=lowVal,
                                        upper=highVal,
                                        scaler=iscale)
        else:
            self.prob.driver.add_desvar(varStr, lower=lowVal, upper=highVal)

        # Internal record keeping
        self.design_variables.append([varStr, lowVal, highVal])

    def add_constraint(self, varStr, lowVal, highVal, eqVal):
        if self.optimizer is None:
            raise RuntimeError(
                'Must set the optimizer to set the driver first')

        assert type(varStr) == type(''), 'Input variable must be a string'
        if not lowVal is None:
            assert isinstance(lowVal,
                              (float, int, np.float32, np.float64, np.int32,
                               np.int64)), 'Invalid lower bound'
        if not highVal is None:
            assert isinstance(highVal,
                              (float, int, np.float32, np.float64, np.int32,
                               np.int64)), 'Invalid upper bound'
        if not eqVal is None:
            assert isinstance(eqVal,
                              (float, int, np.float32, np.float64, np.int32,
                               np.int64)), 'Invalid equality value'

        self.prob.driver.add_constraint(varStr,
                                        lower=lowVal,
                                        upper=highVal,
                                        equals=eqVal)

        # Internal record keeping
        self.constraints.append([varStr, lowVal, highVal, eqVal])

    def set_optimizer(self, optStr):
        assert type(optStr) == type(''), 'Input optimizer must be a string'
        self.optimizer = optStr.upper()

        # Reset all design variables and constraints
        self.design_variables = []
        self.constraints = []
        self.objective = None

        # Establish the optimization driver
        if self.optimizer in ['SOGA', 'SOPSO', 'NM', 'SUBPLEX']:
            from openmdao.api import HeuristicDriverParallel
            self.prob.driver = HeuristicDriverParallel()
        #elif self.optimizer in ['SUBPLEX']:
        #from openmdao.api import HeuristicDriver
        #self.prob.driver = HeuristicDriver()
        elif self.optimizer in ['COBYLA', 'SLSQP']:
            self.prob.driver = ScipyOptimizer()
        elif self.optimizer in ['CONMIN', 'PSQP', 'SNOPT', 'NSGA2', 'ALPSO']:
            from openmdao.api import pyOptSparseDriver
            self.prob.driver = pyOptSparseDriver()
        else:
            raise ValueError('Unknown or unworking optimizer. ' +
                             self.optimizer)

        # Set default options
        self.prob.driver.options['optimizer'] = self.optimizer
        if self.optimizer == 'CONMIN':
            self.prob.driver.opt_settings['ITMAX'] = 1000
        elif self.optimizer in ['PSQP']:
            self.prob.driver.opt_settings['MIT'] = 1000
        elif self.optimizer in ['SOGA', 'SOPSO', 'NM', 'SUBPLEX']:
            self.prob.driver.options['population'] = 50
            self.prob.driver.options['generations'] = 500
        elif self.optimizer in ['NSGA2']:
            self.prob.driver.opt_settings['PopSize'] = 200
            self.prob.driver.opt_settings['maxGen'] = 500
        elif self.optimizer in ['SNOPT']:
            self.prob.driver.opt_settings['Iterations limit'] = 500
            self.prob.driver.opt_settings['Major optimality tolerance'] = 1e-4
            self.prob.driver.opt_settings['Major feasibility tolerance'] = 1e-6
            self.prob.driver.opt_settings['Minor feasibility tolerance'] = 1e-6
            self.prob.driver.opt_settings['Function precision'] = 1e-8
        elif self.optimizer in ['COBYLA', 'SLSQP']:
            self.prob.driver.options['tol'] = 1e-6
            self.prob.driver.options['maxiter'] = 1000

    def set_options(self, indict):
        if self.optimizer is None:
            raise RuntimeError(
                'Must set the optimizer to set the driver first')
        assert isinstance(
            indict,
            dict), 'Options must be passed as a string:value dictionary'

        for k in indict.keys():
            if self.optimizer in [
                    'SOGA', 'SOPSO', 'NM', 'SUBPLEX', 'COBYLA', 'SLSQP'
            ]:
                self.prob.driver.options[k] = indict[k]
            elif self.optimizer in [
                    'CONMIN', 'PSQP', 'SNOPT', 'NSGA2', 'ALPSO'
            ]:
                if k in ['title', 'print_results', 'gradient method']:
                    self.prob.driver.options[k] = indict[k]
                else:
                    self.prob.driver.opt_settings[k] = indict[k]

    def get_constraints(self):

        conlist = [
            # Try to get tower height to match desired hub height
            ['tow.height_constraint', -1e-2, 1e-2, None],
            ['tow.tip_deflection_margin', 1.0, None, None],

            # Ensure that draft is greater than 0 (spar length>0) and that less than water depth
            ['main.draft_margin', None, 1.0, None],
            ['off.draft_margin', None, 1.0, None],
            ['main.wave_height_freeboard_ratio', None, 1.0, None],
            ['off.wave_height_freeboard_ratio', None, 1.0, None],
            ['wave_height_fairlead_ratio', None, 1.0, None],
            ['main_offset_spacing', 1.0, None, None],

            # Ensure that the radius doesn't change dramatically over a section
            ['main.manufacturability', 0.0, None, None],
            ['main.weldability', None, 0.0, None],
            ['off.manufacturability', 0.0, None, None],
            ['off.weldability', None, 0.0, None],
            ['tow.manufacturability', 0.0, None, None],
            ['tow.weldability', None, 0.0, None],

            # Ensure that the spar top matches the tower main
            ['tower_transition_buffer', 0.0, None, None],
            ['nacelle_transition_buffer', 0.0, None, None],

            # Make sure semisub columns don't get submerged
            ['offset_freeboard_heel_margin', 0.0, None, None],
            ['offset_draft_heel_margin', 0.0, None, None],

            # Ensure max mooring line tension is less than X% of MBL: 60% for intact mooring, 80% for damanged
            ['axial_unity', 0.0, 1.0, None],

            # Ensure there is sufficient mooring line length, MAP doesn't give an error about this
            ['mooring_length_max', None, 1.0, None],

            # API Bulletin 2U constraints
            ['main.flange_spacing_ratio', None, 1.0, None],
            ['main.stiffener_radius_ratio', None, 0.5, None],
            ['main.flange_compactness', 1.0, None, None],
            ['main.web_compactness', 1.0, None, None],
            ['main.axial_local_api', None, 1.0, None],
            ['main.axial_general_api', None, 1.0, None],
            ['main.external_local_api', None, 1.0, None],
            ['main.external_general_api', None, 1.0, None],
            ['off.flange_spacing_ratio', None, 1.0, None],
            ['off.stiffener_radius_ratio', None, 0.5, None],
            ['off.flange_compactness', 1.0, None, None],
            ['off.web_compactness', 1.0, None, None],
            ['off.axial_local_api', None, 1.0, None],
            ['off.axial_general_api', None, 1.0, None],
            ['off.external_local_api', None, 1.0, None],
            ['off.external_general_api', None, 1.0, None],

            # Pontoon tube radii
            #['main_connection_ratio', 0.0, None, None],
            #['offset_connection_ratio', 0.0, None, None],

            # Pontoon stress safety factor
            ['pontoon_stress', None, 1.0, None],
            ['pontoon_wave_height_depth_margin', 1.0, None, None],

            # Frame3DD estimates of stress and buckling
            ['tower_stress', None, 1.0, None],
            ['tower_shell_buckling', None, 1.0, None],
            ['tower_global_buckling', None, 1.0, None],
            ['main_stress', None, 1.0, None],
            ['main_shell_buckling', None, 1.0, None],
            ['main_global_buckling', None, 1.0, None],
            ['offset_stress', None, 1.0, None],
            ['offset_shell_buckling', None, 1.0, None],
            ['offset_global_buckling', None, 1.0, None],

            # Achieving non-zero variable ballast height means the semi can be balanced with margin as conditions change
            ['variable_ballast_height_ratio', 0.0, 1.0, None],
            ['variable_ballast_mass', 0.0, None, None],

            # Metacentric height should be positive for static stability
            ['metacentric_height', 1.0, None, None],

            # Center of buoyancy should be above CG (difference should be positive, None],
            #['buoyancy_to_gravity', 0.1, None, None],

            # Surge restoring force should be greater than wave-wind forces (ratio < 1, None],
            ['offset_force_ratio', None, 1.0, None],

            # Heel angle should be less than 6deg for ordinary operation, less than 10 for extreme conditions
            ['heel_moment_ratio', None, 1.0, None],

            # Wave forcing period should be different than natural periods and structural modes
            ['period_margin_low', None, 1.0, None],
            ['period_margin_high', 1.0, None, None],
            ['modal_margin_low', None, 1.0, None],
            ['modal_margin_high', 1.0, None, None]
        ]
        #raise NotImplementedError("Subclasses should implement this!")
        return self.constraints if len(self.constraints) > 0 else conlist

    def constraint_report(self, tol=2e-2, printFlag=True):
        passStr = 'yes'
        noStr = 'NO'
        activeStr = 'ACTIVE'
        eps = 1e-12

        deslist = self.design_variables
        for k in deslist:
            myval = self.params[k[0]]
            normLin = (myval - k[1]) / (k[2] - k[1])
            normLog = (np.log(myval + eps) - np.log(k[1] + eps)) / (
                np.log(k[2] + eps) - np.log(k[1] + eps))
            margin = np.maximum(np.minimum(normLin, 1.0 - normLin),
                                np.minimum(normLog, 1.0 - normLog))
            lowStr = str(k[1]) + ' <\t'
            highStr = '< ' + str(k[2]) + '\t'
            conStr = activeStr if np.any(margin <= tol) else ''
            valStr = str(self.params[k[0]])
            if printFlag: print(conStr, lowStr, k[0], highStr, valStr)

        #print('Status\tLow\tName\tHigh\tEq\tValue')
        conlist = self.constraints if len(
            self.constraints) > 0 else self.get_constraints()
        allPass = True
        for k in conlist:
            lowStr = ''
            highStr = ''
            eqStr = ''
            passFlag = True
            myval = self.prob[k[0]]
            margin = 0.0 if type(myval) == type(0.0) else np.inf * np.ones(
                myval.shape)
            if not k[1] is None:
                lowStr = str(k[1]) + ' <\t'
                passFlag = passFlag and np.all(myval >= k[1])
                normVal = myval - k[1]
                if k[1] != 0.0: normVal /= k[1]
                margin = np.minimum(margin, np.abs(normVal))

            if not k[2] is None:
                highStr = '< ' + str(k[2]) + '\t'
                passFlag = passFlag and np.all(myval <= k[2])
                normVal = k[2] - myval
                if k[2] != 0.0: normVal /= k[2]
                margin = np.minimum(margin, np.abs(normVal))

            if not k[3] is None:
                highStr = '= ' + str(k[3]) + '\t'
                passFlag = passFlag and np.all(myval == k[3])
                normVal = np.abs(k[3] - myval)
                if k[3] != 0.0: normVal /= k[3]
                margin = np.minimum(margin, np.abs(normVal))

            valStr = str(myval)
            if not passFlag:
                conStr = noStr + '-' + activeStr
            elif np.any(margin <= tol):
                conStr = activeStr
            else:
                conStr = passStr
                valStr = ''
            if printFlag:
                print(conStr, '\t', lowStr, k[0], '\t', highStr, eqStr, '\t',
                      valStr)

            # Record summary statistic
            allPass = allPass and passFlag

        return allPass

    def add_objective(self, varname='total_cost', scale=1e-9):
        if (len(self.prob.driver._objs) == 0):
            self.prob.driver.add_objective(varname, scaler=scale)

        self.objective = (varname, scale)

    def set_inputs(self):
        # Load all variables from local params dictionary
        localnames = self.params.keys()

        for ivar in localnames:
            try:
                self.prob[ivar] = self.params[ivar]
            except KeyError:
                print('Cannot set: ', ivar, '=', self.params[ivar])
                continue
            except AttributeError as e:
                print('Vector issues?: ', ivar)
                print(e)
                raise e
            except ValueError as e:
                print('Badding setting of: ', ivar, '=', self.params[ivar])
                print(e)
                raise e

        # Check that everything got set correctly
        namesAssembly = self.prob.root._unknowns_dict.keys()
        for ivar in namesAssembly:
            if self.prob.root._unknowns_dict[ivar].has_key(
                    '_canset_'
            ) and self.prob.root._unknowns_dict[ivar]['_canset_']:
                selfvar = ivar.split('.')[-1]
                if not selfvar in localnames:
                    print('WARNING:', selfvar, 'has not been set!')
        '''
        namesAssembly = self.prob.root._unknowns_dict.keys()
        for ivar in namesAssembly:
            if self.prob.root._unknowns_dict[ivar].has_key('_canset_') and self.prob.root._unknowns_dict[ivar]['_canset_']:
                selfvar = ivar.split('.')[-1]

                selfval = self.prob[selfvar]
                if ( (type(selfval) == type(0.0)) or (type(selfval) == np.float64) or (type(selfval) == np.float32) ) and selfval == 0.0:
                    print(selfvar, 'is zero! Did you set it?')
                if ( (type(selfval) == type(0)) or (type(selfval) == np.int64) or (type(selfval) == np.int32) ) and selfval == 0:
                    print(selfvar, 'is zero! Did you set it?')
                elif type(selfval) == type(np.array([])) and not np.any(selfval):
                    print(selfvar, 'is zero! Did you set it?')

                selfval = getattr(self, selfvar, None)
                if selfval is None:
                    print('Variable not found:', ivar, selfvar, self.prob[ivar])
                else:
                    self.prob[ivar] = selfval
        #raise NotImplementedError("Subclasses should implement this!")
        '''

    def store_results(self):
        localnames = self.params.keys()
        optDict = self.prob.driver.get_desvars()
        for ivar in optDict.keys():
            ival = optDict[ivar]
            if type(ival) == type(np.array([])) and len(ival) == 1:
                ival = ival[0]
            selfvar = ivar.split('.')[-1]
            self.prob[ivar] = ival
            if selfvar in localnames:
                self.params[ivar] = ival
                ivalstr = 'np.array( ' + str(ival.tolist()) + ' )' if type(
                    ival) == type(np.array([])) else str(ival)
                print(ivar, '=', ivalstr)

    def init_problem(self, optFlag=False):
        self.prob.root = self.get_assembly()

        if optFlag:
            if self.optimizer is None:
                raise RuntimeError(
                    'Must set the optimizer to set the driver first')
            if len(self.design_variables) == 0:
                raise RuntimeError(
                    'Must set design variables before running optimization')
            if len(self.constraints) == 0:
                print('Warning: no constraints set')
            if self.objective is None:
                self.add_objective()

        # Recorder
        #recorder = DumpRecorder('floatingOptimization.dat')
        #recorder.options['record_params'] = True
        #recorder.options['record_metadata'] = False
        #recorder.options['record_derivs'] = False
        #self.prob.driver.add_recorder(recorder)

        # Note this command must be done after the constraints, design variables, and objective have been set,
        # but before the initial conditions are specified (unless we use the default initial conditions )
        # After setting the intial conditions, running setup() again will revert them back to default values
        self.prob.setup()

        self.set_inputs()

        # Checks
        self.prob.check_setup()
        #self.prob.check_total_derivatives()

    def run(self):
        if self.optimizer is None:
            print('WARNING: executing once because optimizer is not set')
            self.evaluate()
        else:
            self.init_problem(optFlag=True)
            self.prob.run()
            self.store_results()
            print(self.prob.driver.get_constraints())
            #print(self.prob.driver.get_desvars())
            print(self.prob.driver.get_objectives())

    def evaluate(self):
        self.init_problem(optFlag=False)
        self.prob.run_once()

    def visualize(self, fname=None):
        raise NotImplementedError("Subclasses should implement this!")

    def init_figure(self):
        from mayavi import mlab
        mysky = np.array([135, 206, 250]) / 255.0
        mysky = tuple(mysky.tolist())
        #fig = plt.figure()
        #ax = fig.add_subplot(111, projection='3d')
        #fig = mlab.figure(bgcolor=(1,)*3, size=(1600,1100))
        #fig = mlab.figure(bgcolor=mysky, size=(1600,1100))
        fig = mlab.figure(bgcolor=(0, ) * 3, size=(1600, 1100))
        return fig

    def draw_ocean(self, fig=None):
        from mayavi import mlab
        if fig is None: fig = self.init_figure()
        npts = 100

        #mybrown = np.array([244, 170, 66]) / 255.0
        #mybrown = tuple(mybrown.tolist())
        mywater = np.array([95, 158, 160
                            ]) / 255.0  #(0.0, 0.0, 0.8) [143, 188, 143]
        mywater = tuple(mywater.tolist())
        alpha = 0.3

        # Waterplane box
        x = y = 100 * np.linspace(-1, 1, npts)
        X, Y = np.meshgrid(x, y)
        Z = np.sin(100 * X * Y)  #np.zeros(X.shape)
        #ax.plot_surface(X, Y, Z, alpha=alpha, color=mywater)
        mlab.mesh(X, Y, Z, opacity=alpha, color=mywater, figure=fig)

        # Sea floor
        Z = -self.params['water_depth'] * np.ones(X.shape)
        #ax.plot_surface(10*X, 10*Y, Z, alpha=1.0, color=mybrown)
        #mlab.mesh(10*X,10*Y,Z, opacity=1.0, color=mybrown, figure=fig)

        # Sides
        #x = 500 * np.linspace(-1, 1, npts)
        #z = self.params['water_depth'] * np.linspace(-1, 0, npts)
        #X,Z = np.meshgrid(x,z)
        #Y = x.max()*np.ones(Z.shape)
        ##ax.plot_surface(X, Y, Z, alpha=alpha, color=mywater)
        #mlab.mesh(X,Y,Z, opacity=alpha, color=mywater, figure=fig)
        #mlab.mesh(X,-Y,Z, opacity=alpha, color=mywater, figure=fig)
        #mlab.mesh(Y,X,Z, opacity=alpha, color=mywater, figure=fig)
        ##mlab.mesh(-Y,X,Z, opacity=alpha, color=mywater, figure=fig)

    def draw_mooring(self, fig, mooring):
        from mayavi import mlab
        mybrown = np.array([244, 170, 66]) / 255.0
        mybrown = tuple(mybrown.tolist())
        npts = 100

        # Sea floor
        r = np.linspace(0, self.params['anchor_radius'], npts)
        th = np.linspace(0, 2 * np.pi, npts)
        R, TH = np.meshgrid(r, th)
        X = R * np.cos(TH)
        Y = R * np.sin(TH)
        Z = -self.params['water_depth'] * np.ones(X.shape)
        #ax.plot_surface(X, Y, Z, alpha=1.0, color=mybrown)
        mlab.mesh(X, Y, Z, opacity=1.0, color=mybrown, figure=fig)
        cmoor = (0, 0.8, 0)
        nlines = int(self.params['number_of_mooring_connections'] *
                     self.params['mooring_lines_per_connection'])
        for k in xrange(nlines):
            #ax.plot(mooring[k,:,0], mooring[k,:,1], mooring[k,:,2], 'k', lw=2)
            mlab.plot3d(mooring[k, :, 0],
                        mooring[k, :, 1],
                        mooring[k, :, 2],
                        color=cmoor,
                        tube_radius=0.5 * self.params['mooring_diameter'],
                        figure=fig)

    def draw_pontoons(self, fig, truss, R, freeboard):
        from mayavi import mlab
        nE = truss.shape[0]
        c = (0.5, 0, 0)
        for k in xrange(nE):
            if np.any(truss[k, 2, :] > freeboard): continue
            mlab.plot3d(truss[k, 0, :],
                        truss[k, 1, :],
                        truss[k, 2, :],
                        color=c,
                        tube_radius=R,
                        figure=fig)

    def draw_column(self,
                    fig,
                    centerline,
                    freeboard,
                    h_section,
                    r_nodes,
                    spacingVec=None,
                    ckIn=None):
        from mayavi import mlab
        npts = 20

        nsection = h_section.size
        z_nodes = np.flipud(freeboard -
                            np.r_[0.0, np.cumsum(np.flipud(h_section))])

        th = np.linspace(0, 2 * np.pi, npts)
        for k in xrange(nsection):
            rk = np.linspace(r_nodes[k], r_nodes[k + 1], npts)
            z = np.linspace(z_nodes[k], z_nodes[k + 1], npts)
            R, TH = np.meshgrid(rk, th)
            Z, _ = np.meshgrid(z, th)
            X = R * np.cos(TH) + centerline[0]
            Y = R * np.sin(TH) + centerline[1]

            # Draw parameters
            if ckIn is None:
                ck = (0.6, ) * 3 if np.mod(k, 2) == 0 else (0.4, ) * 3
            else:
                ck = ckIn
            #ax.plot_surface(X, Y, Z, alpha=0.5, color=ck)
            mlab.mesh(X, Y, Z, opacity=0.7, color=ck, figure=fig)

            if spacingVec is None: continue

            z = z_nodes[k] + spacingVec[k]
            while z < z_nodes[k + 1]:
                rk = np.interp(z, z_nodes[k:], r_nodes[k:])
                #ax.plot(rk*np.cos(th), rk*np.sin(th), z*np.ones(th.shape), 'r', lw=0.25)
                mlab.plot3d(rk * np.cos(th) + centerline[0],
                            rk * np.sin(th) + centerline[1],
                            z * np.ones(th.shape),
                            color=(0.5, 0, 0),
                            figure=fig)
                z += spacingVec[k]
                '''
                # Web
                r   = np.linspace(rk - self.params['stiffener_web_height'][k], rk, npts)
                R, TH = np.meshgrid(r, th)
                Z, _  = np.meshgrid(z, th)
                X = R*np.cos(TH)
                Y = R*np.sin(TH)
                ax.plot_surface(X, Y, Z, alpha=0.7, color='r')

                # Flange
                r = r[0]
                h = np.linspace(0, self.params['stiffener_flange_width'][k], npts)
                zflange = z + h - 0.5*self.params['stiffener_flange_width'][k]
                R, TH = np.meshgrid(r, th)
                Z, _  = np.meshgrid(zflange, th)
                X = R*np.cos(TH)
                Y = R*np.sin(TH)
                ax.plot_surface(X, Y, Z, alpha=0.7, color='r')
                '''

    def draw_ballast(self, fig, centerline, freeboard, h_section, r_nodes,
                     h_perm, h_water):
        from mayavi import mlab
        npts = 40
        th = np.linspace(0, 2 * np.pi, npts)
        z_nodes = np.flipud(freeboard -
                            np.r_[0.0, np.cumsum(np.flipud(h_section))])

        # Permanent ballast
        z_perm = z_nodes[0] + np.linspace(0, h_perm, npts)
        r_perm = np.interp(z_perm, z_nodes, r_nodes)
        R, TH = np.meshgrid(r_perm, th)
        Z, _ = np.meshgrid(z_perm, th)
        X = R * np.cos(TH) + centerline[0]
        Y = R * np.sin(TH) + centerline[1]
        ck = np.array([122, 85, 33]) / 255.0
        ck = tuple(ck.tolist())
        mlab.mesh(X, Y, Z, color=ck, figure=fig)

        # Water ballast
        z_water = z_perm[-1] + np.linspace(0, h_water, npts)
        r_water = np.interp(z_water, z_nodes, r_nodes)
        R, TH = np.meshgrid(r_water, th)
        Z, _ = np.meshgrid(z_water, th)
        X = R * np.cos(TH) + centerline[0]
        Y = R * np.sin(TH) + centerline[1]
        ck = (0.0, 0.1, 0.8)  # Dark blue
        mlab.mesh(X, Y, Z, color=ck, figure=fig)

    def draw_buoyancy_tank(self, fig, centerline, freeboard, h_section, loc,
                           r_box, h_box):
        from mayavi import mlab
        npts = 20
        z_nodes = np.flipud(freeboard -
                            np.r_[0.0, np.cumsum(np.flipud(h_section))])
        z_lower = loc * (z_nodes[-1] - z_nodes[0]) + z_nodes[0]

        # Lower and Upper surfaces
        r = np.linspace(0, r_box, npts)
        th = np.linspace(0, 2 * np.pi, npts)
        R, TH = np.meshgrid(r, th)
        X = R * np.cos(TH) + centerline[0]
        Y = R * np.sin(TH) + centerline[1]
        Z = z_lower * np.ones(X.shape)
        ck = (0.9, ) * 3
        mlab.mesh(X, Y, Z, opacity=0.5, color=ck, figure=fig)
        Z += h_box
        mlab.mesh(X, Y, Z, opacity=0.5, color=ck, figure=fig)

        # Cylinder part
        z = z_lower + np.linspace(0, h_box, npts)
        Z, TH = np.meshgrid(z, th)
        R = r_box * np.ones(Z.shape)
        X = R * np.cos(TH) + centerline[0]
        Y = R * np.sin(TH) + centerline[1]
        mlab.mesh(X, Y, Z, opacity=0.5, color=ck, figure=fig)

    def set_figure(self, fig, fname=None):
        from mayavi import mlab
        #ax.set_aspect('equal')
        #set_axes_equal(ax)
        #ax.autoscale_view(tight=True)
        #ax.set_xlim([-125, 125])
        #ax.set_ylim([-125, 125])
        #ax.set_zlim([-220, 30])
        #plt.axis('off')
        #plt.show()
        #mlab.move([-517.16728532, -87.0711504, 5.60826224], [1.35691603e+01, -2.84217094e-14, -1.06547500e+02])
        #mlab.view(-170.68320804213343, 78.220729198686854, 549.40101471336777, [1.35691603e+01,  0.0, -1.06547500e+02])
        if not fname is None: mlab.savefig(fname, figure=fig)
        mlab.show(stop=True)