def inference_vibration_route():
fftPoints = request.json['fftPoints']
samplingInterval = request.json['samplingInterval']
scalingCoeff = request.json['accelerationCoeff1']
offsetCoeff = request.json['accelerationCoeff0']
modelId = request.json['modelId']
output = parse_vibration(scalingCoeff, offsetCoeff, fftPoints,
samplingInterval)
xInference = lin_log_interp(np.array(output['fftAmps']))
# if modelId == 'CNN-AE-Lite':
# value = autoencoder_lite(xInference)
# elif modelId == 'CNN-MLP-Lite':
# value = classifier_lite(xInference)
#if modelId == 'PCA-GMM':
# value = model_gmm(xInference)
#elif modelId == 'PCA-GNB':
# value = model_gnb(xInference)
#else:
value = 0.
output['values'] = value
return jsonify(output), 201
def select_fft_features(
self,
minDate=None,
stdev=False,
labeled=True,
sensorId=None,
limit=None,
descending_order=True,
fft_interval=None,
):
if self.debug:
print(self.minQueriedDate)
if fft_interval is not None:
interval_str = "and frequencyInterval" + fft_interval
else:
interval_str = ""
if stdev == True:
table = 'FFTSTD'
else:
table = 'FFT'
if labeled == True:
vibStateSelect = """exists
(SELECT VibrationState.values from """ + self.database + """.VibrationState WHERE VibrationState.dateTime = """ + table + """.dateTime)"""
else:
vibStateSelect = table + """.dateTime < '""" + self.minQueriedDate + """'
and not exists
(SELECT VibrationState.values from """ + self.database + """.VibrationState WHERE VibrationState.dateTime = """ + table + """.dateTime)"""
if sensorId is not None:
sensorId_str = "and " + self.database + ".RMS.sensorId = '" + sensorId + "'"
else:
sensorId_str = ''
if minDate is not None:
minDate_str = "and " + table + ".dateTime > " + minDate
else:
minDate_str = ""
if limit is not None:
limit_str = "limit {}".format(limit)
else:
limit_str = ""
if descending_order == True:
desc_str = "desc"
else:
desc_str = "asc"
query = """SELECT """ + table + """.sensorId,
""" + table + """.dateTime AS dateTime,
""" + table + """.values AS fftVals,
RMS.values AS rmsVals,
VibrationState.values as vibState
FROM """ + self.database + """.""" + table + """
INNER JOIN """ + self.database + """.RMS ON
""" + self.database + """.RMS.dateTime = """ + self.database + """.""" + table + """.dateTime
and """ + self.database + """.RMS.assetId = '""" + self.assetId + """'
""" + sensorId_str + """
LEFT JOIN """ + self.database + """.VibrationState ON
""" + self.database + """.VibrationState.dateTime = """ + self.database + """.""" + table + """.dateTime
where """ + vibStateSelect + """
""" + self.debug_str + """
and """ + table + """.assetId = '""" + self.assetId + """'
""" + minDate_str + """ """ + interval_str + """
order by """ + table + """.dateTime """ + desc_str + " " + limit_str + """; """
print(query)
cursor = self.execute_query(query)
data = cursor.fetchall()
if len(data) <= 0:
return pd.DataFrame(), pd.DataFrame(), pd.DataFrame()
sensorId = np.array([[data[i][0] for i in range(len(data))]]).T
dateTime = np.array([[data[i][1] for i in range(len(data))]]).T
fftVals = np.array(
[ast.literal_eval(data[i][2]) for i in range(len(data))])
rmsVals = np.array([[data[i][3] for i in range(len(data))]]).T
vibState = np.array([[data[i][4] for i in range(len(data))]]).T
columns = ['FFT-{}'.format(i) for i in range(fftVals.shape[1])]
columns = ['dateTime'] + columns + ['RMS', 'sensorId', 'VibState']
fftVals = lin_log_interp(fftVals)
# self.minQueriedDate = np.amin(dateTime).strftime('%Y-%m-%d %H:%M:%S.%f')
statsFeatures = np.hstack(
(dateTime, fftVals, rmsVals, sensorId, vibState))
featuresDF = pd.DataFrame(data=statsFeatures, columns=columns)
featuresDF = featuresDF.set_index('dateTime')
featuresDF.index = pd.to_datetime(featuresDF.index)
cursor.close()
self.disconnect()
sensorIdDF = featuresDF.loc[:, featuresDF.columns == 'sensorId']
vibStateDF = featuresDF.loc[:, featuresDF.columns == 'VibState']
featuresDF = featuresDF.drop(['RMS', 'sensorId', 'VibState'],
axis=1).astype(float)
return featuresDF, sensorIdDF, vibStateDF
def select_ml_features(self, labeled=True, sensorId=None, limit=None):
if self.debug:
print(self.minQueriedDate)
if labeled == True:
vibStateSelect = """exists
(SELECT VibrationState.values from db18.VibrationState WHERE VibrationState.dateTime = FFT.dateTime)"""
else:
vibStateSelect = """FFT.dateTime < '""" + self.minQueriedDate + """'
and not exists
(SELECT VibrationState.values from db18.VibrationState WHERE VibrationState.dateTime = FFT.dateTime)"""
if sensorId is not None:
sensorId_str = "and db18.VibrationSkewness.sensorId = '" + sensorId + "'"
else:
sensorId_str = ''
if limit is not None:
limit_str = "limit {}".format(limit)
else:
limit_str = ""
query = """SELECT FFT.sensorId,
FFT.dateTime AS dateTime,
FFT.values AS fftVals,
RMS.values AS rmsVals,
VibrationMean.values as vibMeanVals,
VibrationSkewness.values as vibSkewVals,
VibrationKurtosis.values as vibKurtVals,
VibrationVariance.values as vibVarVals,
VibrationState.values as vibState
FROM db18.VibrationSkewness
INNER JOIN db18.FFT ON
db18.FFT.dateTime = db18.VibrationSkewness.dateTime
and db18.VibrationSkewness.assetId = '""" + self.assetId + """'
""" + sensorId_str + """
INNER JOIN db18.RMS ON
db18.RMS.dateTime = db18.VibrationSkewness.dateTime
INNER JOIN db18.VibrationMean ON
db18.VibrationMean.dateTime = db18.VibrationSkewness.dateTime
INNER JOIN db18.VibrationKurtosis ON
db18.VibrationKurtosis.dateTime = db18.VibrationSkewness.dateTime
INNER JOIN db18.VibrationVariance ON
db18.VibrationVariance.dateTime = db18.VibrationSkewness.dateTime
LEFT JOIN db18.VibrationState ON
db18.VibrationState.dateTime = db18.VibrationSkewness.dateTime
where """ + vibStateSelect + """
""" + self.debug_str + """
order by FFT.dateTime desc """ + limit_str + """; """
cursor = self.execute_query(query)
data = cursor.fetchall()
if len(data) <= 0:
return pd.DataFrame(), pd.DataFrame(), pd.DataFrame()
sensorId = np.array([[data[i][0] for i in range(len(data))]]).T
dateTime = np.array([[data[i][1] for i in range(len(data))]]).T
fftVals = np.array(
[ast.literal_eval(data[i][2]) for i in range(len(data))])
rmsVals = np.array([[data[i][3] for i in range(len(data))]]).T
vibMeanVals = np.array([[data[i][4] for i in range(len(data))]]).T
vibSkewVals = np.array([[data[i][5] for i in range(len(data))]]).T
vibKurtVals = np.array([[data[i][6] for i in range(len(data))]]).T
vibVarVals = np.array([[data[i][7] for i in range(len(data))]]).T
vibState = np.array([[data[i][8] for i in range(len(data))]]).T
vibKurtVals = np.abs(vibKurtVals)
vibSkewVals = np.abs(vibSkewVals)
fftVals = lin_log_interp(fftVals)
columns = ['FFT-{}'.format(i) for i in range(fftVals.shape[1])]
columns = ['dateTime'] + columns + [
'RMS', 'Mean', 'Skew', 'Kurtosis', 'Variance', 'sensorId',
'VibState'
]
self.minQueriedDate = np.amin(dateTime).strftime(
'%Y-%m-%d %H:%M:%S.%f')
statsFeatures = np.hstack(
(dateTime, fftVals, rmsVals, vibMeanVals, vibSkewVals, vibKurtVals,
vibVarVals, sensorId, vibState))
featuresDF = pd.DataFrame(data=statsFeatures, columns=columns)
featuresDF = featuresDF.set_index('dateTime')
featuresDF.index = pd.to_datetime(featuresDF.index)
cursor.close()
self.disconnect()
sensorIdDF = featuresDF.loc[:, featuresDF.columns == 'sensorId']
vibStateDF = featuresDF.loc[:, featuresDF.columns == 'VibState']
featuresDF = featuresDF.drop(['sensorId', 'VibState'],
axis=1).astype(float)
return featuresDF, sensorIdDF, vibStateDF