Companion repository to the paper Enhancing Siamese Neural Networks through Expert Knowledge for Predictive Maintenance.
Please note, that the MS-SNNs approach is here referred to as CaseBasedSimilarity (CBS) and CNN2D + MAR is called cnn2dWithAddInput. The implementation of some components is based on the one presented in NeuralWarp (GitHub).

• The sub directory supplementary_resources of this repository contains additional information about the data sets used and the architecture of the CNN2D + MAR model.
• An overview of all relevant conducted experiments
• The detailed logs for each of those experiments
• The raw data recorded with this simulation factory model used to generate the training and evaluation data sets.
• The preprocessed data set we used for the evaluation.

1. Clone the repository
2. Download the preprocessed data set and move it to the data folder
3. Navigate to the neural_network folder and start the training and test procedure via python TrainAndTest.py > Log.txt

Used python version: 3.6.X
Used packages: See requirements.txt

• All settings can be adjusted in the script Configuration.py, whereby some rarely changed variables are stored in the file config.json, which is read in during the initialization.
• The hyperparameters of the neural networks can be defined in the script Hyperparameter.py or can be imported from a file in configuration/hyperparameter_combinations/ (this can also be changed in the configuration).
• For training, the desired adjustments should first be made at the parts mentioned above and then the training can be started by running Training.py.
• The evaluation of a trained model on the test dataset can be done via Inference.py. To do this, the folder which contains the model files, must first be specified in the configuration.
• For executing the real-time data processing using RealTimeClassification.py first a kafka server must be configured and running. Also the topic names and mappings to prefixes must be set correctly.
• The data/ directory contains all required data. Central are the pre-processed training data in data/training_data/ and the trained models in data/trained_models/. A detailed description of what each directory contains is given in corresponding parts of the configuration file.

The following section gives an overview of the packages, directories and included Python scripts in this repository.

The archive contains currently unused code fragments that could potentially be useful again, old configurations and such.

Used to store the outputs/logs of inference/test runs for future evaluation.

Due to the high amount of different models and configuration options, not all components can be used together. The following table shows the current compatibility status of the different models with the execution variants. Also please note that the real time classification is still under development and may not work currently.