This data repository supports the in review publication titled "Predicting Transfer Learning Suitability in ANN-based Control of FOPDT Industrial Processes" part of the E-TROTLE project. This project is funded under MCIN/AEI/10.13039 /501100011033 project TED2021-129134B-I00 co-funded with the European Union ‘‘NextGenerationEU’’/PRTR funds. ETROTLE aims to develop data-driven control systems using transfer learning and optimization methods to enhance the operation of wastewater treatment plants (WWTPs), focusing on environmental impact and sustainability. This repository contains the data used in the obtention of the transferability metrics, which entails the ANN models and simulations, pid parameters and scalers.

This dataset provides the used data for obtaining the results presented in our publication. It includes MATLAB simulation files for PID tuning and control actuation, model weight dictionaries for baseline Artificial Neural Networks (ANNs) used for comparative analysis, MATLAB source code and models, and data scalers.

METHODOLOGICAL INFORMATION

1. Description of methods used for collection-generation of data: SIMULINK simulation jointly with our tuning functions allowed for selection of the PID parameters and obtention of the simulation vectors. Python for obtention of .h5 ANN models and posterior analysis of generated .csv with the data.

2. Instrument- or software- specific information needed to interpret the data: We used MATLAB_2023a for the simulations and the usort2 tuning rules for the PIDs based on the process sets. Scheduled sampling framework for training the .h5 LSTM_FF networks.

ETROTLE_Dataset_TSM25_MAT_files: Contains necessary MATLAB data objects to run source MATLAB files for data generation. ETROTLE_Dataset_TSM25_models: Includes the ANN models trained for the 10 FOPDT and 5 test FOPDT control processes that will be compared in terms of transferability. ETROTLE_Dataset_TSM25_Scalers_5i_10fopdt and ETROTLE_Dataset_TSM25_Scalers_5i_extra5: Includes the scalers for the ANN data inputs and outputs used in training for all first-order processes. The scaling values are based on historical data, located in ETROTLE_Dataset_MAT_files under the 1st_order directory. .m and .slx in the main directory: in order to obtain data for TSM_BOOST and TSM_DELTAS. Also to directly calculate TSM_COR and TSM_MI.