Denoising Autoencoders and LSTM-Based Artificial Neural Networks Data Processing for Its Application to Internal Model Control in Industrial Environments—The Wastewater Treatment Plant Control Case

Abstract

<jats:p>The evolution of industry towards the Industry 4.0 paradigm has become a reality where different data-driven methods are adopted to support industrial processes. One of them corresponds to Artificial Neural Networks (ANNs), which are able to model highly complex and non-linear processes. This motivates their adoption as part of new data-driven based control strategies. The ANN-based Internal Model Controller (ANN-based IMC) is an example which takes advantage of the ANNs characteristics by modelling the direct and inverse relationships of the process under control with them. This approach has been implemented in Wastewater Treatment Plants (WWTP), where results show a significant improvement on control performance metrics with respect to (w.r.t.) the WWTP default control strategy. However, this structure is very sensible to non-desired effects in the measurements—when a real scenario showing noise-corrupted data is considered, the control performance drops. To solve this, a new ANN-based IMC approach is designed with a two-fold objective, improve the control performance and denoise the noise-corrupted measurements to reduce the performance degradation. Results show that the proposed structure improves the control metrics, (the Integrated Absolute Error (IAE) and the Integrated Squared Error (ISE)), around a 21.25% and a 54.64%, respectively.</jats:p>

Industrial Control under Non-Ideal Measurements: Data-Based Signal Processing as an Alternative to Controller Retuning

“… , and have been considered because these variables are involved in the mass balance equation of the concentration, its conversion rate and the biological processes described in [ 36 , 37 ]. In addition, and have also been considered due to the fact that they are two of the variables showing the highest mutual information with respect to (see Figure 5 in [ 25 ]). For more details about the selection of input variables taking into account the mutual information readers are referred to [ 43 ].…”

“…This is an important process due to the fact that the performance of the control strategy is directly related to the measurements quality. For instance, DAEs have been considered in [ 25 ], where the improvement achieved in the control performance is around a 16.84% in average with respect to the situation where DAEs are not adopted. Besides, the control performance is also dependant on the denoising quality.…”

“…Later, they have been adopted either to determine the actuation variable of the control process, or to determine the next state of the system before being considered by a controller. Another example corresponds to [ 25 ], where two LSTM networks have been proposed to control the concentration by means of an Internal Model Controller structure. However, these nets have been designed to track a variable set-point.…”

“…In such a context, one of the common points among model-based controllers and the ANN-based ones is that they are sensible to perturbations of the signals involved in the control process [ 3 , 25 ]. The main perturbations that affect these signals are the noise and delays introduced by the different non-ideal sensors placed over the industrial plants.…”

“…Thus, a high knowledge of the processes being controlled is required at the same time that the control solution will not be generalisable. This could be approached with the denoising and control approaches proposed in [ 25 ], where an ANN-based Internal Model Controller has been proposed to control the dissolved oxygen in the fifth reactor tank of the WWTP plant. Realistic sensors have been considered and as a consequence, the cleaning of noise corrupted measurements has also been proposed.…”