M.J. Fuente

Fuzzy control of a neutralization process

This paper studies the control of pH neutralization processes using fuzzy controllers. As the process to be controlled is highly nonlinear the usual PI-type fuzzy controller is not able to control these systems adequately. To solve this problem, based on prior knowledge of the process, the pH neutralization process is divided into several fuzzy regions such as high-gain, medium-gain and low-gain, with an auxiliary variable used to detect the process operation region. Then, a fuzzy logic controller can also be designed using this auxiliary variable as input, giving adequate performance in all regions. This controller has been tested in real-time on a laboratory plant. On-line results show that the designed control system operates the plant in a range of pH values, despite perturbations and variations of the plant parameters, obtaining good performance at the desired working points.

Neural networks applied to fault detection of a biotechnological process

Abstract This paper deals with a new method for the detection and diagnosis of faults when they develop in different parts of a wastewater treatment plant situated in Manresa, Spain. When a fault occurs, estimates of parameters in a non-linear mathematical model of the plant change. In this paper, a new approach to fault detection and diagnosis is presented and assessed by using real data from experiments on the plant. The method is based on the use of an on-line estimation technique and a backpropagation neural network to analyse the frequency content of some fault-indicating signal derived from the identification step, allowing correct fault detection, diagnosis and isolation.

Fault tolerance in the framework of support vector machines based model predictive control

Model based predictive control (MBPC) has been extensively investigated and is widely used in industry. Besides this, interest in non-linear systems has motivated the development of MBPC formulations for non-linear systems. Moreover, the importance of security and reliability in industrial processes is in the origin of the fault tolerant strategies developed in the last two decades. In this paper a MBPC based on support vector machines (SVM) able to cope with faults in the plant itself is presented. The fault tolerant capability is achieved by means of the accurate on-line support vector regression (AOSVR) which is capable of training an SVM in an incremental way. Thanks to AOSVR is possible to train a plant model when a fault is detected and to change the nominal model by the new one, that models the faulty plant. Results obtained under simulation are presented.

Neural-Networks-Based Fault Detection and Accommodation in a Chemical Reactor

Abstract This paper discusses how recurrent neural networks can be successfully used for the modelling, constrained multivariable predictive control and fault detection of a non-linear dynamic process. A chemical reactor is modelled via recurrent neural networks. This model is used to build a predictive controller and to detect and to accommodate sensor failures without physical redundancy. The residuals based on this model and the implementation of the incidence matrix detect and diagnosis the fault, whereas the output of the ith neural model accommodates for the failure by replacing the signal from the failed ith sensor with its estimate.

Fault detection and identification method based on multivariate statistical techniques

Multivariate statistical methods such as principal component analysis (PCA) and partial least squares (PLS) have been widely applied to the statistical process monitoring and their effectiveness for fault detection is well recognized, but they have a drawback: the fault diagnosis. In this paper a new method to detect and diagnosis faults is proposed that is composed of two parts: first the PLS method is used for detecting faults and the Fishers discriminant analysis (FDA) is used for diagnosing the faults. FDA provides an optimal lower dimensional representation in terms of discriminating between classes of data, where, in this context of fault diagnosis, each class corresponds to data collected during a specific, known fault. A real plant is used to demonstrate the performance of the proposed method.

CHECKING ORTHOGONAL TRANSFORMATIONS AND GENETIC ALGORITHMS FOR SELECTION OF FUZZY RULES BASED ON INTERPRETABILITY-ACCURACY CONCEPTS

Fuzzy modeling is one of the most known and used techniques in different areas to model the behavior of systems and processes. In most cases, as in data-driven fuzzy modeling, these fuzzy models reach a high performance from the point of view of accuracy, but from other points of view, such as complexity or interpretability, they can present a poor performance. Several approaches are found in the bibliography to reduce the complexity and improve the interpretability of the fuzzy models. In this paper, a post-processing approach is carried out via rule selection, whose aim is to choose the most relevant rules for working together on the well-known accuracy-interpretability trade-off. The rule relevancy is based on Orthogonal Transformations, such as the SVD-QR rank revealing approach, the P-QR and OLS transformations. Rule selection is carried out using a genetic algorithm that takes into account the information obtained by the Orthogonal Transformations. The main objective is to check the true significance, drawbacks and advantages of the rule selection based on the orthogonal transformations via the rule firing strength matrix. In order to carry out this aim, a neuro-fuzzy system, FasArt (Fuzzy Adaptive System ART based), and several case studies, data sets from the KEEL Project Repository, are used to tune and check this selection of rules based on orthogonal transformations, genetic selection and accuracy-interpretability trade-off. This neuro-fuzzy system generates Mamdani fuzzy rule based systems (FRBSs), in an approximative way. NSGA-II is the MOEA tool used to tune the proposed rule selection.

Fault detection in a real wastewater plant using parameter-estimation techniques

Abstract This report deals with the detection and diagnosis of faults when they develop in different parts of a wastewater plant, situated in Manresa, Spain. When a fault occurs, estimates of the parameters in a non-linear mathematical model of the plant change. In this paper, three methods for detecting and tracking changes in different physical parameters are compared and assessed by using real data from experiments on the wastewater treatment plant.

Fault detection in a wastewater treatment plant based on neural networks and PCA

In this paper, a neural network PCA method that integrates neural networks (NN) and principal component analysis (PCA) is used to detect faults in a wastewater treatment plant. The neural networks are used to calculate a non-linear and dynamic model of the process in normal operating conditions. PCA is used to generate monitoring charts based on the residuals calculated as the difference between the process measurements and the output of the networks. It can evaluate the current performance of the process and detects the faults. This technique has been applied to the simulation of a benchmark of a biological wastewater treatment process, a highly non-linear process. The simulation results clearly show the advantages of using this NNPCA monitoring in comparison with classical PCA monitoring.

Fault diagnosis in a wastewater treatment plant using dynamic Independent Component Analysis

In this paper Dynamic Independent Component Analysis (DICA) is used for detecting and diagnosing faults in a wastewater treatment plant. ICA is a method in which the goal is to decompose a set of multivariate data into a base of statistically independent components without loss of information. The DICA monitoring method applies ICA to the augmenting matrix with time-lagged variables. The basic idea of this approach is to use DICA to extract the essential independent components that drive the process in each situation, i.e, in normal operation and in different faulty situations, and to combine them with process monitoring techniques as I2, I2e and SPE for fault detection and diagnosis. The considered charts are compared with their minimum thresholds, with and without faults. The only one that does not violate its threshold tells us the actual system situation, i.e., identifies the fault. This method is applied to the simulation of a benchmark of the biological wastewater treatment process, which is characterized by a variety of fault sources with non-Gaussian characteristics. The simulation results clearly show the advantages of DICA monitoring in comparison with DPCA monitoring.

Support Vector Based Novelty Detection for Fault Tolerant Control

This article presents a method for achieving fault tolerant control in the frame of the nonlinear IMC control scheme. The method consists on the minimization of a cost function in order to compute new controller parameters values. In addition to this, a support vector based novelty detector is used to detect faults in the plant. Simulation results for a continuous stirred tank reactor are presented.