Optimum parameters for fault detection and diagnosis system of batch reaction using multiple neural networks (original) (raw)

ARTIFICIAL NEURAL NETWORKS FOR FAULT DETECTION AND DIAGNOSIS IN A PROCESS PLANT

Artificial Neural Networks (ANNs) consist of simple processing elements or nodes that work in unison to solve specific problem. Information processing is achieved through activation and inhibition of the interconnections among the units in the network. The effective computational ability of the Feed-forward neural networks based on the back-propagation algorithm was used to develop a neural network computer program for fault detection and diagnosis in complex chemical plants. Fault detection and diagnosis are mapped as a pattern recognition and classification problem. The catalytic dehydrogenation of heptanes to toluene was simulated as a case study to illustrate the great potential of the program developed. A three-layered network was used in this case study, the network was made up of 18 units in the input layer, 5 units in the first hidden layer, 7 units in the second hidden layer and 6 units in the output layer. Data abstracted from sensor measurements for normal and faulty operating conditions was used to train the network. After training the network for 8,000 epochs, the network learning was seen to be stable and was able to recognize and classify all patterns for the fault scenarios presented with accuracy above 99%.

Neural Networks for Process Monitoring, Control and Fault Detection: Application to Tennessee Eastman Plant

2001

This paper discusses the application of artificial neural networks in the area of process monitoring, process control and fault detection. Since chemical process plants are getting more complex and complicated, the need of schemes that can improve process operations is highly demanded. Artificial neural network can provide a generic, non-linear solution, and dynamic relationship between cause and effect variables for complex and non-linear processes. This paper will describe the application of neural network for monitoring reactor temperature, estimation and inferential control of a fatty acid composition in a palm oil fractionation process and detection of reactor sensor failures in the Tennessee Eastman Plant (TEP). The potential for the application of neural network technology in the process industries is great. Its ability to capture and model process dynamics and severe process non-linearities makes it powerful tools for process monitoring, control and fault detection.

Detection of multiple sensor faults in a palm oil fractionation plant using artificial neural network

2002

Given the universal approximation properties, simplicity as well its intrinsic analogy to the non-linear state space form, a recurrent Elman network is derived and applied as process predictor for fault detection in process plants. In this paper, a two-stage scheme integrating a neural Elman network dynamic predictor and a feedforward neural network fault classifier is proposed to overcome the problem of multiple sensor faults. The scheme was implemented to detect sensor failures in a palm oil fractionation process. To generate the required simulation data, Hysys.Plant dynamic process simulator was employed. The use of the output prediction error, between a neural network model and a non-linear dynamic process, as a residual for detecting sensor faults is analysed. A second neural network classifier is developed to detect the sensor faults from the residuals generated, and results are presented to demonstrate the satisfactory detection of two sensor faults achieved simultaneously using this scheme.

A neural-network approach to fault detection and diagnosis in industrial processes

IEEE Transactions on Control Systems Technology, 1997

Using a multilayered feedforward neural-network approach, the detection and diagnosis of faults in industrial processes that requires observing multiple data simultaneously are studied in this paper. The main feature of our approach is that the detection of the faults occurs during transient periods of operation of the process. A two-stage neural network is proposed as the basic structure of the detection system. The first stage of the network detects the dynamic trend of each measurement, and the second stage of the network detects and diagnoses the faults. The potential of this approach is demonstrated in simulation using a model of a continuously well-stirred tank reactor. The neural-network-based method successfully detects and diagnoses pretrained faults during transient periods and can also generalize properly. Finally, a comparison with a model-based method is presented.

Fault Diagnosis In Complex Chemical ProcessesUsing Hierarchical Neural Networks

WIT Transactions on Information and Communication Technologies, 1970

BackPropagation Neural Networks (BPN) with linear activation functions have some shortcomings including long training time, neuron size determination problems with hidden layers, extrapolation problems which reduces their applicability for real time fault diagnosis of complex processes. Two hierarchical diagnostic approaches that are based on hierarchically ordered BPNs and elliptical neural networks are developed to overcome some of these limitations. Their applicability and reliability are tested and compared using a hydrocarbon chlorination plant troubleshooting simulator.

Fault diagnosis for a class of chemical batch processes

This paper deals with the problem of fault diagnosis for a class of chemical reactions taking place in jacketed batch reactors. An integrated diagnosis scheme, using redundant temperature measurements and a bank of state observers, is adopted to detect, isolate and identify faults. A unified framework is developed to take into account faults affecting sensors, actuators and process. In order to test the effectiveness of the approach, a detailed simulation case study is developed.

Fault detection and diagnosis for continuous stirred tank reactor using neural network

Kathmandu University Journal of Science, Engineering and Technology, 2010

The paper focuses on the application of neural network techniques in fault detection and diagnosis. The objective of this paper is to detect and diagnose the faults to a continuous stirred tank reactor (CSTR). Fault detection is performed by using the error signals, where when error signal is zero or nearly zero, the system is in normal condition, and when the fault occurs, error signals should distinctively diverge from zero. The fault diagnosis is performed by identifying the amplitude error of the CSTR output error.