Computational Intelligence for Automation of Industrial Processes

Abstract

Automation of industrial processes plays a vital role in the present scenario. Monitoring and Control of such processes are significant because there are many real-time challenges that are encountered. In this chapter, broad investigative studies are carried out for four different industrial processes. The first category of industrial application is about the Fisher's Linear Discriminant Analysis (FLDA) and Radial Basis Function Network method used for identifying an approximate temperature of the boiler flame. The image is acquired from the video camera fixed to the inner portion of the boiler. The features from the flame image serve as the basis for flame temperature measurement. The features are compared with a template of information stored in the database to identify the temperature of the flame. The accuracy of the temperature identification is ±10% of the expected flame temperature. The second application is to design a Fuzzy Logic Controller (FLC) for a level process. Automatic tuning of the parameters is carried out to ensure optimal output without compromising the performance under disturbances. The traditional Proportional, Integral and Derivative (PID) controller is also incorporated to control the level process. The result for implementing the conventional PID controllers is used as a standard to validate the proposed FLC controller. Fuzzy logic based self-tuning controller is used for controlling the level process and it does not require a mathematical model or a difficult algorithm. A fuzzy inference mechanism is responsible for changing the variable parameters of the controller with respect to the disturbances ensuring optimum performance from the controller. The third industrial application is about the traditional manufacturing sectors in the field of Electronics and Electrical, implementing a scheme which is a combination of automated Image Processing (IP) algorithms, Artificial Intelligence(AI) techniques and human visual inspection to guarantee the product quality. Nowadays, Machine vision techniques that are used for monitoring and control of industrial applications have become a part of automation. An automated system is one in which high resolution cameras are used to ensure the quality check so as to make specific and instantaneous decisions. Design and development of such schemes are challenging and it is made possible using object recognition algorithms and AI classifier. Here, the chief objective is to identify an effective IP algorithm and AI classifier to investigate the industrial parameters. The fourth category of industrial process discussed here is a non-contact, investigative method using indigenous deep learning algorithm which became initially known to the researchers in the year 1991. This method is effectively used to detect the flaws during the fabrication of steel sheets which takes place inside a furnace at very high temperature. The infra-red camera is used to capture the images during the processing inside the furnace. To detect the flaws, the images of the steel sheets must be divided into segments to locate the exact region. It helps to map and measure their thickness. These measurements help with the diagnosis and provide proper guidance for identification of flaws. In this work, the Deep Learning technique which uses Recurrent Neural Networks (RNN) for segmentation of images is proposed. The proposed method is inferred to be a suitable method to detect the presence of flaws with high precision.