Article Preview
TopIntroduction
The overall process control objectives, such as the quality and the quantity of product, have been left in the hands of human operators in the past. Nowadays, computational intelligence techniques are used to solve several complex problems by extracting expert’s knowledge and developing intelligent and modern supervision systems.
Fuzzy logic has proved to be a powerful tool for decision-making systems, especially expert and pattern classification systems. Fuzzy set theory has been used in some chemical processes. In traditional rule-based approaches, knowledge is encoded in form of antecedent-consequent structure. When new data are encountered, it is matched to the antecedent’s clause of each rule, and those rules where antecedents match a data exactly are fired, establishing the consequent clauses. This process continues until the desired conclusion is reached, or no new rule can be fired. In the past decade, fuzzy logic has proved to be useful for intelligent supervision systems in chemical engineering. Most control situations are more complex than we can deal with mathematically. In this situation, fuzzy control can be developed, providing a body of knowledge about the existing control process, in the form of a number of fuzzy rules.
In the first part of this work, expert’s knowledge was extracted by interviews with operators and manuals. This knowledge is used in control by integrating fuzzy logic in the supervision system to control the industrial process, which is milk production, to resolve problems and replace the old supervision system by a new one that satisfies all needs.
In the second part of this work, the big advantage is to integrate machine learning (ML) using SVM as a binary classification of different measurement coming from sensors to the Programmable Logic Controller (PLC). These measurements are classified after training in two classes. The first one is when the process is in a good situation and the second one indicates that a default or an alarm is occurred.
The advantages of such architecture is its flexibility in control, its ability to data process a lot of information in order to improve the productivity and to reduce maintenance costs. In the second Section, related works concerning fuzzy logic, machine learning and SVM are presented. Third section is dedicated to the case study and the proposed approach. Implementation and results of the developed system are discussed in the fourth section. We conclude and discuss the results in a conclusion. To obtain all objectives and realize the approach we followed steps presented in Figure 1.
Figure 1. Steps of realization of the new approach
Expert Knowledge
Knowledge Management is a set of methods and techniques to collect, identify, analyze, and organize, store and share knowledge among members of the organization to achieve the objectives. In industry Knowledge Management is used in the industrial supervision in diagnostic task to identify the probable cause of the (or) failure (s) with one based on a set of information from a logical reasoning inspection, control or maintenance.
Knowledge-based approaches often rely on monitoring residuals between multiple sensor measurements (Frank, 1990). However, due to the high number of sensors used on modern industrial plants and other complex industrial systems, the adoption of additional redundant sensors is prohibitively expensive (Zhang et al., 2017). Different works analyze the knowledge extraction process for the application using expert and induced knowledge from data collected during navigation tasks (Alonso et al., 2007) and applying extracted knowledge in monitoring and supervision (Zermane et al., 2017a). The main aim is to modernize enterprises that needs to modernize their supervision systems to ensure diagnostic of alarms, control and maintenance, where we can store data and prepare reports. The term “heuristic” refers to knowledge that is acquired by experimentation or trial and error. Expert knowledge is knowledge possessed by human experts about a situation or problem. Conventional model–based controllers, although invaluable in solving complicated problems, cannot utilize expert knowledge. However, it is one of the great strengths of fuzzy control that expert knowledge can be easily incorporated into fuzzy controllers.