Expert system for assessment of technical condition of electric centrifugal pump assemblies based on productive presentation of knowledge and fuzzy logic

Introduction. Condition assessment of the electric submersible pump units is one of the many tasks that need to be solved to increase the efficiency of oil production business processes. Improving the efficiency of condition monitoring processes and failure prediction of electric submersible pumps often requires the development of special mathematical and software engineering tools. Aim. To research the use of knowledge-based expert systems for technical condition assessment of the electric submersible pumps units. Materials and methods. The expert system is considered as an auxiliary tool that minimizes errors associated with the false positive errors of predictive analytics. The expert system, based on trends for each indicator of the electric submersible pump (pressure, vibration, amperage, etc.), makes an assessment of the technical condition, for example, diagnosing a certain type of malfunction. To store knowledge in the expert system, frames and production rules are considered. The production rules knowledge representation is considered in detail and the possibility of using fuzzy inference is proposed. Results. The application of knowledge-based expert systems, including fuzzy knowledge representation models and logical inference algorithms, is investigated. The use of a production-based rules for representing expert’s knowledge is justified, the application of the fuzzy inference is shown. The concept of an intelligent information system is proposed, which includes an expert knowledge-based decision support system, as well as a preliminary and deep data processing unit, including a component of predictive analytics based on neural network technologies. A demonstration example of the expert system application is presented, and the features of its implementation in the FuzzyCLIPS shell are also considered. Conclusion. The methods and models under study were tested on real data, which confirms the possibility of their use in the development of an intelligent information system.