Article


Article Code : 13980507191155

Article Title : An Intelligent Autonomous System for Condition-Based Maintenance- Case Study: Control Valves

Journal Number : 26 Spring 2019

Visited : 206

Files : 701 KB


List of Authors

  Full Name Email Grade Degree Corresponding Author
1 Hamidreza Naseri Hamid.R.Naseri@gmail.com Post Graduate Student PhD
2 Ali Shahidinejad A.shahidinejad@qom-iau.ac.ir Assistant Professor PhD
3 Mostafa Ghobaei-Arani m.ghobaei@qom-iau.ac.ir Assistant Professor PhD

Abstract

Maintenance process generally plays a vital role to achieve more benefits to the enterprises. Undoubtedly, this process has a high value-added in oil and gas industries. Process owner expectations and new technology acquisition have been changing the mindset of domain experts to the new maintenance approaches and different newer methods such as condition-based maintenance models for improving the reliability and decreasing the cost of maintenance. Because of the high dynamic behavior of the gas and the instability of the input parameters, the need to apply a model with self-healing behavior is a serious demand in the gas industry. However, to the best of our knowledge, despite its importance, there is not any comprehensive study in the literature. In this paper, we present a new neuro-fuzzy model and a self-management control loop using real world data to meet the mentioned targets for a specified control valve in a gas refinery. ANFIS model is employed for the reasoning process which has six inputs (Inlet/outlet Pressures, temperature, flow rate, controller output and valve rod displacement), and one output that is a type of failure of the control valve and the most failures are considered based on domain expert knowledge. A suitable control loop is used to unceasingly monitor, analyze, plan and finally execute the process of prediction of failures. Due to undertaken improvement, there is a considerable change in reliability and financial indices. Moreover, the proposed approach is compared with two different methods. The results show that our proposed model comprehensively improves accuracy by 24%.