Developing an expert prognosis system of the reciprocating compressor based on associations among monitoring parameters and maintenance records

Highlights

• For the demonstration purpose, failure information from published literatures mimics the maintenance history.

• Use of the association analysis to create equivalent functionalities of the IF-Then rule with the certainty value.

• Based on the association rules, establish diagnosis expert system for rotating equipment.

• The diagnosis expert system can quickly infer faults of rotating equipment from single and double parameters.

Abstract

The reciprocating compressor is, in general, a critical equipment in a process plant. For certain ultra-high-pressure process, if the reciprocating compressor fails, often it will cause serious impact to not just the compressor itself, but also the process surrounds it. To prevent compressors from failures, an expert diagnosis system is needed. However, the traditional rule-based expert system is quite inefficient and difficult to create.

For an expert prognosis system that is customized to meet needs of a specific process, one needs to refer to plant maintenance history, which is hard to come by due to the fact that most maintenance was poorly documented. This research attempt to demonstrate the feasibility of developing an expert prognosis system through implementation of association rules. Rather than mining from maintenance history, records of failure cases were collected from technical journal articles by extracting information containing failure symptoms and causes on failed components, that mimicking repair history. In total, 115 failure information out from 41 journal articles were gathered. Applications of this approach to practical use in a process plant is easy by replacing the failure information table with that from datamining the repair history. The failure information was first tabulated and then put through association analysis for support, confidence, and lift between two parameters. The demonstration program has been successful with 1-to-1, many-to-1, and many-to-many analysis among failed components, failure modes, and operation parameters.

Introduction

Manufacturing is the backbone of our modern society (Yao et al., 2019).The petrochemical and chemical industries are one of the most important manufacturing industries. Therefore, their production capacity has always been an important topic. Any abnormal shutdown or catastrophic accident for a factory will not only damage the reputation of the enterprise but also cause a huge economic impact (Jiao et al., 2018). Shu et al. (2016) indicated that the annual loss caused by the abnormal shutdown in the United States was at least 2 billion U.S. dollars.

Reciprocating compressor is critical equipment in the petrochemical industry (Bloch, 2006; Bloch and Hoefner, 1996; Tran et al., 2014). For certain ultra-high-pressure process, if the reciprocating compressor faults, it might cause severe impact to equipment or even to the whole process.(CSB, 2011; Doyle, 1972; Kletz, 1979; Yi and Zhu, 2014). In order to deal with the aforementioned problems, an expert diagnosis system (ES) was developed.

In fact, expert systems have long been applied to handle various process control or safety matters. Mohammad Yazdi et al. (2019) proposed an ingenious approach by combining the Z-number and fuzzy logic to ensure reliable result. Mohammad Yazdi et al. (2019) Used method which combine Z-number and fuzzy logic to improve accuracy result of expert system that apply risk analysis. Sohag Kabir et al. (2019) proposed a framework that combines intuitionistic fuzzy set theory and expert elicitation to enable quantitative analysis of TFTs of dynamic systems with uncertain data. Rich, S. H. et al. (1989). proposed a diagnostic expert system for the whipped toppings process. Wang, F., & Gao, J. (2012). presented a novel knowledge database construction method for an operation guidance expert system based on the HAZOP analysis and the accident analysis. Rahman, S. et al. (2009). developed a based-HAZOP expert system.

In summary, the expert system has been continuously improved and strengthened, and has so far been applied to address process safety related issues.

In the core of the diagnosis engine is a complex logical operation of If … Then … rules that are used to deduce and determine failure categories based on anomaly features entered by the user. An expert system typically comprised of an inference engine, a knowledge base, an explanation facility, and a user interface (Kadhim et al., 2013; Lamatsch et al., 1988; Tripathi, 2011).

Among them, the inference engine and knowledge base were structured in a rule-based format at that time, and thus, were also referred to as the “rule-based expert system."(Haen et al., 2012).First of all, knowledges have to be expressed as sets of IF-THEN rules (Zhou, 2010). In order to perform diagnosis that leads symptoms to potential failures, all relevant rules need to be linked through an exhaustive search in the rule base and then allow the inference engine to deduce and diagnose (Noh et al., 2012). In practice, a great percentage of the information required to create rules came from expert experience (Alonso et al., 2012), or maintenance manuals (Cowan, 2001; Haen et al., 2012), which however poses some inherent constraints in rule base development due to the lack of comprehensive experience and information. Knowledge from the maintenance manual are composed of individual troubleshooting guidance and are often insufficient for fault diagnosis. A successful implementation of the expert system needs to address the following issues: 1. unclear relationships and links among diagnostic rules (Pepper, 1987), 2. difficult to achieve automated learning (Noh et al., 2012), 3.computation inefficient and time consuming for exhaustive search to create linkage among rules (Negnevitsky, 2011), 4. tough for system maintenance, and 5. inference errors caused by incorrect interpretation and rule forming from empirical knowledge.

In recent years, on the basis of mature data science and technology, numerous research achievement which used big data algorithms to establish data-driven fault diagnosis be proposed (Ling et al., 2003; Qiu et al., 2005; Shu et al., 2016; Song and Qi, 2012; Verron et al., 2010; Z. Zhang et al., 2013). Aforementioned literatures proposed various approaches for inferring equipment faults by integrating big data technology and historical records, which, in practice, are similar to functionalities of the expert system. As a typical data mining method, association rules can function as classification rules except that association rules can be used to predict any attribute and combinations of attributes as well. Therefore, it is one of the most commonly used algorithms and can be applied to determine what records commonly occur close together.

Market basket analysis is the most classic application of association rules, followed by stock analysis, medical diagnosis, and customer market analysis (Solanki and Patel, 2015). As this method can establish correlation between data, applications have been created base on the enormous amount of medical history in the medical community to study correlations between disease symptoms and characteristics. Ordonez et al. (2000) applied the association rule to determine the correlation between considerable medical data, such as basic patient information and medical records, and took the correlation as the basis of diagnosis. Nahar et al. (2013) adopted association rule mining in the UCI Cleveland dataset to analyze information about patients and healthy people and used the confidence as an indicator to determine the factors that led to heart diseases in men and women. Huang et al. (2019) proposed a model based on association rule mining (ARM), which was used to discover the association rule of chronic diseases from the data that were continuously collected during the medical examination and medical treatment process. In addition, Yairi et al. (2001) combine the time-series data, pattern clustering, ARM, and other methods to detect fault of spacecraft.

In various aspects with successful applications, the method derived from the association rules have the following characteristics 1. Correlations between features can be drawn from data. 2. It can establish causal relationships between multiple features and abnormalities. 3. Relationships can be presented quantitatively. 4. This approach is highly fault tolerant. In addition, Mazid et al. (2009) concluded that for classification rules, accuracy of results from association rule can be significantly greater than that from rule-based approaches.

The aim of this research is to develop an expert system based on association rules to reinforce the diagnosis quality and eliminate deficiencies conventional rule-based expert systems suffered. The goal is to establish a many-to-many diagnostic rules for predicting potential failures prognostically.

The main contributions of this paper are as follows: 1. This system have the same inferring functionalities of the IF-THEN rules. 2. Limitations of traditional expert systems has been lifted. 3. The expert system can quickly predict possible failure modes, failure location, and failure cause of the equipment through process parameter. Goal of the expert diagnosis system is shown in Fig. 1. Maintenance staff can conduct maintenance with the prognosis results from the expert system.