Advanced Analytics and Diagnostic Rules Automatically Notify Operators About Developing Failures in Rotating and Reciprocating Machines

Abstract

With the paradigm shift towards digitalization, Operators and service providers are inclined to use technologies that can optimize efforts from workforce by providing meaningful information rather than ‘just’ data, transition subject matter knowledge into machines rather than limiting to people, deploy machine learning techniques to improve systems and leverage this big data to serve on wide scale. Historically, condition monitoring knowledge has primarily been people-centric and Reliability personnel have to spend hours in front of screen reviewing terabytes of data. Unfortunately, most of the time is spent to find problems rather than finding solutions. Need of the hour is to define automated mechanisms for triggering alerts pointing towards developing malfunctions for which systems are created with embedded knowledge to run the data through pre-configured diagnostic rules and analytics. Through these online systems, operators are able to receive meaningful actionable information about the issue and its source. These analytics are widespread across machinery, auxiliary and process domains. Through this automated diagnostics platform, Data-driven insights can be generated for machine condition monitoring through advanced rule-building and data-mapping capabilities. In addition to packaged algorithms of known failure signatures, users can also create custom rules that help to capture, disseminate, and leverage knowledge of equipment, processes, and business solutions. For turbomachinery, trending of process parameters, bearing temperature and overall vibration have been used for decades to monitor condition of assets, whereas knowledgeable diagnostic personnel are required to review dynamic data like orbit shape, vibration precession, along with other attributes together to really monitor condition of machine. Now meaningful information from dynamic data can be digitized and attributes can be used in rule logics for automated diagnosis of typical malfunctions like unbalance, misalignment, rubbing, fluid induced instability, rotor bow etc. For reciprocating compressors, automated diagnosis of typical malfunctions like pressure packing leak, valve failures, crosshead pin / frame overloading, debris/liquid ingestion, auxiliary systems (lube oil, cylinder cooling system, unloader etc.) failures and several process related issues can be realized. In this paper, case studies will be demonstrated where users were able to capitalize these systems to identify some of above stated malfunctions and save their assets from expensive secondary repercussions. An operational analytics software will be demonstrated in detail with elaboration on built-in library of pre-packaged algorithms. A primary consideration is maximizing return-on-investment and minimizing payback period. Through use case studies, it will be further demonstrated on how the users were able to identify anomalies and relish 100% payback in less than 2 months of deployment.