Applying Automated Model Building to Predictive Maintenance in Oil and Gas

Abstract

Abstract Predictive maintenance has become a major focus for the largest industrial companies because of the value it derives, including reduced downtime, improved efficiency, reduced maintenance costs, and others. Success of predictive maintenance programs is achieved when data, analytics, and subject matter expertise intersect. While data and subject matter expertise are always available, analytics talent is often lacking or facing numerous challenges which hinders the success of predictive maintenance programs. Automated model building (AMB) aims at delivering artificial intelligence to the fingertips of industrial companies and hence ensuring the success of predictive maintenance programs without the need of large data science organizations. The automated model building platform ingests the operational (sensor) and failure/fault data and automatically builds AI models to predict the remaining useful life for the asset. The patented technology behind the platform drives feature engineering and model selection which allows customers to automatically create numerous new variables from the sensor data and tests thousands of different models. The platform will then select the optimal set of variables and the model that will achieve the best performance. The entire process can be performed in a matter of few minutes without the need to know the details of all AI models. The platform also gives details on the selected models, which aids with interpretability. This paper will discuss why automated model building and artificial intelligence are needed to deliver effective, scalable predictive maintenance to the oil and gas industry, as well as specific use cases in which AI-powered automated model building has been applied.