Anomaly diagnosis with multi-type information fusion based on MLP for oil pumps

Abstract As the primary power equipment in the oil transport system, the running status of the oil pump is that such a system is safe and reliable operation. The monitored samples in the oil pump are often from various sensors, in which the key signal features need to be extracted and integrated to enhance the capability of anomaly diagnosis. Owing to the complexity and concealment of running information, it is crucial to explore an effective method for anomaly diagnosis of the oil pump. With the rapid progress of artificial intelligence, an intelligent anomaly diagnosis method is proposed for the oil pump by fusing multi-type information in this paper. First, the canonical correlation analysis (CCA) is employed to extract key signal features from multi-type data information. Then, the multi-layer perceptron (MLP) with one output node is implemented to identify anomalies by fusing various signals of the oil pump. Additionally, the anomaly warning decision is achieved by the sequential probability ratio test (SPRT). Finally, the proposed intelligent anomaly diagnosis of the oil pump is validated in a real-world case, demonstrating the effectiveness for diagnosing and warning anomaly information.