Precision Fault Diagnosis in Screw Compressors Using ANN

Abstract

Screw compressors are critical assets in petroleum and chemical industries, where the ineffectiveness in the functioning may result in the considerable financial losses. The study introduces an effective fault diagnosis model, which is founded on the Artificial Neural Networks (ANNs) technology to distinguish between the malfunctions of the major compressor subsystems: engine, compressor block (CB), cooling system, and oil control circuit. Custodial ANN were trained on large industrial datasets, and attained high classification accuracy: 100% in the engine (3-12-7-3, MSE: 3.47e-17), 100% in the CB (2-13-8-5-3, MSE: 4.76e-18), 99.4% in the cooling system (2-11-6-3, MSE: 6.76e-14), and 100% in The system proved to be capable of working in severe operational conditions, such as high vibration (RMS 5.1 mm/s, gA 6 mm/s2 ) and high temperatures (up to 130 °C ), which was tested extensively. The findings support the argument that the suggested ANN-based solution is a scalable, high-fault diagnosis, and real-time solution, which is significantly superior to the traditional methods in order to minimize maintenance expenses and maximize operational efficiency of industrial screw compressors.