Real-Time Classification of Diesel Marine Engine Loads Using Machine Learning

Abstract

An engine control system is responsible for controlling the combustion parameters of an internal combustion engine to increase the efficiency of the engine. An optimized parameter setting of an engine control system is highly influenced by the engine load. Therefore, with a change in engine load, the parameter settings need to be updated for higher engine efficiency. Hence, to optimize parameter settings during operation, engine load information is necessary. In this paper, we propose a real-time engine load classification from sensed signals. For the classification, an artificial neural network is used and trained using processed, real, measured data. To that end, a magnetic pickup sensor extracts the rotational speed of the prime mover of a four-stroke V12 marine diesel engine. The measured signal is then converted into a crank angle degree (CAD) signal that shows the behavior of the combustion strokes of firing cylinders at a particular engine load. The CAD signals are considered an input feature to the designed network for classification of engine loads. For verification, we considered five classes of engine load, and the trained network classifies these classes with an accuracy of 99.4%.