Engine Vibration Anomaly Detection in Vessel Engine Room

Unsupervised and autonomous operations require new, more efficient, and smarter technical solutions. Digitalization is one of the enablers of autonomous operation, with installing new sensors and using the data for AI and machine learning. The trend is to move the handling the vast amount of data from sensors, installed in machines, and devices, close to the sensors. This way, data communication can be heavily reduced, while the latency of decisions based on data can be minimal. This is achieved by moving the advanced data analysis capabilities from the cloud to the edge. In this paper, the driver of technological development is the concept of an unmanned machine room, where the normal inspections performed by personnel are automated using sensors and algorithms. We present an analysis of vibration data, recorded at a cruise ferry’s engines, using a signal processing method called cepstrums. A cepstrum is a version of frequency analysis, intending to identify and visualize periodic structures in data using cepstrograms. Compared to spectrogram analysis, usage of cepstrogram enables better visualization of engine behavior into the shape of audio signals. This helped us discover the run-time vibration routine of the engines from where we could observe unusual vibration sensing generated from the engine. We started studying the unusual behavior and came to the conclusion that the engine is sometimes misfiring the cylinders, generating anomalies.