The Characteristics of Instantaneous Angular Speed of Diesel Engines for Fault Diagnosis

Early fault detection and diagnosis of diesel engines are paramount now, especially with countries like the UK and France, in-line with the 2015 Paris agreement on climate change, making plans to ban the use of an automobile with diesel and petrol engines before the year 2040. This ban could affect other sectors where diesel engines are the prime mover and result in more stringent exhaust emission regulations. The instantaneous angular speed (IAS) model based fault diagnosis has shown more prospects of fault detection and location. However, there are serious gaps in available knowledge regarding IAS model based fault diagnosis which takes into account the effect of the system’s modal properties. Hence, this research focuses on the online modal properties identification of a typical engine-load system for an improved performance of IAS based fault diagnosis.

Having acknowledged the essentials of IAS based fault diagnosis techniques through a comprehensive literature study, this research firstly investigates the impact of modal properties on the IAS of a four-cylinder engine. This is achieved through a three degree of freedom (DOF) torsional vibration model of the engine-load system, which allows for the modal properties of the system to be calculated and analysed. The calculated modal properties of the system showed one rigid and two flexible modes which had a low (<13Hz) and high (<92Hz) frequencies. The mode shape of the low frequency resonance shows more amplitude on the flywheel-load reference point of the system while that of the high frequency resonance shows more amplitude on the engine-flywheel reference point of the system. It then simulated the IAS which represents the torsional vibration signature with altered modal properties. The simulated result demonstrated that the low frequency resonance is more sensitive to the peak and trough values of the IAS waveform.

After identifying the deployment merits of operational modal analysis (OMA) techniques through a comprehensive literature study, this research then explore the prospect of an IAS based output-only modal properties identification of a typical engine-load system. This was done through both experimental and simulation evaluations, which allowed simulated and experimental IAS to be used for implementing covariance-driven reference based stochastic subspace identification (SSI). The simulated result using pseudo-random input shows that the identified resonance frequencies and mode shapes are 80% correlated with the calculated ones. The simulation results also demonstrated that the accuracy of the identified modal properties is dependent on the number of IAS responses used for implementing the covariance-driven reference based SSI technique. The experimental result using estimated IAS during engine shutdown operation showed that both high and low frequency vibration mode can be identified. The identified resonance frequencies with their mode shapes are 80% correlated with the predicted ones.

Having identified the modal properties of the engine-load system online through the implementation of an IAS based covariance-driven SSI, this research then investigates the impact of misfire on the system’s modal properties especially the mode shape of the low frequency resonance. This was achieved experimentally by inducing a complete misfire in respective cylinders (1st, 3rd and 4th) and the IAS estimated during engine’s transient shutdown operation was used for implementing a covariance-driven reference based stochastic subspace modal properties identification. While the mode shape of the identified high frequency resonance (<80Hz) showed no characteristics for cylinder misfire detection, that of the low frequency resonance (<13Hz) did.

Faults in the engine’s injection system and an abnormal clearance valve train conditions significantly affects its combustion process. The cylinder by cylinder pressure torque obtained from measured IAS through order domain deconvolution technique can be used to detect and diagnose injection faults. In the interim, this research has also recognised that the closer the low-resonance frequency of the model used for the order domain deconvolution gets to its real time value the more accurate the pressure torque becomes. The reconstructed pressure torque which takes into consideration the real time low frequency resonance can be used to detect faulty injection system with different severities and abnormal clearance valve conditions of several severities. Furthermore, the importance of an accurate modal properties utilisation in IAS model