Reconstruction of cylinder pressure using crankshaft speed fluctuations

In this work a Recurrent Neural Network (RNN) is proposed for cylinder pressure reconstruction using crankshaft speed fluctuations. It is shown that single RNN is not capable of reconstructing pressure for complete operating domain of engine distributed over load torque and engine speed. The capability of RNN to estimate cylinder pressure with varying speed and load is enhanced by applying the interpolation on weight parameters. For, this a separate RNN is trained for each specific load and speed of the engine and the weights of RNN are mapped over operating domain of engine. The model is validated on a test rig consisting of single-cylinder engine coupled with eddy current dynamometer. It is shown that the method has potential to estimate cylinder pressure estimation which can be used for future engine controls and diagnostic purpose.