(2017) Engine cylinder pressure reconstruction using crank kinematics and recurrently-trained neural networks. Mechanical Systems and Signal Processing, This version is available from Sussex Research Online: http://sro.sussex.ac.uk/63580/ This document is made available in accordance with publisher policies and may differ from the published version or from the version of record. If you wish to cite this item you are advised to consult the publisher's version. Please see the URL above for details on accessing the published version. Copyright and reuse:Sussex Research Online is a digital repository of the research output of the University.Copyright and all moral rights to the version of the paper presented here belong to the individual author(s) and/or other copyright owners. To the extent reasonable and practicable, the material made available in SRO has been checked for eligibility before being made available.Copies of full text items generally can be reproduced, displayed or performed and given to third parties in any format or medium for personal research or study, educational, or not-for-profit purposes without prior permission or charge, provided that the authors, title and full bibliographic details are credited, a hyperlink and/or URL is given for the original metadata page and the content is not changed in any way. ABSTRACTA recurrent non-linear autoregressive with exogenous input (NARX) neural network is proposed, and a suitable fully-recurrent training methodology is adapted and tuned, for reconstructing cylinder pressure in multi-cylinder IC engines using measured crank kinematics. This type of indirect sensing is important for cost effective closed-loop combustion control and for On-Board Diagnostics. The challenge addressed is to accurately predict cylinder pressure traces within the cycle under generalisation conditions: i.e. using data not previously seen by the network during training. This involves direct construction and calibration of a suitable inverse crank dynamic model, which owing to singular behaviour at top-dead-centre (TDC), has proved difficult via physical model construction, calibration, and inversion. The NARX architecture is specialised and adapted to cylinder pressure reconstruction, using a fully-recurrent training methodology which is needed because the alternatives are too slow and unreliable for practical network training on production engines. The fully-recurrent Robust Adaptive Gradient Descent (RAGD) algorithm, is tuned initially using synthesised crank kinematics, and then tested on real engine data to assess the reconstruction capability. Real data is obtained from a 1.125 litre, 3-cylinder, in-line, direct injection spark ignition (DISI) engine involving synchronised measurements of crank kinematics and cylinder pressure across a range of steady-state speed and load conditions. The paper shows that a RAGD-trained NARX network using both crank velocity and crank acceleration as input information, provides fast and robust training. By using the optimum epoch identified ...

Closed-loop combustion control (CLCC) in gasoline engines can improve efficiency, calibration effort, and performance using different fuels. Knowledge of in-cylinder pressures is a key requirement for CLCC. Adaptive cylinder pressure reconstruction offers a realistic alternative to direct sensing, which is otherwise necessary as legislation requires continued reductions in CO2 and exhaust emissions. Direct sensing however is expensive and may not prove adequately robust. A new approach is developed for in-cylinder pressure reconstruction on gasoline engines. The approach uses Time-Delay feed-forward Artificial Neural Networks trained with the standard Levenberg-Marquardt algorithm. The same approach can be applied to reconstruction via measured crank kinematics obtained from a shaft encoder, or measured engine cylinder block vibrations obtained from a production knock sensor. The basis of the procedure is initially justified by examination of the information content within measured data, which is considered to be equally important as the network architecture and training methodology. Key hypotheses are constructed and tested using data taken from a 3-cylinder (DISI) engine to reveal the influence of the data information content on reconstruction potential. The findings of these hypotheses tests are then used to develop the methodology. The approach is tested by reconstructing cylinder pressure across a wide range of steady-state engine operation using both measured crank kinematics and block accelerations. The results obtained show a very marked improvement over previously published reconstruction accuracy for both crank kinematics and cylinder block vibration based reconstruction using measurements obtained from a multi-cylinder engine. The paper shows that by careful processing of measured engine data, a standard neural network architecture and a standard training algorithm can be used to very accurately reconstruct engine cylinder pressure with high levels of robustness and efficiency. 27 main-section pages (double spaced) 23 references Figures 1-13 No appendices

A new inverse model is proposed for reconstructing steady-state and transient engine cylinder pressure using measured crank kinematics. An adaptive nonlinear timedependent relationship is assumed between windowed-subsections of cylinder pressure and measured crank kinematics in a time-domain format (rather than in crank-angledomain). This relationship comprises a linear sum of four separate nonlinear functions of crank jerk, acceleration, velocity, and crank angle. Each of these four nonlinear functions is obtained at each time instant by fitting separate m-term Chebychev polynomial expansions, where the total 4m instantaneous expansion coefficients are found using a standard (over-determined) linear least-square solution method. A convergence check on the calibration accuracy shows this initially improves as more Chebychev polynomial terms are used, but with further increase, the over-determined system becomes singular. Optimal accuracy Chebychev expansions are found to be of degree m=4, using 90 or more cycles of engine data to fit the model. To confirm the model accuracy in predictive mode, a defined measure is used, namely the 'calibration peak pressure error'. This measure allows effective a priori exclusion of occasionally unacceptable predictions. The method is tested using varying speed data taken from a 3-cylinder DISI engine fitted with cylinder pressure sensors, and a high resolution shaft encoder. Using appropriately-filtered crank kinematics (plus the 'calibration peak pressure error'), the model produces fast and accurate predictions for previously unseen data. Peak pressure predictions are consistently within 6.5% of target, whereas locations of peak pressure are consistently within ± 2.7˚ CA. The computational efficiency makes it very suitable for real-time implementation. main-section pages (double-spaced) 33 references 15 FiguresNo Appendices