Search

The control of internal combustion engines is becoming increasingly challenging to the customer’s requirements for growing performance and ever-stringent emission regulations. Therefore, significant computational efforts are required to manage the large amount of data coming from the field for engine optimization, leading to increased operating times and costs. Machine-learning techniques are being increasingly used in the automotive field as virtual sensors, fault detection systems, and performance-optimization applications for their real-time and low-cost implementation. Among them, the combination of long short-term memory (LSTM) together with one-dimensional convolutional neural networks (1DCNN), i.e., LSTM + 1DCNN, has proved to be a promising tool for signal analysis. The architecture exploits the CNN characteristic to combine feature classification and extraction, creating a single adaptive learning body with the ability of LSTM to follow the sequential nature of sensor measurements over time. The current research focus is on evaluating the possibility of integrating virtual sensors into the on-board control system. Specifically, the primary objective is to assess and harness the potential of advanced machine-learning technologies to replace physical sensors. In realizing this goal, the present work establishes the first step by evaluating the forecasting performance of a LSTM + 1DCNN architecture. Experimental data coming from a three-cylinder spark-ignition engine under different operating conditions are used to predict the engine’s in-cylinder pressure traces. Since using in-cylinder pressure transducers in road cars is not economically viable, adopting advanced machine-learning technologies becomes crucial to avoid structural modifications while preserving engine integrity. The results show that LSTM + 1DCNN is particularly suited for the prediction of signals characterized by a higher variability. In particular, it consistently outperforms other architectures utilized for comparative purposes, achieving average error percentages below 2%. As cycle-to-cycle variability increases, LSTM + 1DCNN reaches average error percentages below 1.5%, demonstrating the architecture’s potential for replacing physical sensors.

To carry out increasingly sophisticated checks, which comply with international regulations and stringent constraints, on-board computational systems are called upon to manipulate a growing number of variables, provided by an ever-increasing number of real and virtual sensors. The optimization phase of an ICE passes through the control of these numerous variables, which often exhibit rapidly changing trends over time. On the one hand, the amount of data to be processed, with narrow cyclical frequencies, entails ever more powerful computational equipment. On the other hand, computational strategies and techniques are required which allow actuation times that are useful for timely and optimized control. In the automotive industry, the ‘machine learning’ approach is becoming one the most used approaches to perform forecasting activities with reduced computational effort, due to both its cost-effectiveness and its simple and compact structure. In the present work, the nonlinear dynamic system we address is related to the torque estimation of an ICE through a nonlinear autoregressive with exogenous inputs (NARX) approach. Preliminary activities were performed to optimize the neural network in terms of neurons, hidden layers, and the number of input parameters to be assessed. A Shapley sensitivity analysis allowed quantification of the impact of each variable on the target prediction, and therefore, a reduction in the amount of data to be processed by the architecture. In all cases analyzed, the optimized structure was able to achieve average percentage errors on the target prediction that were always lower than a critical threshold of 10%. In particular, when the dataset was augmented or the analyzed cases merged, the architecture achieved average prediction errors of about 1%, highlighting its remarkable ability to reproduce the target with fidelity.