ANALYSIS OF THE POSSIBILITIES OF IMPLEMENTATION OF ARTIFICIAL INTELLIGENCE TO MAPPING OF THE SPARK IGNITION ENGINE
The need for simulation engine is indisputable. Each simulation requires a mathematical model. Building a comprehensive mathematical model of the motor is difficult, because it was used here artificial intelligence. It was noted gaps in the literature about such a description of the engine. Therefore it conducted appropriate studies and compiled the results. Obtained mathematical model of the engine in the form of a neural network. Research shows that it is best to build a network of multi-way because they provide accurate and quick to carry out analyzes. The result is close to reality, as confirmed by the work. It was found after a preliminary analysis, that the best is three-tier network. Developed network has a first layer 3 of the neurons, the neurons in the second 10 and the third 5 neuron network with two inputs and five outputs. It is noted that the neural network results in averaging, or smoothing characteristics. For this reason, the results generated are different from real ones. Another factor influencing the formation of the differences can be inaccuracy and lack of reproducibility of measuring devices. On the other hand, the characteristics of the engine, in fact, is not as "smooth" as presented so far. Approximations that were obtained are acceptable. For more accurate results should be in the future to collect a larger data set for a variety of engines and carry out more research engines of various types and capacities. The network can be used to create various types of controls and for educational purposes as a model for the internal combustion engine.
Tematyka artykułu: Badania, diagnostyka, trwałość i niezawodność silników
Autor: Lech J. SITNIK
Współautor(zy):