Research On Anti-interference-oriented Ignition Timing Control For HCCI Engines S130301040

Nowadays, energy shortage and environmental pollution are becoming increasingly serious. In this context, homogeneous charge compression ignition(HCCI) engines with both high efficiency and low permission draw a lot of attention worldwide. To function effectively, HCCI engines need an appropriate ignition timing control strategy, which should be able to reject the various disturbances from both inside and outside the engine. However, existing control strategies are unable to meet this requirement due to their lack of consideration for disturbances or due to the limitations of their internal algorithms. Therefore, research into the ignition timing control strategies of HCCI engines is carried out in this thesis, with the focus on the improvement of the anti-interference of the strategy.1. Model Construction of HCCI engineAs the basis for the design of control strategies, a model of the HCCI engine, with a state-space form, is firstly constructed. The model's inputs are the crank angles at which the intake valve opens and the exhaust valve closes and the output is the ignition timing. The single state variable of the model is the mixture's temperature when the intake valve closes. The mathematic expressions of the model are obtained through detailed analysis of the chemical kinetics inside the cylinder. The model built here fully considers the effect of residual exhaust gas on the ignition timing. Moreover, the model is easy to be computed while keeping a relatively high accuracy. Therefore, it is more suitable than other existing models to serve as the basis of the control strategy research of this thesis.2. Research on PI ignition timing control strategy based on BP neural networkA PI control strategy based on BP neural network is researched, aimed at improving the anti-interference capacity of the traditional PI controller. First, a PI controller is designed, with the error between the desired and the actual ignition timing as the controller's input and the incremental crank angle when the intake valve opens as the output. In order to adapt the parameters of the PI controller for different operating conditions, a BP neural network is constructed, whose inputs are the desired ignition timing, actual ignition timing and the error between them and whose outputs are the paramters of the PI controller. Simulation results show that when the desired ignition timing has a step change, the PI controller based on BP neural network can drive the actual ignition timing to track the desired timing without a steady error. In the meanwhile, compared with the traditional PI controller, the controller proposed here has a shorter and smoother transient process, demonstrating its superior tracking performance. When various external disturbances emerge, the PI controller based on BP neural network can control the ignition timing to return to its previous steady value more quickly and smoothly than the traditional PI controller, showing a better anti-interference capacity of the proposed controller.3. Research on feedforwad ignition timing control strategy based on disturbance observerThere is some room for improvement in the anti-interference capacity of the PI controller based on BP neural network. Besides, the transient response of this controller is not fast enough. To handle this problem, a feedforward control strategy based on disturbance observer is researched. According to this strategy, the internal parameter perturbations and external disturbances can be viewed as a single compound disturbance. Then a disturbance observer is designed to observe the compound disturbance. The observed value of the compound disturbance will be added to the feedforward control law to offset the effect of the compound disturbance on control performance. Results from simulation show that when the desired ignition timing has a step change, the feedforward controller based on disturbance observer only needs one cycle to adjust the actual timing to track the desired timing, indicating that its transient response is much faster than the PI controller based on BP neural network. Simulation results also show that when compound disturbances emerge, the feedforward controller based on disturbance observer is able to regulate the ignition timing to a smaller range, showing a better anti-interference capacity compared with the PI controller based on BP neural network.