This work demonstrated the following novel contributions:
(a) A transient diesel engine model has being created using AVL-BOOST simulation software. The model was configured based on the specifications of the CAT3126B engine, and it is able to simulate the engine in both steady state and transient running conditions. The timings and lifts of intake and exhaust valves were derived from the profiles of the intake cam and exhaust cam. The timing and lift were further tuned by calibrating them for the best performances of the engine. The fuel injection map for the ECU sub-model was created based on the natural air intake curve. The VGT sub-model was configured using the OEM maps of the compressor and turbine, the EGR valve was simulated by defining a look-up table between the flow coefficients and valve positions. The engine model was validated against the CAT3126B engine, the results show that the engine model is able to represent the CAT3126B engine with relatively small differences. There is a full range of parameters of the engine are accessible across engine. Temperatures, pressures and heat transfer are accessible by adding measuring points to the positions where needed. The combustion behaviour and energy analysis are also available. More interested results would be BMEP, engine torque, BSFC, soot and NOx productions of the engine. The PID controller, fuzzy logic controller, ANFIS controller can be tested and validated using the engine model before they can be tested in the real engine, also the performance of the
on-board emission predictors can also be assessed using the engine model.
(b) Traditional PID controllers are good at resolving the control target at a predefined value (set-point) precisely when running in steady state. However, diesel engines usually run at highly transient conditions, such as the NRTC cycle used in this work. Hence, a map of predefined values for the control target has to be defined firstly. Because the set-point of the control target changes rapidly under transient running conditions as defined in the map, the PID controller with fixed coefficients may struggle to track the change of the set-points and the stability of the control would not be guaranteed, especially in this particular problem where there is interactive effect between the VGT and EGR as demonstrated in Section 4.1.3. The proposed real-time fuzzy logic controller is a novel approach to control the VGT and EGR in a diesel engine. Instead of regulating the VGT and EGR positions to achieve the predefined set-points of the inlet pressure and EGR mass fraction, the fuzzy logic controller is able to rapidly determine the optimized positions of the VGT and EGR by monitoring the engine running conditions and emission outputs in real time. Hence, the fuzzy logic controller is much more robust compared to the traditional PID controllers. There are still PID controllers working in the inner loop to position the VGT vanes and EGR valves at the positions determined by the fuzzy logic controller, but compared to regulating the VGT and EGR positions to achieve the set-points of inlet pressure and EGR mass fraction, this local position control has fewer nonlinearities, the hysteresis for example. Hence, the PID control is adequate. Results have shown that compared to the engine performances and emissions when the VGT and EGR are controlled by the PID controller, the engine performances has been improved when the VGT and EGR are controlled by the fuzzy logic controller, the fuel consumption and emissions have been reduced.
(c) ANFIS controllers have been developed in this work. By learning from a data set collected from transient experimental tests on the CAT3126B engine when the VGT and EGR are controlled by the fuzzy logic controller, the ANFIS controller is able to determine the VGT and EGR positions without monitoring the emission outputs from the engine, and hence the ANFIS controller can replace the fuzzy logic con- troller when online emission measurement is not available. It has been anticipated that the performances and emissions of the engine when the VGT and EGR are con- trolled by the ANFIS controller should be very close but not better than the engine performances and emissions when the VGT and EGR are controlled by the fuzzy
logic controller. However, both simulation and experimental results show that the performances of the engine when the VGT and EGR are controlled by the ANFIS controller can be improved compared to the performances of the engine when the VGT and EGR are controlled by the fuzzy logic controller, emissions can also be slightly reduced when the VGT and EGR are controlled by the ANFIS controller. (d) On-board emission predictors for soot and NOx have also been developed in this
work, and the predictors are used to provide the fuzzy logic controller with predicted emission outputs as the control inputs required by the fuzzy logic controller. The ANFIS modelling technique is used to develop the predictors by learning from a data set collected from transient experimental tests that include engine speed, inlet oxygen concentration, intake air flow, and load as input variables, with the soot and NOx as the output variables for soot predictor and NOx predictor, respectively. Experimental results show that the engine performances and emissions when the VGT and EGR are controlled by the fuzzy logic controller with emission predictors are very close to those of the engine when the VGT and EGR are controlled by the fuzzy logic controller with emission analysers. Most of the time during the tests, the engine performances and emissions are slightly better with emission predictors than those with emission analysers. This is caused by the differences between predicted values and measured values of soot and NOx, and the predicted values are generally slightly higher than the measured values.