Abstract
A fast experimental procedure for the calibration of internal combustion engines microprocessor control systems is proposed. This procedure is based on using a new approach to the solution of many-parametric optimization problems with constraints. The main benefit of this approach is that it requires a minimum number of experimental measurements where the measurements can have a high level of noise stability to the inaccuracy of measurements and adjustments. The efficiency of the suggested procedure is demonstrated by the search for air-fuel ratio, ignition timing and exhaust recirculation optimum bounded control that provide for the minimum fuel consumption for each speed-load regime at a given emission level. The optimization problem was defined by three variables and three constraints. The optimization criterion extremum value was found in 14 direct calls to the mathematical model computing the objective function. Ten tasks with different initial points of search have confirmed these results. The comparative analysis of the method efficiency with other well-known non-linear programming methods (the methods of Powell and Nelder-Mead) has shown the advantage of the proposed procedure that required only one-third of the experiments needed by these algorithms.