| Volume of automotive cylinder head combustion chamber not onlycasts great influence on engine’s power performance, but also is closelylinked to defects such as detonation, leakage and so on. Consistency ofcombustion chamber volume is of significance to the performance ofproduced engines as well as rate of pass.In regarding to the fact that the control of combustion chamber volumeis difficult and studies on this respect is insufficient at this stage, we presenta systematic research based on data mining techniques in this paper. Anoptimal control scheme is built by combining mold control method andmachining parameter determination. This study can not only provide ascheme which can be applied to cylinder heads used on engines produced ina plant in China, but also can provide an example for a systematic study onthe respective component in engines made by other manufacturers.This paper concentrates on the follow aspects:（1）The study of cylinder head manufacturing processIn this section, the manufacture process of cylinder head is investigatedand introduced chain by chain. The monitoring and controlling schemes arestated in detail as well. Based on statistical testing, variation patterns ofcombustion chamber volume are extracted. Then volume measurement datais analyzed with ensemble empirical mode decomposition and underlyingvariation cycles are unveiled. Finally, factors which account for thesevariation patterns are given according to expert knowledge.（2）Devising cylinder head casting mold gauging scheme In casting procedure, the molds are used almost24hours per day.Although the maintenance expands through the whole life cycle of a mold,the state of the mold would be changing from time to time, thus cast a greatinfluence on the cylinder head product. In this section, current adopted molddetection and evaluation method is studied and its weaknesses are presented.To overcome these drawbacks, we devised a new scheme to better controlvolume variation brought about by the change in mold state.Based on the point cloud generated through optical measurement, weemploy methods in reverse engineering to calculate the volume change ofthe cavity in the molds. Then spectral clustering is applied to divide thewhole cavity into several regions. We penetrate the influence of the changeof point coordinate in z direction in different regions on the whole volume.Through this study, we can understand that which part in the cavity wouldbe the most predominate region in affecting the volume cavity. Local andglobal features of different regions are taken into consideration to make areasonable arrangement of gauging points. In this way, a new and morelogical detect scheme is placed in substitution of the current empiricalscheme.（3）The construction of optimal method for determining machiningparameterIn the manufacturing of cylinder heads, we indirectly control thecombustion chamber by the distance from two pre-designed bulges to theunderside of a cylinder head. And after the whole process, opticalmeasurement comes to check whether the volume is under control. Thoughthis method is practical in engineering, it is far from sophisticated andreasonable since this parameter is not proportional to the volume.To take use of this data and try to penetrate the mechanism of errorpropagation from casting to machining and finally to the product, we use thedistance data to build a time-series relationship between the measurement ina particular product and a product manufactured right before it. Then thisrelationship can help us evaluated the change of volume in time order. Dynamic neural network is adopted to describe the change of volumeof products made by different molds and in different time periods. In thispart, a discrete Hopfield network is constructed. It takes the abstracted time-series data, the gauged distance data and the number of mold as inputs andpredicts the class of volume. After the model is built, it can be takenadvantage of to help determine the machining parameter. |
PDF Full Text Request