A modified RBF-CBR model considering evaluation index for gear grinding process with worm grinding wheel decision support system

The grinding process of transmission gear with worm grinding wheel usually relies on the manual parameter setting by experienced engineers, but the reliability and validity are quite hard to control due to many grinding factors that affect the surface quality (e.g. structure, size, and materials of gear, heat treatment process, surface accuracy requirements). Therefore, it is necessary to set up a decision system for assisting engineers with designing the process plan efficiently and reasonably. A modified model of the revision method based on the radial basis function (RBF) neural network and case-based reasoning (CBR) considering evaluation index was proposed in this paper. A similar case retrieval mainly relies on the soft likelihood function considering a compound evaluation index model, where the score value of each new case execution test is converted into the corresponding index to achieve effective case storage. Process solutions for similar cases that perform better will be recommended from the case database by means of CBR. These candidates can be revised systematically based on a self-learning model. A modified RBF neural network trained by the existing cases will apply the learned experience to revise solution in part or in whole, which utilizes a self-filtering attribute method based on the correlation coefficient matrix so as to gain the primary attributes. In the case of the grinding gear with worm grinding wheel, the applicability of this technology was demonstrated. A decision-making system applying this method was developed by using.Net framework4.0 and SQL Server. Consequently, the technique can quickly generate a feasible process plan for specific gear.