Abstract:

Condition evaluation usually runs into excessive calculation and difficulty in weight determination. This problem is caused by information redundancy when the evaluation information is composed by multi-dimensional characteristic parameters. For this, the self-organizing feature map (SOM) network is opted to discretize parameters with continuous ranges, for the SOM network is capable of clustering without supervision. At the same time, the redundant algorithm of attributes based on rough set is used to eliminate redundant characteristic parameters. Mutative scale chaos optimization algorithm is utilized in optimization of weights for resulting parameters of reduction. The outcome of case study shows that the method applied is notably able to solve the optimization of weights for multi-dimensional parameters and enhance the accuracy of evaluation.