Tooth Reducer, Swing Based On Neural Network Multi-objective Fuzzy Robust Optimization Design

The robust design method is a kind of engineering design method which have developed in recent years, the guidelines are the use of the situation when factors (causes) change in a small scale as a result of modest deterioration variables (results) that do not affect the sensitivity of product performance, quality and cost, and make the best integrated into the design to improve product quality and reduce costs. This shows that the robust design is a real optimization problem. This paper takes the robust design method as the robust design optimization. At present, the robust design method of the mechanical transmission used in the optimal design is the traditional method, which did not consider all the design parameters of the fuzzy and randomness, and in the process of optimization, it still uses the traditional optimization such as these traditional optimization algorithm with a local phenomenon optimal and optimization slowly, the optimal design of such traditional methods which are often inconsistent with the actual situation and even the program of the design is not feasible. Based on the shortcomings of the traditional method, this paper considers the design parameters of fuzzy and randomness, combine robust design and fuzzy design into fuzzy robust optimal design, And using of the neural network of optimization method to find the optimal solution of the multi-objective fuzzy and robust optimal design of the mechanical transmission .Swing movable teeth is a new type of high-performance transmission device, which is used to transfer between the two coaxial rotary movement of new less -tooth difference Planetary Gear (movable teeth transmission), which has many advantages such as a compact structure, a wide range of transmission ratio, carrying capacity, and high transmission efficiency . These characteristics will have broad application prospects in the industry but immaturity in some ways, and do not acquire any results by general optimal design. Neural network optimization application is to use neural network energy function corresponds to a very small point system of stable equilibrium. The minimum point solution of the energy function can take into the stability and balance of neural network system .As the time changes , the movement track of neural networks in phase space is always toward to the direction of reducing the energy function movement, eventually reach the point of balance- the minimum point of energy function. So if the neural network dynamic system of stable equilibrium to be considered to an appropriate energy function (or generalized objective function)is a very small point, optimal design is to find a very small point from an initial target function as it obtained for the most optimization solution. Therefore this paper will be applied this method to the multi-objective fuzzy and robust optimal design of the Swing movable teeth reducer.This paper will take the present more popular large-scale scientific computing software-MATLAB to optimal and calculate. First to design it in MATLAB condition, then to analyze optimal ends by robust method of Taguchi—parameter design in inner table and outer table, to educe optimal ends. Based on the fuzzy robust optimal design take the neural network used in this paper to solve optimization problems, it is the final results of optimization. Then take the optimize results back to the objective function and restraint function. Verified the results it is reasonable. The results show that this optimization methods improve the quality of the optimization and then use the software MATALB to analysis image, finally design virtually swing movable teeth reducer in Pro/ENGINEER condition.