| Summary: | 碩士 === 國立高雄應用科技大學 === 電機工程系博碩士班 === 106 === This study applies intelligent modeling methods in two different cases: “low-voltage electrical analysis” and “robotic manufacturing problems”. This study uses similar modeling strategy because both cases use big data. And study the quality of modeling using artificial neural networks. Neural networks data cut with time and combine the temperature and humidity data that make analysis content is closer to the actual application level. Using fuzzy method and artificial neural networks to analyze low-voltage user data. Cutting axis space of machine arm and using lot of neural networks and forward kinematics to modeling inverse kinematics of machine arm. It can quickly solve the joint value of terminal that machine arm arrives to processing point.
In our study, neural networks modeling should pretreatment large data before training this model. And compare the quality of solution. We can know that regression analysis and neural networks have their own advantages and disadvantages from the experimental results of case one. Regression analysis has high stability in R analysis and it can execution acceptable modeling results stably. Neural networks can get better R value result than regression analysis after different parameter adjustment. However, the stability of neural networks analysis is lower than that of regression analysis. The final goal is to build the best model for future. Therefore, we recommend that you can use neural networks and regression analysis to find the best model. The results of case two show that neural networks training model after cutting has lower error result than training model without cutting. It can provide smaller error on inverse kinematics. From these two cases, In these two cases, we can see that the case of big data modeling, neural networks use cutting methods and cooperate with multiple neural networks can achieve more accurate and better results.
|
|---|
