支持向量迴歸對鋁合金輪圈的加工參數研究

壓鑄在現代的鑄造業中，占了很重要的角色及地位，然而鑄造程序引起成品的良率及鑄造成本的考量，導致尋找出一種能提前預測的方法成為迫切的需求，而在大部份的工廠中，要分析出鑄造程序哪裡出問題，通常都要廠內有資深的壓鑄師傅經過試誤法花費時間來找出問題所在，或者由外聘專業人員來做檢測也是一筆可觀的支出，況且若在開模鑄造後才發現一大批產品都不能用，那對企業來說絕對是一種致命的傷害。近年來隨著人工智慧的崛起及更多演算法的精進，支持向量回歸被廣泛應用在多種領域，而在CNC這一方面，被用來預測加工溫度、粗糙度、加工時間等，本論文利用支持向量回歸來預測鑄造時的鑄孔不良程度，利用所得到的數據進行模型的訓練，其中考量到各種塗模劑、鋁湯溫度、模溫、壓力段、時間段所產生的影響，最後使用訓練完成的模型來預測鑄孔不良程度，並從中找出最佳的塗模劑、鋁湯溫度、模溫、壓力段、時間段組合，並在實驗進行中印證，支持向量回歸結合高斯核函數(rbf)的預測準確率可以達到99.56%，研究證實，透過此論文的方法，能夠找出壓鑄機台的最佳化預測模型。

關鍵字

支持向量回歸；鑄孔不良；鋁合金輪圈

並列摘要

Die casting plays a very important role and position in the modern casting industry. However, the casting process causes the yield of finished products and the cost of casting to be considered, resulting in an urgent need to find a method that can be predicted in advance. In the factory, analyze the problems in the casting process is usually necessary to have a senior die casting master in the factory to spend a trial and error method to find out the problem, or an external professional to perform the test is also a considerable expense. Moreover, if a large number of products are found to be unusable after the mold is made, it is definitely a fatal injury for the company. In recent years, with the rise of artificial intelligence and the advancement of more algorithms, support vector regression is widely used in many fields. In the aspect of CNC, it is used to predict processing temperature, roughness, processing time, etc. This paper use the support vector regression to predict the degree of casting hole defects during casting and use the obtained data to train the model. In this paper, takes the effects of various coating agents, liquid aluminum temperature, mold temperature, pressure section, time period , and finally use the trained model to predict the degree of casting hole failure. Finally, I confirm it by actual experiments, the prediction accuracy rate can reach 99.56%. Research has confirmed that through the method of this paper, the optimal prediction model of die casting machine can be found.