Sorting of Portable Small Metallic Components using Machine Learning Technique

摘要

Modern auto-mobile industries have challenge of increasing manufacturing rate in order to meet the increasing demand of market. Auto-mobile uses plenty of small castings, fabricated components, and sub-assemblies. These components are required to be tested thoroughly before deployment. The parent production industry generally offloads this task to vendors. The quality control of these produced components plays vital role in this process. The use of Non-destructive testing (NDT) methods helps to accelerate production process to a greater extent. The NDT based on acoustic resonance facilitates to test products in-line. It follows the principle of acoustic resonance (AR). The component under test (CUT) is triggered by an impact or induced with a shock wave, and the acoustic response is acquired and analysed. In addition to this the weight and Brinell hardness (HB) of the same CUTs is also measured and documented. Hence every CUT has its unique signature composed of acoustic resonance frequency (ARF), weight and (HB) hardness number. The signatures of the CUTs are compared with the master component's signature. The result is used to detect faulty CUT on the basis of Go-NoGo decision. In this paper we have proposed, developed and tested the machine learning techniques which help to take the same decision quickly. The consistency of the method is tested on the set of similar components. The responses are tested by introducing a fault in a test-component. The results are useful to sort the CUTs in 'Go' and 'NoGo' class. The decision saves labour-cost and time.