A Prestudy of Machine Learning in Industrial Quality Control Pipelines

Today’s fast paced industrial production requires automation atmultiple steps during its process. Involving humans during thequality control inspection provides high degree of confidence thatthe end products are with the best quality. Workers involved inthe control process may have an impact on production capacityby lowering the throughput, depending on the complexity of thecontrol process at the time the control is carried out, during theprocess which is a time-critical operation, or after the process iscompleted. Companies are striving to fully automate their qualitycontrol stages of production and it comes naturally to focus onusing various machine learning methods to help build a qualitycontrol pipeline which will offer high throughput and high degreeof quality. In this paper we give an overview of applying severalmachine learning approaches in order to achieve an autonomousquality control pipeline. The applications for these approacheswere used to help improve the quality control pipeline of two ofthe biggest manufacturing companies in Slovenia. One of the mostchallenging part of the study was that the tests had to be performedonly on a small number of defective products, as is in reality. Themotivation was to test several methods to find the most promisingone for later actual application.