Artificial Intelligence in Quality Assurance
With the help of a systematic analysis, defects in the product or the manufacturing process and their causes can be detected at an early stage. Artificial Intelligence methods can support such an analysis in quality assurance.
Failure Mode and Effects Analysis (FMEA) is a method of preventive quality assurance and a proven method for identifying and evaluating potential product defects. It is applied in the run-up to product creation in the development and production planning phase. In the method, system functions and elements are analyzed for potential failure causes, types and consequences.
Then, the potential product failures are evaluated based on their probability of occurrence, probability of detection, and severity of failure, and assigned a resolution action. In complex systems, a large number of possible product defects can occur.
Artificial Intelligence methods will be used to automatically generate suggestions for possible product defects based on the specified product functions and elements.
Here, the information from previous analyses is used to give a suggestion for occurring faults and fault evaluation. To do this, inputs such as system functions, system elements and past analyses must be collected and processed. Natural Language Processing (NLP) techniques such as Text Mining and Named Entity Recognition (NER) can be used for this purpose. Text mining is particularly suitable for the analysis of unstructured text data. Automatic analysis can be used, for example, to extract core statements from texts without having to read the texts themselves. Named entity recognition aims to find named entities (such as names of people, organizations, time references) in texts and classify them into predefined categories.
The resulting entities and keywords can be merged via Similarity Learning to identify potential product defects for the current product. Similarity Learning is an area of machine learning. The goal of this method is to learn a similarity function that measures how similar or related two objects are. Similarity Learning is used in ranking, recommendation services, or even face recognition.
To identify patterns in error scoring (especially for error severity and frequency of occurrence), the mathematical method Support Vector Machines (SVM) can be used. Thanks to SVM, it is possible to analyze data and assign objects to specific classes.
This is what companies need to consider
The effort for companies lies mainly in the availability of a sufficient database in usable quality. The entries in the error database must contain a detailed description of the causes and must also be made accessible to external systems via an API. In doing so, the high data and access protection regulations must be observed.
So what are the concrete advantages of AI methods in quality assurance? The automated and thus more efficient construction of the initial fault tree increases the process performance of the product development process. In addition, consistent automated reuse of field knowledge leads to optimization of the final product. The resulting sustainable product design has a lower susceptibility to defects.