https://www.mdu.se/

With the manufacturing industry moving towards digitalization, it has opened the possibility for new opportunities for improvements technologies. Where one such promising technology that has emerge is Artificial intelligence (AI) and with recent progress in the field, a subset of AI, machine learning (ML) has become a useful tool and greatly enhancing data processing. Because product quality is regarded as an important competitive factor in the manufacturing industry and combining this with the ever-increasing amount of data that can be collected in manufacturing, ML can enable further improvements in quality control by enabling improvements of quality data handling. However, the implementation of ML is limited by many factors, the complexity of the process in which its applied, the inhouse knowledge the manufactory company processes of ML, as this is in general limited to how much resources they have. Combining this with ML-complexity and continuous change suggests that this may be a difficult task. 

This thesis will investigate implementation of machine learning for quality control, covering challenges and enablers that may occur before and during the implementation. Machine learning implementation processes for quality control are rare in literature. Thus, this project was carried out through a case study at a heavy vehicle company located in Eskilstuna Sweden. Challenges found for machine learning implementation in quality control was: Poor quality of data, Insufficient process for data gathering, Inability to take ownership of the complex case, Lack of Knowledge about machine learning, Unclear usage of data, Data visualization issue. Enabling step that should be considered when implementing machine learning for quality control found during the project was: Create a cross-functional team, Map out the problem, Involve external parties, Create in-house knowledge of data usage, Assist decision-makers in making decisions. During this thesis a ML-model was created and tested, as well as an implementation process for ML-implementation.