**5. Conclusions**

In this research, auto-machine learning was utilized to optimize failure modes handling by automatically identifying the failure mode, obtain its RPN and identify the manufacturing process related to the root cause of the issue. Three multiclass-classification machine learning models were developed to predict values for the RPN three elements namely severity, occurrence and impact. A fourth multiclass-classification model was developed to classify failures to their root cause process. The models' evaluation indicated relatively high accuracy models that can be deployed and integrated to enhance the company's ERP system. One of the features of the selected AutoML platform is its simple integration through the API, which is o ffered on the cloud. Such technology performs e fficiently for large applications at the macro level of the factory. Utilizing such a solution enhanced the capabilities of the quality managemen<sup>t</sup> team to handle any volume of claims data under high flow velocity. Such a solution will allow the quality team to focus on other strategic issues which will enhance the team's performance and results. The benefits of such technology do not end by this, but also could be furtherly extended to link claims and defects to the relevant manufacturing machine and operator. Once a claim is reported to the quality managemen<sup>t</sup> it will be processed by the deployed model and instantly will be communicated to the relevant operators or managers and deeper to the shop floor in the factory. One more result for this research is that the manufacturing quality checklists for the selected product can be dynamically updated to include the top ten issues which are updated continuously according to their RPN. Such improvement enhanced the quality of processes and products. The factory in this study uses large screens on the shop floor to display quality checklists at every manufacturing process. These are used to review the quality issues while manufacturing processes are in place. A final check is being made at the quality gate of every process. The operators can watch the screens which

are updated every while and learn instantly about recently reported issues and making immediate correction actions. Finally, it is important to highlight the factors that impact the quality and accuracy of the developed models. For example, the accuracy of the model is strongly dependent on the quality of the data originated at the first point where the failure or defect was initially detected. Empty data rows, ambiguous information, or mistyping could forfeit important features and therefore, result in inaccurate prediction and reduce the system credibility. Therefore, a recommendation was suggested to the company to develop the data gathering platform (ERP system) in order to ensure higher quality prediction in the future. Furthermore, it is also essential to keep updating and maintaining the model by conducting periodical review sessions for the predicted RPN values and correct them when needed. Retraining the model using a larger volume of data will accumulate the model experience and improve model accuracy.

**Author Contributions:** S.S. (Conceptualization, Methodology, Investigation, Writing—Original Draft, Writing—Review & Editing). I.H. (Conceptualization, Supervision, Validation). M.D. (Supervision, Project administration). All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Stipendium Hungaricum Programme and by the Mechanical Engineering Doctoral School, Szent István University, Gödöll˝o, Hungary.

**Acknowledgments:** Special thanks to CLAAS Hungaria Kft, especially to Mr. Robert Csombordi, Head of Quality Management, for their endless support in conducting this research work.

**Conflicts of Interest:** The authors declare no conflict of interest.
