OL24 Analyzing root-cause failure through event log data from machine stops using AI models

Aim/goal/research question

This master’s thesis seeks to employ artificial intelligence, to analyze event logs from breakdowns of industrial machines. The study aims to discover the root cause of a machine or robot breakdown with specific stop events generated from sensor-based data in a machine or robot. Event log data from two industrial companies will be utilized, supplemented by insights obtained through interviews with engineering technicians and operators. Included in the thesis is to validate the effectiveness of the proposed AI-based approach through testing with datasets and real-world information. The thesis project is part of the Restart II project. Expected Results: An enhanced decision model for supporting the root cause of breakdowns of automated manufacturing systems, incorporating sensor-based system data for analyzing deviations from normal behavior. Validation and evaluation results demonstrate the effectiveness of the approach with diverse datasets.

Method

Event Log Analysis: In-depth exploration and understanding of the structure and content of event logs from two industrial companies. Feature Extraction: Identify relevant features in event logs and integrate them with the common procedures obtained through interviews with engineers. AI Model Development: Design and implement an AI model capable of analyzing event logs, incorporating procedural information for improved accuracy. Validation and Evaluation with Datasets: Validate the model using event logs from two distinct companies, ensuring robustness and generalizability. Interview-based Validation: Verify the model predictions with established procedures through interviews with engineers and operators to ensure practical relevance.

Recommended past experience/interest

Machine Learning, Data Analysis.

Other comments

Suggested timeline: Week 8 to week 9: Data collection, and initial interviews at the companies (the cases). Literature review. Week 10: Data preprocessing and continued literature review. Week 11 to week 14: AI model development and training from event logs. Week 15: Validation with datasets from both companies, initial evaluation. Week 16 to week 17: Verification through interviews, refinement of model and system. Week 18 to week 20: Result writing, final evaluation, and submission.

Partners

Sansera AB, Trollhättan Husqvarna AB, Brastad (Lysekil)

Contact

bassam.massouh@hv.se