Dependable Troubleshooting Assistant - LLM-Based Support for Troubleshooting

Our approach enables manufacturing companies to perform complex troubleshooting faster and more reliably by integrating relevant knowledge sources (machine documentation, machine data, and domain knowledge) as well as live and log data into an LLM-based troubleshooting assistant that provides clear instructions on the cause of the quality issue.

The troubleshooting assistant provides an intuitive, natural language interface and allows operators to focus on solving the problem at hand. We focus on LLM-supported assistance in high-stakes industrial applications to ensure rapid root cause analysis, practical troubleshooting measures, and more flexible, downtime-reduced production.

Challenge: Targeted and efficient troubleshooting

• When production disruptions occur – such as unexpected quality issues – it is often difficult to identify the actual cause. This is due to the continuously changing manufacturing systems, complex processes, and varying degrees of digitalization within the factory.
• Best practices for troubleshooting are usually only implicitly available to experienced employees and have often not been documented or passed on. When qualified personnel leave the company, this knowledge is lost.
• Troubleshooting often requires time-consuming investigation of documentation and tedious cross-checking of various information.
• However, troubleshooting must be carried out as quickly as possible, as production downtime incurs high costs.

Solution: Reliable and trustworthy troubleshooting assistant

To perform the troubleshooting process, they interact with the assistant and receive guidance on identifying the cause of the quality issue. To do this, the assistant uses techniques such as retrieval pipelines or knowledge graph integrations to capture domain knowledge. In addition, it leverages additional sources of information such as real-time and historical data, logs, and manufacturing software systems to provide tailored, real-time insights. To ensure the dependability of the LLLM-based system, we focus on using novel and cutting-edge technologies to ensure reliable and trustworthy answers about the current system status, potential causes, and the recommended steps for troubleshooting.

Key Benefits:

• Accelerated troubleshooting and reduced downtime
• Improved knowledge distribution and retention
• Improved decision-making through dependable guidance
• Integration of live/historical data and logs for real-time insights

Insight Background

At the forefront of innovation, we want to create a safer and more efficient future in industrial environments. Our research addresses the crucial question: “How can LLM-based troubleshooting assistance reliably coexist with flexible and resilient manufacturing systems?” Through a structured approach to dependability assurance, we enable companies to effectively manage the complexity of LLM adoption and human-machine interaction. To achieve these goals, we develop tools and methods that accelerate troubleshooting and make the necessary knowledge reliably accessible to employees. We integrate troubleshooting knowledge from various sources, such as machine documentation, machine data, and codified domain knowledge, to ensure a comprehensive knowledge base. The focus is on making the troubleshooting process as reliable and intuitive as possible so that employees can quickly understand and resolve production problems. Our research focuses on trustworthy artificial intelligence, safety assurance, and robust software systems. Join us in revolutionizing the integration of AI into safety-critical environments and making troubleshooting accessible to all industries.

In-depth information

• Troubleshooting Assistance

Sponsored by

Bavarian Ministry of Economic Affairs, Regional Development and Energy