LLM-Based Support for Troubleshooting

Modern manufacturing systems consist of many heterogeneous and complex processes. At the same time, there is a growing need to adapt and integrate novel, disruptive technologies. This increasing complexity makes it more difficult to troubleshoot problems when they arise. It makes troubleshooting time-consuming and therefore expensive, leads to downtime, and requires extensive expertise. Fraunhofer IKS supports companies in accelerating and simplifying the troubleshooting process to ensure flexible and resilient manufacturing.

Contact us now

• When production malfunctions occur, it is often difficult to identify the actual cause. This is due to the ongoing reorganization of manufacturing systems, complex processes, and varying degrees of digitization within the factory.
• Knowledge about troubleshooting is usually only implicitly available to experienced employees and has often not been documented or shared. When qualified personnel leave the company, this knowledge is lost.
• Even if documentation exists, it must be reliable and trustworthy to avoid false conclusions.
• In addition, troubleshooting must be carried out as quickly as possible, as production downtimes incur high costs.

• To overcome these challenges, employees need tools that accelerate the troubleshooting process, make the necessary knowledge easily accessible, and provide reliable guidance for identifying faults.
• A key prerequisite is the integration of troubleshooting knowledge from various sources, such as device documentation, machine data, and codified knowledge from experts.
• The troubleshooting process must be as intuitive as possible so that employees can quickly understand and correct production problems.

An LLM-based troubleshooting assistant provides an intuitive, natural language interface and allows employees to focus on solving the problem at hand. To perform the troubleshooting process, they interact with the assistant and receive guidance on identifying the cause of the error. To do this, the assistant uses techniques such as retrieval augmented generation (RAG) pipelines or knowledge graph pipelines to capture both explicit and implicit knowledge. In addition, it leverages additional sources of information such as live and historical data, logs, and manufacturing software systems to deliver tailored, real-time insights. Central to this is the use of novel and cutting-edge technologies to ensure reliable and trustworthy answers about the current system state, possible causes, and next steps for troubleshooting.