Industrial Workflow Replay Using Process Mining and Generative AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The aging workforce in Process and Automation Engineering (PAEN) and Control Engineering leaves a knowledge and experience gap that needs to be saved and utilized by an aging workforce in Process and Automation Engineering (PAEN) and Control Engineering, as experienced engineers' knowledge and experience is not effectively preserved and utilized.
Innovation Solution
A method utilizing process mining and generative artificial intelligence to capture and replay expert knowledge by analyzing historical processes, identifying patterns, and suggesting next steps or sequences, supported by user feedback for improved support in industrial plants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If expert knowledge is saved through traditional documentation methods, then knowledge preservation is achieved, but knowledge utilization and accuracy are insufficient
Solution Approach 1:
The patent creates digital copies of expert knowledge by capturing actual process execution data from industrial plants. Instead of manual documentation, the system records real process steps, decisions, and outcomes as they occur, creating accurate digital replicas of expert workflows that can be analyzed and reproduced.
Solution Approach 2:
The patent replaces traditional mechanical documentation methods (manual writing,纸质 records) with automated digital data capture and process mining technology. This substitution enables automated analysis of expert behaviors and generation of actionable insights without human intervention in the knowledge extraction process.
2Productivity
If process mining and generative AI are used to analyze historical processes, then knowledge utilization is improved, but data processing complexity increases
Solution Approach 1:
The patent segments the complex task of knowledge extraction into distinct phases: data collection from operational logs, process mining analysis to identify patterns, generative AI modeling to create knowledge representations, and deployment as actionable insights. This segmentation makes the overall complex system manageable and implementable.
Solution Approach 2:
The patent introduces process mining technology as an intermediary between raw operational data and generative AI models. This intermediary layer preprocesses and structures the data, extracting meaningful process patterns before feeding them to the generative AI, thereby simplifying the overall data processing pipeline.
3Measurement precision
If historical process data is collected and analyzed, then next step prediction accuracy is improved, but data storage and processing requirements increase
Solution Approach 1:
The patent extracts only the essential and relevant features from historical process data using process mining techniques. Instead of storing and processing all raw operational data, the system identifies and extracts key process steps, transitions, and patterns that are critical for prediction, thereby reducing data storage requirements while maintaining prediction accuracy.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
There is disclosed a method for saving and utilizing expert process knowledge with regard to industrial plant. The method comprises obtaining first data indicative of historical processes comprising sequences of historical process steps occurred in an industrial plant. The method comprises obtaining second data based on analysing the first data, wherein the second data is indicative of occurrence statistics and thereon-based probabilities for several historical processes and several historical process steps. And the method comprises, based on the second data, suggesting, for at least one of a process step, a sequence of process steps and a process, at least one of a next process step, a next sequence of process steps and a next process. The analysing comprises analysing the first data using process mining and/or processing the first data using generative Al. The suggesting comprises suggesting the at least one of the next process step, the next sequence and the next process by using process mining and/or by using generative Al.