Adaptive and (semi-)autonomous modelling of causal graphs and causal inference for flexible production in SMEs
Small and Medium-sized Enterprises (SMEs) are struggling with the ongoing shortage of skilled workers. Furthermore, to react and adapt to frequent changes within the production process, flexible production is becoming a prominent part of manufacturing which requires profoundly skilled employees to operate the machinery. Since there is a need and shortage of qualified laborers, it is crucial for the workflow optimization to hold onto the acquired knowledge, transfer it to a new employee, and utilize the benefits of new methodologies simultaneously. An Expert System (ES) exists that formalizes the professional knowledge; however, these knowledge bases are used only to infer queries. Thus, they do not provide graphical understanding of the knowledge nor existing relations between entities and the user. Similarly, these systems infer results from observable information. Hence, it fails to capture the output for unobserved scenarios that professionals did not experience which could help in improving the process.
Ontologies allow knowledge to be formalized and represented in subject-predicate-object form. Likewise causal Bayesian network helps in inferring in unobserved scenarios; therefore, the main idea of this thesis is to use an ontology as a knowledge base to provide graphical representation as well as to create causal graphs and processing those causal graphs with observation data to create a causal Bayesian Network and infer causality.
As part of thesis, the following task should be done to solve the above-mentioned challenges:
- Preliminary view:
- Research on existing ontologies in the manufacturing domain
- Research on state of the art in deriving causal graphs from ontologies
- Building demonstrator tool:
- Analysis of extracting causal relations from ontology and generating causal graph (modelling of the ontology is not part of the work)
- Performing causal inference calculations on the example of manufacturing.
- Secure user input as basis of an integrated suggestion system (exemplary, only approach/concept should become apparent)
- Develop a user interface to perform causal inference calculations efficiently
- Verification based on two application scenarios from the manufacturing domain:
- Upload the ontology and visualize the RDF-Graph
- Select causal relations (causal triples) and safe user input
- Visualize causal graph and perform causal inference
- Documentation of the work
The objective of this master thesis is to find an approach or combination of approaches to solve the previously mentioned problem in the context of knowledge demonstration and causal inference especially for SMEs. This particularly includes the state of the art regarding knowledge representation and causal analysis. The demonstration of feasibility with an implementation demonstrator of the concept is part of this thesis as well as a suitable evaluation with exemplary use cases.