84 % less Scrap and Rework: Cost Reduction in Cylinder Head Production thanks to Causal AI Algorithms from Munich-based startup
Munich, Dec. 1, 2022 – By using AI algorithms from Munich-based startup Xplain Data GmbH, Schwäbische Werkzeugmaschinen GmbH (SW) has succeeded in reducing scrap and rework rates by 84% in the production of cylinder head covers. Xplain Data provides a causal discovery process that uncovers previously unknown causes of defects and optimizes production processes.
The supply chain around Schabmüller Automobiltechnik GmbH, a Bavarian automotive supplier with a focus on series production of high-precision components, benefited from this. By eliminating the identified causes, the production defect rate in cylinder head cover production was reduced to 1.6%.
It turned out that leak testing in the manufacturing process is strongly dependent on the component temperature after the washing process: A short waiting time between these steps is crucial. Furthermore, significant quality differences were found in the casting nests of the supplied blanks. Due to the complexity of the plant, this influencing factor was previously difficult to test.
“The manufacturing industry wants to use its extensive production data to understand cause-and-effect relationships and solve problems. SW took the plunge and implemented our AI algorithms – and won,” said Dr. Michael Haft, CEO Xplain Data GmbH.
Schabmüller plans to further automate the algorithms used:
- AI-Bot: Counteracting new sources of error early on through ongoing monitoring.
- Integration of additional data along the life cycle of a workpiece into the analysis: from material composition during casting to further processing and final assembly at one of the leading German automotive manufacturers.
The goal is a 360° perspective. Full SW case study.
Causal relationships as a critical factor in process optimization
Since 2015, the Munich-based startup Xplain Data has been developing algorithms that can detect causal relationships in “real world data” (causal discovery) and use them for intelligent interventions. For example, causes of errors can be eliminated and desired effects can be generated. The prerequisite for this is an object-centric consolidation of all company data. This is accomplished by Xplain Data’s Object Analytics Database.
Xplain Data achieved initial success with “Root Cause Analysis” in the healthcare sector. Now, in collaboration with SW, Xplain Data has been able to transfer analysis methods from the healthcare sector to production data for the first time: the aim here is to identify and eliminate the causes of production errors and to perform predictive maintenance on machines. The first real-world application in industry turned out to be a great success, which is now being followed up.
Also applicable in industry: AI techniques from healthcare
Using the patented Xplain Data methods, it was possible to distinguish potential causal relationships from ordinary correlations in the healthcare environment, based on extensive patient data. For example, it identified unwanted side effects of medications, uncovered risks for future illnesses, or assisted insurance providers with CRM, rate design, and quality of life offers.
The challenge behind this is similar: millions of patients must be correlated with countless prescribed medications, diagnoses made, hospitalizations, and numerous different data sources.
Xplain Data GmbH
Marketing & Communication
About Xplain Data
Xplain Data GmbH, founded in 2015, focuses on the development of innovative technologies in the field of Causal AI. Xplain Data algorithms enable companies in all industries to identify the few, potentially causal relationships in their “real world data” that are hidden behind a plethora of trivial correlations. Users can use these cause and effect insights to intervene in their business processes to eliminate causes of errors or achieve a desired effect. Xplain Data customers include leading companies in mechanical engineering and manufacturing as well as in healthcare, which use the technologies not only for sophisticated data analyses but also, for example, for production and yield optimization as well as care management analyses.
- Example from production: Representation of the production flow in order to track faulty parts
- Michael Haft, CEO Xplain Data GmH