Hella Gutmann – Finding the Right Spare Part with a Click

Hella Gutmann Solutions has been providing solutions for the repair and maintenance of cars and motorcycles since 1968, primarily collaborating with independent workshops. Within the AI Marketplace, the company focuses on AI-supported vehicle diagnostics to automatically detect faulty components across a variety of vehicle brands and support workshop personnel.

Traditionally, potential faulty components in vehicles are identified based on error codes and sensor readings. Mechanics in workshops typically require significant time and extensive automotive knowledge to create a comprehensive diagnosis based on retrieved error codes or measured sensor values. Hella Gutmann’s devices perform 66 million automotive diagnostics per year, generating data from 35,000 vehicle models from various manufacturers, including information on fuel injection quantities and mileage. “This large amount of data has led us to explore where we can apply AI methods to simplify and optimize vehicle diagnostics,” says David Aymanns, Head of Data Science at Hella Gutmann.

In the process, vehicle and diagnostic data are collected and stored in a database through an Extract-Transform-Load (ETL) process implemented at different locations in the workshops. In collaboration with experts from the AI Marketplace, a machine learning model has been developed and is now being trained using the collected data. Simultaneously, the predictive model is gradually optimized through feedback from automotive experts and integrated into a demonstrator for AI-supported vehicle diagnostics.

No AI without Data Engineering

However, before a predictive model can be developed, it is essential to identify relevant data sources, validate and cleanse the collected data for completeness and changes, i.e., preprocess it. “From our experience, data engineering, which involves preprocessing the data, is a crucial step towards AI and constitutes a significant part of our work in the project,” explains Aymanns.

At Hella Gutmann, critical parameters for fault diagnosis are already known, as are tolerance ranges for error codes. With this information, the AI model can now detect anomalies in the data and diagnose faults in vehicles. Currently, the predictive model is being tested and validated in practice with the support of automotive experts.

No Artificial Intelligence without Human Intelligence

The so-called “active learning” approach is employed, where the machine learning model suggests a possible fault diagnosis to employees, who can review and potentially correct it.

It is crucial to understand the introduction of AI as a cultural change and to involve employees in transparent dialogues when implementing AI methods. This approach helps to address any reservations towards automation processes, enabling harmonious cooperation between human and artificial intelligence. “The AI Marketplace has provided us with information resources that helped us gain a comprehensive and well-rounded understanding, especially regarding the ethical and legal framework for introducing AI,” adds Aymanns.

Product Innovation with the Involvement of the AI Marketplace

The Hella Gutmann “Automated Diagnosis” (AD), developed with the participation of the AI Marketplace, will be integrated into the mega macs Diagnosis Software SDI by the end of 2022 and rolled out to SDI customers. The system continuously improves itself by leveraging and integrating feedback from automotive experts. Currently, the AD feature is based on around 2 billion historical error codes and approximately 5 million statistically recorded causalities from Hella Gutmann’s Technical Call Center. With this data, defective components can be identified with high probability in over 80% of diagnosed cases. The “Automated Diagnosis” enables automotive workshops to save significant time through more efficient workflows while improving diagnosis quality.