CastControl: Reducing production errors in casting processes

Accurate AI-based recommendations in the manufacturing process

© Fraunhofer IIS

In the project »CastControl«, a collaboration between the Center for Applied Reasearch on Supply Chain Services, the Development Center for X-ray Technology at Fraunhofer IIS, and Ronal AG, manufacturing processes are optimized using AI. The focus is on creating an AI-powered model that is based on neural networks and combines production and X-ray inspection data. This allows not only for accurate identification of production defects in casting processes but also for the prediction of potential defects. Additionally, the model provides recommendations for process optimization. »CastControl« significantly reduces scrap rates, improves process efficiency, and lowers manufacturing costs.

Comprehensive data analysis for improved prediction accuracy and optimized recommendations

To increase the prediction accuracy of the model, extensive data analysis was conducted, including the evaluation of over 16,000 sensor data records from the casting process. These data were used to understand the relationship between X-ray defect features and the quality of the castings. The results of these analyses were directly incorporated into the training data for the neural network, thereby laying the foundation for the generated optimization recommendations.

Use of »Explainable AI« for targeted adjustments

A unique aspect of the project is the use of »Explainable AI« (XAI) techniques. With XAI, it is possible to understand the decision-making process of the neural network. The influences of individual sensor data on predictions can be visualized and interpreted, allowing for targeted adjustments of production parameters. For example, XAI can show which sensor value should be increased or decreased to minimize the defect size.

Cast Control - The next step in quality control and process optimization in production

The CastControl project takes the application of artificial intelligence to quality control and process optimization in production to a new level. In a follow-up project, the predictive reliability is now being further improved and a prototype for real-time recommendations during production is also being developed. Through collaboration with additional industry partners, the model will also be expanded to cover additional production processes. This expands the application scope and increases production quality across industries.

That might also interest you


AI-based demand forecasts for logistics, retail and transport

We bring AI-based demand forecasting to logistics, retail and production to improve predictions and quantify forecast uncertainties.