The use cases address the engine block manufacturing process, in terms of detecting and predicting defects in aluminium injection and machining operations. Two industrial partners are involved in this use-case – the Martinrea Honsel (MRHS), located in Madrid, specialized in manufacturing of aluminium cylinder blocks and FORD Valencia Engine Plant finishing the rough Cylinder Blocks received from MRHS to produce engines. The main types of defects occurring are categorized into two groups: porosity and leakage. The goal is to improve the quality through the scrap output reduction, but also to provide the quality predictions.
The process data required for anomaly detection and production process optimization are gathered from multiply sources both in MRHS and in FORD and aggregated to predict the quality of the product and a rejection probability. Based on the data gathered, the model for parameters’ optimization is generated to achieve a certain, user-defined, objective.
Sse-cases offer an additional services for quality assessment and prediction based on the aggregated data from the participating industrial partners (Martinrea Honsel (MRHS) and FORD) that will be provided by the ZDMP platform. In this case, the service covers only specific product lifecycle stage, when the product – aluminium cylinder blocks, is produced and arrives at FORD to be used in engine production, as the defected blocks are filtered both at MRHS and FORD. Thus, the goal is to reduce the amount of defected parts through collaboration between two industrial partners.
In UC 1.1 / UC 1.2 use-cases, both industrial partners exchange the data from the various end devices (sensors) in order to achieve the cumulative effect in terms of defects reduction. Thus, the MRHS – supplier of aluminium cylinder blocks, can benefit from the data sharing between the MRHS and FORD to improve the quality of products needed for the engines assembly.
The service proposed doesn’t require big hardware installation, but rather makes use of the data gathered from already installed field level devices (sensors, smart meters, cameras) having the supportive or supplementary role in the production process. However, the added value is in reduction of the number of defected parts or products through extensive utilization of the data-rich environment.
For the quality assessment and prediction service is crucial to aggregate the data from both the Martinrea Honsel (MRHS) and FORD industrial partners. The quality forecasting or prediction allows optimising and adjusting the production process to achieve better quality of the product – aluminium cylinder blocks, which is not possible or less accurate without digitalized data exchange between industrial partners.
ZDMP platform supports the industrial partners by collecting the process data from MRHS and FORD and providing reasoning to be able to detect and predict anomalies and provide some basic recommendation on improvement of the manufacturing process on the MRHS side. Both companies make use from this service, as they are long-term value-chain partners and interested in the costs and scrap output reduction resulting in the improved manufacturing process.
To be able to make prediction and automated quality assessment, process data need to be gathered and presented in the form suitable for processing. Process data are gathered from various sensors and smart meters, as well as from PLCs at MRHS and automatically uploaded to the database. As the production cycle takes around 2 minutes, subsequently data are uploaded every 2 minutes. The ultimate goal is to receive the anomaly warnings close to real-time.