Speaker
Description
The production of cast-iron disc brakes demands tight control over metallurgical and process parameters, as defects can compromise critical product properties. Defects compromising mechanical strength, alter natural frequency response, promote the formation of cementite and cause micro-shrinkage porosity, which represent frequent quality issues in ferrous metal foundries. They are particularly critical since their root causes lie in the liquid metal processing stage, but their effects often become detectable only through downstream testing. This results in production waste and rework costs, which can be mitigated through anticipatory digital tools.
This work presents a Decision Support System (DSS) designed for foundry environments, integrating machine learning models to enable predictive quality control in cast-iron manufacturing. The considered process includes cupola furnace melting, secondary refining of liquid metal through ferroalloy additions, and a final inoculation stage during mould filling. The DSS continuously ingests real-time process data, including chemical and thermal analysis of the refined melt, pouring temperature, inoculant flow rate, and casting duration, useful for monitoring the current production state and feeding predictive models based on Artificial Intelligence (AI). The models are trained on historical data and forecast the occurrence of specific casting defects such as micro-shrinks, cementite and deviations on natural frequency, up to three to four hours in advance compared to the production window. This predictive horizon is operationally significant, as it provides process engineers and metallurgists with actionable time to adjust melt chemistry, refine inoculation parameters, or intervene in secondary treatment before non-conforming parts are cast.
The system was validated against historical production data from an industrial foundry environment, demonstrating reliable predictive accuracy across multiple defect categories. This work demonstrates the concrete applicability of AI-based approaches in conventional production route, contributing to the ongoing digital transformation of foundries and offering a replicable framework for quality-oriented predictive systems in similar metallurgical contexts.
| Speaker Company/University | Scuola Superiore Sant'Anna, TeCIP Institute |
|---|