Historical and locally tuned model for cycle-time prediction and optimization
Introducing a predictive model combining historical machine learning and locally trained time-series models to optimize injection molding cycle time. This hybrid approach enhances accuracy and adapts to unknown factors, offering improved production efficiency.
Overview
The proposed solution is a cutting-edge predictive model that combines historically trained machine learning models with locally updated time-series models to optimize cycle times in injection molding processes. This hybrid approach aims to enhance the accuracy of predictions by adapting to unknown factors such as equipment degradation and unmeasurable material properties. By leveraging both historical data and local sensor readings, the model improves cycle time predictions and subsequently optimizes production settings for efficiency.
Technical specifications
- Historical Machine Learning Models: Trained per preform design and machine generation, focusing on key variables such as extruder temperature, screw speed, and material properties.
- Local Time-Series Models: Continuously updated with recent sensor data to refine predictions and adapt to changes in setup or material properties.
- Optimization Routine: Utilizes trained regression models to search for optimal settings, reducing cycle time and improving overall efficiency.
- Offline and Online Phases: Includes a planned three-phase validation to assess and implement the model across different preform designs.
Technology readiness level
The technology is currently at TRL 4, having been validated in a lab environment with plans for further testing and refinement through structured offline and online phases.
About Predictronics Corp.
Predictronics Corporation provides an industrial AI and predictive analytics software platform called PDX, designed to assist organizations in monitoring, analyzing, and predicting the health of critical industrial assets. The platform uses a template-driven approach that integrates data acquisition, model-building, and asset-monitoring components to help companies move from reactive maintenance to proactive, data-driven decision-making. By leveraging artificial intelligence and machine learning, their solutions enable users—including those without extensive data science expertise—to visualize performance, identify anomalies, and forecast potential failures or quality issues before they result in significant downtime or production waste. The company’s services are supported by a team of experts that combine industrial domain knowledge with professional analytics to develop and integrate custom predictive solutions tailored to specific operational needs.
This technology is critical for industrial sectors such as manufacturing, automotive, aerospace, and energy, where minimizing unplanned downtime and maximizing operational efficiency are essential for profitability. Predictronics has demonstrated the effectiveness of its solutions through various industry engagements, including work with Fortune 500 companies and collaborative research projects on advanced manufacturing technologies. Case studies show significant operational improvements, such as substantial reductions in scheduled maintenance and production stoppage time. By helping customers move toward agile, just-in-time maintenance environments, Predictronics enables them to optimize resources, reduce unnecessary labor costs, and improve overall equipment effectiveness (OEE) across their facilities.
delivered to your inbox.