Video Summary
In this CSIA Masterclass, Lisa Richter, Director of Industry Outreach and Growth at CSIA, introduces the association’s mission to help systems integrators (SIs) scale businesses through best practices, certification programs, and educational resources like the Learning Hub and CSIA Industrial Automation Exchange. Membership offers access to virtual events, in-person conferences, and peer insights via the Talking Industrial Automation podcast.
Marc Bertrand, Director of Industry Solutions at SmartSights, presents on optimizing production using AI and machine learning (ML). Drawing on 25+ years as a systems integrator, Bertrand emphasizes boosting regression models for predictive, descriptive, and prescriptive analytics. These models handle complex, multivariate production systems, uncover true bottlenecks, and deliver highly accurate predictions for throughput improvement.
The session highlights the integration of large language models (LLMs) like ChatGPT to enhance accessibility, query ML models via natural language, and automate workflow creation. By combining ML and generative AI, SIs can optimize production lines, improve KPIs, reduce downtime, and balance rate versus quality.
Bertrand underscores the importance of high-quality, feature-rich utilization data, proper sample sizing, and pre-cleaning to ensure predictive accuracy. Key categorical variables like SKU, shift, and environmental factors (e.g., temperature, humidity) can be included to refine outputs. Prescriptive and predictive analytics allow “what-if” scenario modeling to guide targeted interventions, from adjusting machine rates to reducing stoppages.
Practical use cases extend beyond throughput optimization to quality control, setpoint optimization, and daily production reporting. SIs can leverage these AI-driven insights to enhance operational decision-making, implement line improvements, and deliver measurable results to clients.
For SIs looking to adopt AI/ML solutions, Bertrand recommends partnering with organizations experienced in ML deployment and data curation, using pre-built templates, and focusing on actionable outputs rather than reinventing models for every project.