Skip to content

AI to Deliver the “Golden Batch”

  • today
  • access_time 2:50 - 3:20 PM ET
  • location_onSmart Zone
  • blur_circularSmart Zone Theater Presentation
  • monetization_onIncluded with Exhibits Registration

A lot has been written about AI relative to manufacturing but except for predicting the health of machines in the asset intensive process industry, little can really be shown in terms of results.  Technologies from the process industry, which has been an early pioneer in the use of AI, are being used to help manufacturer’s maintain high quality at every step of the process. This is a key driver to reduce scrap and achieve high throughput. Early detection of quality issues helps with containment of poor batches and aid in root cause analysis to remediate manufacturing issues, thus delivering the Golden Batch.

With traditional methods of rule based static process control methods, manufacturing errors often go undetected or unreported and are propagated. Application of Statistical Process Control (SPC) has persisted for close to a century now.  However, SPC faces its challenges.  For instance, SPC can be applied only once a statistically significant sample set is collected and analyzed.  This leads to a significant lag of corrective actions being available only after process completion. Secondly, SPC methods are univariate in nature that do not capture the multi-variate dynamics of time-series data. Even though things look good in SPC charts, we see unforeseen quality issues.

To get proactive in predicting batch and quality outcomes, new methods are being implemented. Manufacturing lines emit large amounts of data that characterize the manufacturing process ranging from simple temperature or pressure settings on a line to other more complex settings such as the size of components used, the surface area, etc.  Until recently this data has not been utilized to drive automated analysis.

With the advent of IIoT Platforms and more modern historian data sources, this data can now be captured and cataloged more accurately. By utilizing new AI tools available, manufacturing organizations can predict the outcome of a batch well before the manufacturing process is complete.  AI tools can analyze huge volumes of data across multiple variables and establish deep relations between input process variables and measures of the batch. With deep learning AI models, that uses multiple input variables collected from process of manufacturing (KPIVs) and correlating it with the critical output variables of the batch (KPOVs), a definition of the “Golden Batch” can be established.  Once this correlation is established, the quality of every batch is predicted in advance by monitoring KPIVs with automated recommendations for optimal control.

Symphony Industrial AI has been working with numerous industrial customers in application of AI in various manufacturing operations.  In this talk, we will discuss a real-world case study where one of our customers is relying to AI to make decisions that drive autonomous manufacturing.