Seebo: Modeling the Production Line with Machine Learning for Accurate and Actionable Predictive Insights
One startup is leveraging Industrial AI to create solutions targeted at process manufacturers. “Infusing process flow data into machine learning algorithms lets manufacturers predict and prevent costly manufacturing disruptions, while by-passing weeks of investigation time,” says Lior Akavia, co-founder and CEO of Seebo.
With its Industrial Internet of Things (IIoT) platform, Seebo is pioneering process-based industrial AI solutions to anticipate and mitigate disruptions in manufacturing processes. The process-based approach to AI enriches a manufacturer’s relevant IT and OT data with production line process flows to deliver accurate and actionable insights to production floor staff and management.
The Seebo platform consists of four modules: modeling and simulation, process-based predictive analytics, automated root cause analysis, and digital twin analytics. Users model their production line on the platform - including machines, equipment, sensors, OT and IT data sources, and process data flows. The system then generates a process-based data schema, upon which a digital twin prototype is created to help validate the solution before full implementation.
Seebo’s platform captures production line data into its process-based data lake, and applies machine learning algorithms - including Random Forest, Support Vector Machine, and Recurrent Neural Networks - to identify anomalies and predict problems in production quality and uptime. Automated alerts and customized Digital Twin dashboards deliver actionable insights to the workforce on the production floor.
The final piece of the puzzle, digital twin analytics, paints the larger picture for manufacturers. It serves as a clear, visual representation of the manufacturing KPIs, alerts, and reports, contextualized with production operations. The Digital Twin enables manufacturers to view the line on an operational level, along the AI insights, to improve overall equipment effectiveness (OEE). “The representation of the production line provides manufacturing managers with complete operational intelligence, right from the site level, through production lines, down to specific machines and individual sensors,” adds Akavia.
The applicability of Seebo’s process-based AI is best illustrated through the company’s recent deployment at a large food manufacturer. The client was faced with a problem in the manufacturing line, which produced batches with broken wafers. While the client’s investigation team identified the problem and associated its root cause with overheating in the ovens, the suggested solution of lowering the production temperature proved to be ineffective. Once the Seebo solution was implemented, it accurately identified the source of the problem by highlighting discrepancies in the combination of temperature and slight jumps in conveyor belt speed. The Seebo solution also alerts on such issues ahead of time, enabling the client’s production staff to prevent the issue from occurring. The end results: 10% reduction in production waste, and 45% reduction in unplanned downtime.
This example highlights two critical factors that set Seebo apart from its competitors: The accuracy of insights derived by its process-based AI, and its ease of use, enabling manufacturers to harness its insights without mastering data science. “While there are companies that offer AI solutions to streamline manufacturing processes, they often overlook the importance of process-based contextualization, resulting in many false positive alerts that hinder production. Data pertaining to manufacturing processes — the dependencies between machines, data flows, and the product flows per recipe — must be at the core of any solution for process manufacturing, and that is what we do best,” concludes Akavia.