Application of Industrial AI: A manager's guide.
Colin Koh (許国仁)
Founding Member @ AI Asia Expo | Industry Leader | Book Author/Evangelist/Enthusiast on Smart Manufacturing | Industrial AI | OT Cybersecurity
The ISA-95, or the automation pyramid, provides a useful framework for understanding how data flows through a manufacturing operation and how AI can leverage at each level. Here's a breakdown of the data value and potential AI applications at each level, considering edge, on-premise, hybrid and cloud computing:
Levels 0 & 1: Sensors & Production Process Signals (PLC)
· Data Value: This level captures real-time data from sensors embedded in machines and the production process. Examples include temperature readings, vibration data, robots and production line speeds.
· AI Applications (Edge): Due to limitations in processing power and bandwidth at the edge, this level is ideal for simple AI tasks like:
o Anomaly Detection: AI algorithms can continuously monitor sensor data to identify deviations from normal operating ranges, potentially indicating equipment malfunctions or inefficiencies.
o Predictive Maintenance: By analysing sensor data over time, AI can predict when equipment will likely fail, enabling proactive maintenance and preventing costly downtime.
Level 2: Monitoring & Supervising (SCADA/HMI)
· Data Value: This level aggregates and visualises data from Levels 0 and 1, giving operators a real-time overview of the production process.
· AI Applications (Edge/Premise): With more processing power than Levels 0 and 1, this level can handle more complex AI tasks like:
o Process Optimisation: AI can analyse historical and real-time data to identify areas for improvement in the production process, such as optimising equipment settings or scheduling maintenance for minimal disruption.
o Quality Control: AI can use image recognition or other techniques to inspect products for defects in real-time, improving quality control efficiency.
Level 3: Manufacturing Operations Management (MES)
· Data Value: This level integrates data from Level 2 with higher-level planning systems, providing insights into production performance, resource utilisation, and quality control.
· AI Applications (On-Premise/hybrid/Cloud): This level leverages larger datasets and more powerful computing resources, enabling AI for:
o Production Scheduling: AI can analyse real-time data and historical trends to optimise production schedules, considering demand forecasts, machine availability, and material constraints.
o Yield Management: AI can identify factors impacting production yield and recommend adjustments to optimise output and minimise waste. (OEE implementation)
Level 4: Business Planning & Logistics (ERP)
· Data Value: This level focuses on enterprise-wide data, including production information, inventory, financials, and customer relations.
· AI Applications (hybrid/Cloud): hybrid/Cloud computing offers vast data storage and processing capabilities, ideal for AI applications like:
o Demand Forecasting: AI can analyse historical sales data, market trends, and external factors to predict future product demand, enabling better production planning and inventory management.
o Supply Chain Optimisation: AI can analyse logistics data to optimise transportation routes and inventory levels across warehouses and identify potential disruptions in the supply chain.
Key Considerations for AI Implementation:
· Data Availability and Quality (7V of Data): Reliable and clean data is essential for effective AI applications at any level.
· Connectivity and Security: Ensure secure and reliable communication between devices, controllers, and AI systems at the edge, on-premise, hybrid and cloud.
· Scalability and Adaptability: Choose AI solutions that can scale with your data volume and adapt to changing manufacturing processes.
Manufacturers can gain valuable insights, optimise operations, improve efficiency, and gain a significant competitive advantage by understanding the data value at each level of the ISA-95 pyramid and strategically applying AI at the edge, premise, and cloud.
AI is dynamic, and changes daily; more Industrial AI-related newsletters will be extracted from my new book, "Industrial AI: Applications in Smart Manufacturing," which will be released in May 2024 at Amazon.com. So stay tuned... and subscribe to Smart Manufacturing Newsletters.
Distinguished Professor in Wholistic Learning | Founder WILL | Academic Fellow CMC | Senior Research Fellow RCTI, Tsinghua University | Author | Inventor Wholistic Thinking & Project-Based Accelerated Action Learning
7moHow to purchase your book?
Grid Software ASIA Commercial and APAC Alliance Leader GE Vernova
7moISA still remains a valuable standard but more for easy framework referencing, the definitions inside the standard should be reviewed and modified to suit the industry move especially with AI being the ramping force. Colin Koh (許国仁) it’s great to highlight the criticality of data readiness which will be the key area to tackle in the foreseeable future. Security will be another area for deep dive if we unpill the onion. Zero-trust will be a long journey requiring DNA level reengineering and fusion with the applications. Also agreed on adaptability which means the use case in the context of industry and verticals (discrete, batch and process).