The buzz around Artificial Intelligence (AI) has been hard to ignore, and it’s not just the trade publications fueling the hype. As someone who has been cautiously optimistic about AI, I can’t help but feel myself leaning more towards a pro-AI stance as I witness its transformative impact. With the advent of cutting-edge AI-based tools like ChatGPT, it’s clear that AI is no longer just a futuristic concept, but a powerful tool that’s already changing the game.
While AI has been steadily making its way into automation applications, particularly in vision systems, there are countless other use cases for machine learning and AI at the edge. For years, the limitations of memory, processing power, physical form, and price have hindered progress at the edge. However, with advancements in PLC products, these barriers are rapidly being overcome. The question now is, what data do you need to effectively implement an AI application at the edge?
To answer this question, the table below was compiled based on AI applications in industrial automation application.
| AI Application | Data Points Needed for Implementation |
| Predictive maintenance | Real-time sensor data, historical maintenance records, equipment operating conditions, and environmental conditions. |
| Fault detection and diagnosis | Sensor data, historical data on faults, and their root causes, machine operating parameters, environmental conditions. |
| Quality control | Real-time sensor data, historical quality control data, product specifications, process parameters. |
| Energy optimization | Real-time sensor data on energy consumption, machine operating conditions, historical energy consumption data, environmental conditions. |
| Production planning and scheduling | Production data, inventory data, supplier data, machine data, order data, production schedules, and delivery schedules. |
| Asset tracking | Real-time sensor data on asset location, usage, and maintenance records. |
| Supply chain management | Historical sales data, real-time demand data, inventory data, supplier data, shipping data, transportation data. |
| Process optimization | Real-time sensor data on machine operating conditions, production data, quality data, historical process data, environmental conditions. |
| Root cause analysis | Historical production data, fault data, maintenance records, quality control data, environmental data. |
| Machine learning-based control | Real-time sensor data on machine operating conditions, historical machine data, environmental conditions, and production data. |
| Automated decision-making | Real-time sensor data, historical data on machine performance, historical maintenance records, production schedules, and delivery schedules. |
| Real-time monitoring | Real-time sensor data on machine operating conditions, production data, quality data, and environmental conditions. |
| Autonomous robots | Real-time sensor data, machine operating data, production data, environmental data. |
While this list is a higher level view of relevant data for different applications, there are other data factors. These include data quality, quantity, and accessibility, as well as considerations around data privacy and security. In some cases, a data storage device may be needed from which data is pulled back into the PLC as needed. Also ,the elephant in the room is what type of algorithms and AI functions can a PLC actually execute. More to delve into on all these points. One thing is for sure, the final solution is going to be much more involved than a two letter acronym (AI) infers.
Moving forward, my aim is to be concise and informative in posts. If you’re interested in learning more about the trends, challenges, and opportunities in AI implementation at the edge, be sure to subscribe. In future posts, we’ll explore other aspects of AI at the PLC, and I welcome your comments and requests for specific topics.
