How AI and ML Affect Industrial Automation

Summary

Artificial intelligence (AI) and machine learning (ML) are rapidly changing many industries. While these two terms are often—and incorrectly—used interchangeably, machine learning is a subclass of artificial intelligence. Put simply, AI is an overarching concept of creating intelligent machines that can perform tasks that generally require human intelligence, including problem-solving, pattern recognition, natural language processing and decision making. Machine learning is a specific approach within the broader field of AI that focuses on training machines to learn from data, and to make improvements in processes without being explicitly programmed to do so. 

Although industrial automation has been one of the major drivers of Industry 4.0, the recent advances in AI and ML technology represent a significant shift away from traditional automation processes and methodology. This shift is quite clear in adaptive automation; programming, system design and flexibility; and decision making. 

Rule-based automation versus adaptive automation 

Traditional automation systems rely on rule-based programming. In this framework, machinery and technology are programmed to respond to a set of predefined if-then-else type rules. Engineers and developers explicitly define rules and instructions for the machinery and technology to follow. Rule-based automation is effective for simple processes, but this type of system has limited adaptability when presented with unforeseen scenarios and can struggle in complex, changing environments. 

Adaptive automation systems, specifically those that use AI and ML, can learn from data and adapt to rapidly changing conditions. Adaptive systems can make decisions based on patterns and trends in the data without being explicitly programmed for every possible scenario. 

Programming, system design and flexibility 

Traditional automation systems require a large amount of upfront programming and coding. The if-then-else type rules must be explicitly programmed into the system. Essentially, every potential action and response from the automated machinery must be predefined. This approach is acceptable for very simple systems with one or two possible outcomes. However, as systems become more complex, the amount of coding and programming increases nearly exponentially. Further, the functionality of traditional automation systems remains static unless they are manually reprogrammed. This rigidity has the overall effect of limiting the flexibility and adaptability of traditional automation systems. 

Alternatively, automation systems powered by AI and ML can rapidly adapt to changes, often with little or no human input. Adaptive automation systems are driven by ML algorithms that enable continuous adaptation and process optimization without the requirement of reprogramming or reengineering. 

Decision making 

Perhaps the most important difference between traditional rule-based automation and adaptive automation lies in how each respond to novel or complex problems and edge cases. Traditional automation systems only perform tasks or processes when given explicit rules or commands. As the complexity of a system increases, traditional systems struggle to adapt to dynamic environments. These traditional systems are “dumb” in the sense that they only react based on defined logic. 

Adaptive automation systems powered by AI and ML can impart a measure of device intelligence into these formerly “dumb” systems and devices by making use of data gathered over time. Whereas traditional systems are reactive and follow predefined logic, adaptive systems can refine and optimize processes and react to unforeseen scenarios. By analyzing the data and learning patterns, adaptive systems can handle complex decision making. Through analysis of historical data, these systems can anticipate and predict component or subsystem issues that enables proactive maintenance. According to the International Society of Automation (ISA), adopting a preventive maintenance approach can provide savings from 8-12% over reactive maintenance and can reduce equipment and machinery downtime by 35-45%. 

AI and ML are changing automation 

The use of AI and ML is dramatically changing the industrial automation industry. The traditional approach to automation, which relies on if-then-else, rule-based programming, is limited in terms of flexibility and its ability to respond to novel scenarios that fall outside of explicitly defined parameters. However, the shift toward adaptive automation systems, powered by AI and ML, brings increased flexibility, predictive capabilities and the ability to handle complex decision making, contributing to more efficient and responsive industrial processes. 

Several components make up an industrial automation system using AI and ML. These include sensors or other data gathering and data acquisition devices, data storage, a processing component, AI and ML models and algorithms, control systems that translate the AI and ML decisions into actions, a human-machine interface (HMI) and a communications network. These complete solutions are being deployed across industries, from improving ADAS systems in automotive manufacturing to data acquisition in materials production to communications for pharmaceutical testing. 

_{This feature originally appeared in the March edition of AUTOMATION 2024: IIoT and Digital Transformation.}

Download AUTOMATION 2024