Artificial Intelligence and Machine Learning in the Manufacturing Space

By Dennis Hodges, CIO, Inteva Products

When many of us think of Artificial Intelligence or AI, we jump to ChatGPT and the ability to ask a question to see how smart the technology really is. I hope that my competitors never get beyond this point! We get overwhelmed by the macro possibilities that we often ignore the micro benefits that can come with the technology.

I focus this article on the manufacturing environment. I am in the automotive production space; our company provides a variety of products to car makers globally. In the US alone, the automotive market contributes 3 – 3.5% to the national GDP. Overall, manufacturing contributes around 12% to the US GDP. These are major pieces of the pie that often get overlooked or ignored – technically and from the jobs perspective.

I have been fortunate to work with several VC groups over the past fifteen years. They bring some very interesting technology up for review. Almost universally, however, the companies miss the mark when it comes to fit and function for the demanding manufacturing environment – precision and repeatability.

From the jobs perspective – people are less interested in pursuing opportunities in the manufacturing space. As the experienced skilled trades workers prepare to retire, there is a tremendous shortfall in the required personnel to replace them. The move to offshore manufacturing positions beginning with NAFTA did not help the situation. As we attempt to re-shore work to the US, we will see a dearth of talent.

Along with the intelligent factory, changes to the workforce are critical. The average car, truck, or SUV produced now often contains a much richer mix of content than in the past.

How do AI and Machine Learning (ML) fit into this conversation? There are several critical elements of the industry that will open the door to these technologies that have already emerged. In my world, these are some of those critical elements:

  • Machine & equipment monitoring and management
  • Production visibility and oversight
  • Employee assistance, education, and cooperation

My company has been using one plant-oriented AI/ML tool for several years to aid in a complex production line. Sight Machine provides a product that develops a model of the process and can monitor machine operating and production parameters. The system also learns to identify optimal parameters and can alert when these parameters are drifting away from the optimum. The ability to analyze and predict when and how to operate the equipment begins to rival the experienced tradesman’s understanding of the process. This becomes critical as the skilled worker retires. A cooperative relationship between the system and the operator can provide a strong environment for success.

Additionally, this can be leveraged to show these results across the organization. Plants that utilize identical equipment can start to visualize where the best performers are. Then the company can begin to mimic those performers to improve efficiency and productivity across the organization.

Similarly, production results for assembly and other processes can be retrieved from systems and intelligent interfaces rather than manual data entry. Downtime reason codes that are immediately fed into the visualization system allow for rapid response and improvement. Patterns of behavior and performance variation can be understood systematically to create improvements locally and across the company. No longer will Center of Excellence staff be required to spend long periods in remote facilities to debug operations – if the intelligent systems are fed accurate data in real-time.

Along with the intelligent factory, changes to the workforce are critical. The average car, truck, or SUV produced now often contains a much richer mix of content than in the past. Differentiators in look, feel, and driver/passenger interaction are driving the winners in the auto market. This moves the market to produce “semi-custom” vehicles. From the manufacturing point of view, complexity is increased exponentially.

New materials that are thinner, lighter, and often more luxurious are more difficult to manage and manipulate. The assembly worker must learn the tricks of handling such products or risk introducing scrap problems. Training the operators just in time to produce a variant effectively is key. Training room exercises cannot accomplish this, there can be too many variations to train so far in advance. Short instruction right on the line as the product is in line allows for new possibilities for staff with less experience.

Understanding the skill level of staff and assigning work at the beginning of a shift without significant “tribal knowledge” from the supervisor is becoming more important. Using smart work centers to allow or deny a badged operator to run the station based on their level of certification provides an audit trail that can be reviewed. Online training and certification records combined with production and personnel planning tools allow the shop floor leaders to assign tasks closer to the time of production – a task too complex for manual manipulation if absenteeism issues arise.

A cooperative operating ecosystem with personnel and digital production assistants will allow the enterprise to adapt to the constantly evolving environment. Improving efficiency and driving unnecessary costs out of the operating processes can aid in ensuring the viability and success of the organization.