AI: Enhancing predictive maintenance in production lines

Increasing the service life of equipment machinery and minimising unexpected downtime, can deliver significant reductions in operational costs. This has been supported by application of edge computing and enhanced through the development of Internet of things (IoT) enabled sensors to capture real-time data such as temperature, vibration, and pressure from equipment, allowing faster predictions and immediate actions.

Alongside increasing amounts of production data being gathered across operations, the digital tools to support this have been developing rapidly, with the scope of data becoming wider and the management improved to help with more accurate decision making.

A big advance in predictive maintenance has been the integration of artificial intelligence (AI) and IoT to create smarter, self-sustaining manufacturing ecosystems that support long-term reliability and efficiency of equipment and processes.

AI boosting predictive maintenance capabilities
As with much of the digital development in automotive manufacturing, the adoption of AI is accelerating as companies gain more understanding of its capabilities in supporting and improving operations. Vehicle makers and tier supplies are actively working with AI to deliver more targeted, accessible data to support predictive manufacturing and the related decision-making process.

This is becoming a key tool in Stellantis’ digital transformation of manufacturing. The carmaker’s chief manufacturing officer, Arnaud Deboeuf, shared some insights with AMS into a predictive maintenance project that uses generative AI, which has been in progress for the last two years with positive results and represents a step forward in the approach to predictive maintenance.

“Every night, we download all the problems that we have had in the plants that day, and the processes for solving the problems,” he explained. “All this data is downloaded into the system so, whenever you raise a problem, the system provides specific information on how this has been managed previously in another plant, another country with the same equipment.”

“This is the first step, which is already in place and deployed across 28 locations worldwide, all connecting to this predictive maintenance system. What we now want to develop is how can we have predictive maintenance system that also provides the solution.”

Suppliers are also using AI to support improving data management and accessibility. Vehicle interiors manufacturer, Antolin said it had increased its productivity through IoT technology, using it to capture and monitor process data to perform predictive maintenance and advanced analysis. It is also using AI to improve 3D design and documentation.

“The company is experimenting with AI assistance to improve information queries, report generation and translation of sentences, promoting collaboration and interchange of ideas among employees,” said a spokesperson.

Another example of the application of AI in supporting decision making across plant operations, including predictive maintenance, comes from Fersa Bearings. It’s also a good example of how interconnected this approach can be. The company said it is using what it calls Industry 5.0 technology to implement AI in strategic decision-making in operation planning and optimisation, the application of digital tools in plants, vision systems and automated quality control, as well as developing digital twins in its operations to predict failures. Fersa said that combining that technology with human intuition and knowledge was optimising strategic decision-making.

David Schultz, an industry expert and Director for the SMIIoT Division of the International Society of Automation (ISA) offered some explanation of the differences between machine learning and artificial intelligence in the context of manufacturing operations. He observed that machine learning focuses on identifying patterns in data, while AI seeks to broaden these patterns by making models more human-like. He noted the potential for AI to optimise manufacturing processes but stressed that this requires a strong foundation of integrated and contextualised data to be in place.

Schultz highlighted the significant role AI has in predictive maintenance, amplifing the capabilities by providing advanced algorithms, machine learning models, and analytical tools to process large datasets, identify patterns, and make accurate predictions.

Shifting predictive maintenance from cycle-based to condition-based approach
The approach to predictive maintenance continues to evolve. An important trend in this is the transition from traditional cycle-based maintenance to more sophisticated condition-based approaches.

The familiar cycle-based maintenance sees maintenance tasks performed on predetermined intervals, such as time elapsed, usage hours, or production cycles.

However, this approach has limitations in that it lacks flexibility, which can lead to over-maintenance with unnecessary replacement of components, wasting time and resources. And the opposite is also possible in that the condition of equipment can deteriorate faster than anticipated, leading to failures outside of the maintenance schedule due to unaccounted for variations in operating or environmental conditions.

The benefits of condition-based maintenance are that it uses sensors and data to monitor the real-time state of specific equipment and more accurately predict when maintenance is needed. These sensors can collect data on temperature, vibration, pressure, humidity, and other critical parameters. This approach also allows a more dynamic response, with adjustments or maintenance being triggered if anomalies are detected or predefined thresholds are exceeded, and the data collected is unique the operating conditions of each piece of equipment.

AI supported condition-based maintenance
This approach is further enhanced through the application of AI, and it supports condition-based maintenance in several areas. Scale is the first of these with the AI being able to monitor and process data from thousands of sensors mounted on equipment, importantly, in real-time. In doing this the system can identify any patterns in machine condition that might emerge. It can also analyse historical and real-time data, which can create more accurate predictions in relation to machine failure patterns, supporting proactive maintenance measures. This can also be used to suggest corrective measures based on previous case histories.

The AI enable systems can be integrated with IoT devices to enhance monitoring and data accuracy. They can also create alerts to inform the factory teams if equipment condition deviates outside of set parameters, which helps to reduce response times.

AI-driven digital twins can also provide virtual simulations of equipment and processes, enabling manufacturers to test maintenance strategies in a virtual environment before implementation.

Several vehicle makers are using this AI supported approach to predictive maintenance. For example, BMW has been using an AI supported system at its Regensburg plant to monitor the assembly line conveyor equipment. The load carriers used to transport vehicles through assembly send various data to the carrier control system which is transmitted via the carrier and plant control system to the BMW Group’s own predictive maintenance cloud platform. The algorithm constantly searches for irregularities, such as fluctuations in power consumption, abnormalities in conveyor movements or barcodes that are not sufficiently legible, which could trigger a malfunction.

Challenges in integrating AI into predictive manufacturing
Despite its benefits, the integration of AI into predictive manufacturing comes with some challenges. AI systems require high-quality data to function effectively. Inconsistent or incomplete data can lead to inaccurate predictions. Additionally, managing and processing high volumes of data from IoT devices and sensors is complex.

The initial investment in AI technologies, including hardware, software, and expertise, can be prohibitive for smaller manufacturers. However, the long-term benefits often outweigh these upfront costs. Implementing and managing AI systems also requires specialised skills in data science, machine learning, and industrial engineering. Many manufacturers face a shortage of skilled personnel to bridge this gap.

Legacy systems are another potential hurdle to integration in that they may not be compatible with modern AI technologies. Upgrading these systems or integrating AI solutions poses technical and logistical challenges. The increased use of IoT devices and AI systems in manufacturing also raises concerns about data security and vulnerability to cyberattacks. Manufacturers must implement robust cybersecurity measures to protect sensitive information.

The future of AI in predictive manufacturing
The development and adoption of condition-based maintenance supported by AI systems is a significant step forward in predictive maintenance strategies. As AI technologies continue to evolve, their role in predictive manufacturing is set to expand. Emerging trends include

autonomous manufacturing with AI-powered robots and systems being capable of making decisions and optimising processes without human intervention. But AI will also support collaborative work alongside human operators enhancing productivity and decision-making through shared intelligence.

More developments with edge computing will allow processing data closer to the source, to reduce latency and enhance real-time decision-making. AI is also now seeing serious interest and investment in R&D across industries and from governments seeking to gain a technical advantage. So, AI has seen a dramatic shift from the realms of science fiction to the reality of industrial manufacturing and beyond and we’re only seeing a fraction of its potential being realised.