How AI Can Boost Predictive Maintenance In Manufacturing

What is Predictive Maintenance?

Predictive maintenance is the process of using data and analytics to monitor the condition and performance of equipment and machinery as part of manufacturing processes, and to predict when they might need maintenance or repair. Predictive maintenance can help manufacturers reduce downtime, improve efficiency, optimize resources, and save costs, key factors in the operations of a business. According to a report by McKinsey, predictive maintenance can reduce maintenance costs by 10 to 40 percent, reduce downtime by 50 percent, and increase equipment lifetime by 20 to 40 percent.

How Can AI Enhance Predictive Maintenance?

By enabling more accurate and timely predictions AI can allow manufacturing companies to implement a predictive maintenance process. Having provided actionable insights and recommendations and by reducing the manpower required to achieve this, AI can draw on various data sources, such as sensors, cameras and historical data to create a comprehensive and dynamic picture of the equipment and machinery.

AI can also use advanced techniques, such as machine learning, deep learning, computer vision, and natural language processing, to analyze the data. With a constant state of monitoring, these systems can detect patterns, anomalies, trends, and correlations. AI can then use these insights to forecast the remaining useful life of the equipment, identify potential failures, and suggest optimal maintenance schedules and actions.

Practical Examples of AI Applications for Predictive Maintenance in Manufacturing

As you might expect, AI applications for predictive maintenance in manufacturing, whilst becoming more widespread and diverse, are mostly being led by larger companies. However, as AI tech is becoming widely available, more manufacturers are adopting digital transformation and Industry 4.0 initiatives. Some examples are:

• Siemens uses AI to monitor and optimize the performance of its gas turbines, compressors, and generators. The company uses a cloud-based platform that collects and analyzes data from thousands of sensors, and applies machine learning algorithms to generate predictions and recommendations. Siemens claims that its AI solution can reduce maintenance costs by up to 30 percent and increase availability by up to 99 percent.
• GE Aviation uses AI to predict and prevent engine failures in its aircrafts. The company uses a digital twin approach, which creates a virtual replica of each engine that simulates its behaviour and condition. The digital twin is fed with data from sensors, flight records, weather data, and maintenance logs, and uses deep learning models to detect anomalies and predict failures. GE Aviation says that its AI solution can improve fuel efficiency by up to 1 percent and reduce maintenance costs by up to 10 percent.
• Toyota uses AI to inspect and maintain its welding robots in its factories. The company uses computer vision and deep learning to analyse images and videos of the robots' welding operations, and to identify defects. The AI system can also recommend the best welding parameters and settings for each robot and alert operators when the robots need maintenance or replacement. Toyota reports that its AI solution can reduce inspection time by 70 percent and improve quality by 10 percent.

Predictive Maintenance for the SME

The examples stated in the section above are use cases adopted by some very large manufacturers, so if you are an SME, then you might well be switching off right now, but please bear with me a little longer. Whilst most SMEs won’t have the budget to create ‘digital twins’, there are some factors that you can look at.

Equipment Suppliers

Choose your equipment carefully. Ask your supplier if they have self-diagnostics, remote diagnostics, IoT functionality and whether or not they have invested, or are looking to invest, in AI technology to help with the analysis of the data with the goal of improving uptime and reducing TCO.  Your supplier should have a good coherent answer to the question, and it might be worth paying a little bit more for equipment that is working towards such goals.

Quality Data

‘Quality Data’ could mean the data from your quality measures, or it could mean the quality of your data, in reality both are important.  AI cannot make decent decisions without decent data, so looking now into the quality of your data is a good first step as it will help ensure that an AI can make meaningful assumptions when the time comes.  Data from your quality system is also important, as it is often product quality that shows the first signs of maintenance issues.  Whilst an output may still be within tolerances from a quality point of view, the variances in those tolerances, or a skew in those tolerances, can be an early sign of wear and so it is important that you take actual measurements rather than just record pass and fails. E.g. You calibrate a machine to cut lengths of 1000mm and your accepted tolerance is +/- 1mm, a continuous run of 100 items that measure 1000.50 +/- .25 mm are all in tolerance but are all a bit on the long side and can point to an underlying issue.

Systems

Another important aspect to consider is your IT systems.  If you are looking at replacing your ERP solution, then look at replacing it with a system that is embracing AI. As an SME, spending millions on an AI project for predictive maintenance is not usually an option, replacing your ERP solution with a recognised ERP solution such as Microsoft Dynamics® Business Central to drive your business forward, store all the required data and make use of Microsoft’s investment in AI and their Copilot offerings often is for a the fraction of the investment.  Microsoft are looking at adding AI across their offerings, and utilising a solution, such as Business Central, to store all that valuable production data is a good first step in the process.

Wherever you are on your journey, make sure that your strategic decisions take into account preventative maintenance requirements and that you put some thought into them.  Even if the decision is ‘not now’, make sure you at least do the groundwork to make it possible in the future.