Maximizing Efficiency and Sustainability through Predictive Maintenance and AI

The manufacturing industry is one of the world’s highest producers of carbon emissions. The industry is responsible for approximately 20 percent of the world’s carbon emissions[1]. Energy production or consumption during the manufacturing process are key contributors. But new paradigms can be used for designing, making, and disposing of assets. These paradigms promise to reduce waste by eliminating it from the products and processes or by reusing, recycling, or recovering materials. With the help of data and technology, these goals are much more achievable now than they were a decade ago. Artificial intelligence (AI) has captured headlines for the past few years and has promised to change how we work, live, and interact with our environment. One of the greatest opportunities of AI in business is in manufacturing. SAP Asset Performance Management, enabled by AI, directly supports organizations to transition towards circular manufacturing. In this post, we will learn how.

The interplay of circular economy and circular manufacturing

Circular manufacturing is part of a broader principle called circular economy. Circular economies aim to reduce the amount of waste (e.g., materials or energy) of a product and design them for longevity, ease of material recovery, and recycle. Circular economies are a more sustainable approach to production and consumption of goods and assets compared to the traditional ways of manufacturing (aka linear economies) that do not consider reusability into account. Circular economies typically have four objectives: Reduce, Refurbish/ Reuse, Recycle, and Recover[2]. Circular manufacturing aims to reduce waste created because of machines and to reuse old components when making new products. It also promises to achieve United Nations’ Sustainable Development Goals[3] (SDG) — energy (#6), sustainable consumption and production (#12), and climate change (#13). One way to operationalize circular manufacturing is by establishing new maintenance paradigms, such as condition-based or predictive maintenance.

But what exactly makes these approaches different and much more sustainable?

The circular economy vs. a linear economy (Source: SAP Insights)

The path from preventive to predictive maintenance

Businesses have traditionally maintained their assets and machinery on a fixed schedule. This approach is called preventive maintenance as its objective is to prevent failures. Maintenance is scheduled every six months, for example. The downside of this approach is that a fixed schedule does not take into account the actual asset utilization. Hence, the maintenance interval might be too long if the asset is used more frequently or short, if the asset is underutilized. In addition to potential downtime, an additional aspect makes preventative maintenance a less popular choice if the chosen maintenance window is too short; it is the impact on the environment and sustainability.

Maintenance approaches have evolved with the availability of big data and sensors. A popular new approach is called condition-based maintenance or predictive maintenance. Condition-based maintenance uses thresholds or conditions in rules (e.g., when a vibration exceeds a defined threshold, maintenance is scheduled). The most significant difference compared to preventive maintenance is the use of data to determine the actual usage of the asset (e.g., when the sensor data deviates from its standard normal patterns or schedule maintenance) and SAP Asset Performance Management offers just that. Over time, systems can recognize patterns in the data about asset failures and alert maintenance supervisors to send technicians to the asset before the failure occurs. This maximizes the lifetime of a part or asset and reduces waste. This directly increases sustainability.

To set up a predictive maintenance program, the assets in scope must be equipped with sensors that can capture data. The most common types of sensors monitor vibration, temperature, and pressure. To understand when an asset should be maintained, companies first need to define a baseline of values for the type of asset they want to monitor. Over time, data from asset sensors provide insights into when an asset will likely need to be maintained. This would, for example, be the case when data from sensor readings show deviation. For example, when there is more than the usual vibration or higher than expected temperatures. However, spotting rare but critical events in vast amounts of data is something a person can no longer do by themselves. This is where artificial intelligence in SAP Asset Performance Management comes in. The AI-based predictive maintenance applies AI (Artificial Intelligence) to determine asset's remaining life, predictive failures, detects anomalies, and prescribes recommendations.

AI powers predictive maintenance

Artificial intelligence is an emerging technology that learns from patterns in vast amounts of data. The patterns that AI observes are codified in a so-called model without a human explicitly programming it. New data points are then compared against this model to forecast future trends or determine how similar this new data point is to previously observed ones. In the case of condition-based maintenance, the most common source of data are sensors that are installed on or close to the asset that you want to monitor and maintain. As new sensor data comes in, the AI model analyzes it against the historical data from the asset, such as vibration, temperature, or pressure. The data is then used to predict when the part or asset will most likely need to be maintained before it breaks. This narrows the maintenance window, maximizes asset utilization, and reduces unnecessary waste from premature replacement of parts. This is especially impactful in asset-intensive industries. SAP Asset Performance Management directly integrates into SAP S/4HANA Asset Management, allowing a seamless user experience.

Conclusion

AI in predictive maintenance helps to predict upcoming asset failures before they are likely to occur, based on sensor data from the asset that is being monitored and models that have been built on historic data. This enables predictive maintenance and contributes to the concept of circular manufacturing. Because of a more optimal maintenance schedule, components and replacement parts can be utilized longer and waste can be kept out of the economy, creating a more sustainable future and contributing to the United Nations’ sustainable development goals.