Predictive maintenance promises many benefits, from reducing machine downtime and eliminating unnecessary maintenance to adding revenue streams for equipment vendors with aftermarket services. These benefits are achievable if engineering and business challenges are kept from getting in the way.
This article discusses three common obstacles that prevent businesses from successfully implementing predictive maintenance and how to overcome them.
1. We do not have enough data to create a predictive maintenance system.
Many predictive maintenance approaches rely on AI algorithms, so there must be enough data to train an accurate model. For predictive maintenance, this data usually originates from sensors on machinery, but many organisations struggle with limited or fragmented data.
Take a close look at your data sources
You might find that your department does not collect enough data for a predictive maintenance system. Consider whether other departments collect data as well. Perhaps the Controls division does not collect enough data, but what if it was combined with data from the Services division? Looking farther afield within your organisation could be enough to meet your needs.
Also, consider your agreements with suppliers or customers. Cooperating to prolong the health and efficiency of equipment components may create a win-win situation that fosters data access between business entities.
Change how data is captured
Some systems operate in a “feast or famine” mode, collecting little to no data until a fault occurs. Others only log event codes and time stamps: engineers are notified that an event occurred, but not the sensor values at the time of the failure. Although this data may be useful for diagnostics, it is likely insufficient for developing models that can predict failures.
Consider changing the data logging options to record more data, perhaps on a test fleet, if production data is not available. Depending on the load on existing embedded devices, reconfiguring them to collect and transmit sensor data may be possible, or external data loggers may be necessary to get started.
Use simulation tools to synthesise data
Generate test data using simulation tools and combine that data with operational sensor data to build and validate predictive maintenance algorithms. This is done by creating models for monitoring the mechanical, electrical, or other physical system. Validate synthesised data against measured data to ensure the model is well-calibrated.
Modelling three types of faults: cylinder leaks, blocked inlet, and increased bearing friction.
Considerations
When considering data for a predictive maintenance system, begin to analyse data early to understand which features are important and which may be redundant. By taking a proactive approach to data and leveraging tools like MATLAB and Simulink, organisations can overcome data-related obstacles and lay the foundation for successful predictive maintenance projects.
2. We lack the failure data needed for accurate results.
Historical fault, failure, and degradation data are crucial for designing accurate predictive maintenance algorithms. Failure data may not exist if maintenance is performed so often that no failures have occurred, making collecting this data a significant hurdle.
Simulate failure data
With the right tools, an engineer with domain expertise can generate sample failure data. Using a simulation product like Simulink®, an engineer can build or use a physical model of the machine, as described in the first challenge.
Products like Predictive Maintenance Toolbox™ simplify tasks like failure data generation and provide data ensembles for managing and organising multiple datasets.
Understand the data available
Failure data might not be available, but operational data might show trends about how a machine degrades over time. Looking at the raw sensor data from a component, system, or machine with dozens or hundreds of sensors can be intimidating. Statistical techniques such as principal component analysis (PCA) can help reduce the dimensionality of such datasets and provide valuable insight into how equipment operates over time.
Sometimes, only healthy operational data may be available. In these situations, one-class anomaly detection techniques, like autoencoders or one-class SVMs, can be used to characterise abnormal data. This helps identify data that might represent degradation or failure.
Using principal component analysis to visualise how equipment trends prior to failure.
Considerations
Try to keep the number of variables down to the minimum needed for an accurate model. It can be tempting to include every measured component to make sure nothing is missed, but this may result in a black-box model that is overwrought with complexity.
3. We don’t know how to predict failure
Understanding the cause of a failure is important, but there is a difference between identifying what went wrong and knowing how to predict it. Many organisations struggle with developing and deploying effective predictive maintenance algorithms due to the complexity of the prediction process and the lack of expertise in data science and machine learning.
Define goals and start small
To compare a predictive maintenance algorithm to the old method, you must first set clear goals. Begin with pilot projects focused on well-understood systems or failure modes. When you and your team have the domain knowledge to explain failures, you can identify the features that influence the system’s performance and design an algorithm. When you and your team master the algorithms for a simple problem, you can transfer that knowledge to more complex systems.
Work with the tools your engineers already know
Instead of introducing new technology and techniques, take advantage of new capabilities in existing software. Some tools that engineers already use, such as MATLAB, have specific predictive maintenance capabilities, enabling engineers to continue working in an environment they know. These tools provide functions, reference examples, training, and technical support to get started quickly.
Validate and iterate
When algorithms begin to show promising results, use current and historical data to test and validate models before moving to production. Adopt an iterative approach to algorithm development and validation, continuously refining models based on feedback from operational data and performance evaluations. By iteratively improving models, organisations can ensure that their predictive maintenance solutions are accurate and reliable.
Considerations
It is important not to try to do everything simultaneously because the project can feel too complex and cause frustration. Define clear goals, start small, validate against data, and iterate until you are confident with the results. Repeat this process to build up to more complex systems.
Conclusion
Predictive maintenance is achievable with the right tools, guidance, and motivation. With tools like MATLAB and Simulink and a systematic approach to predictive maintenance development, organisations can overcome the early obstacles to success. Find the approach that works for your business and iterate until you get it right—and remember, you do not have to do it alone.