Pulling down the barriers to predictive maintenance. By Jos Martin
Predictive maintenance promises a lot, from minimising machine downtime to adding revenue streams for equipment vendors with aftermarket services. These benefits are eminently achievable but only if you can keep engineering and business challenges from being the barrier to progress.
So, what are the most common obstacles that stop businesses from successfully implementing predictive maintenance? And how is it possible to solve each of these challenges?
1. We don’t have enough data to create a predictive maintenance system
Many predictive maintenance approaches rely on machine learning algorithms, so there must be enough data to create an accurate model. This data usually originates from sensors on machinery, but if the sensors are new or the way readings are logged limits the information, you will need to think about the best way to access enough data to build your models.
The solution is fourfold:
- Look to other departments in case they collect data that can be amalgamated with other data sets.
- Contemplate changing the data logging options to record more data, perhaps on a test fleet if production data is not a vailable.
- Generate test data using simulation tools and combine that data with what sensor data is a vailable to build and validate predictive maintenance algorithms.
- Analyse data early to understand which features are important and which may be redundant so that unusable data can be deleted.
2. We lack the failure data needed for accurate results
Failure data is a crucial part of teaching algorithms to recognise the warning signs to trigger just-in-time maintenance. However, failure data may not exist if maintenance is performed so often that no failures have occurred, or the system is safety critical and cannot be left to fail. To stop this, it is possible to simulate failure data and learn how to recognise warning signs from operations data.
An engineer with detailed system knowledge of how the physical components work will be able to generate sample failure data with the right tools. Using a simulation product, an engineer can then build or use a physical model of the machine. Tools such as failure mode effects analysis (FMEA) provide useful starting points for determining which failures to simulate. The resulting failure data from those simulations is then labelled and stored for further analysis.
When failure data isn’t present, operations data might show trends about how a machine degrades over time. But looking at the raw sensor data from a system, or machine with 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.
In addition, it is also possible to generate data from existing models which can then be used to improve the amount of failure data.
3. We understand failures, but cannot predict them
Understanding the cause of a failure is important for your business, but there is a significant difference between identifying what went wrong and knowing how to predict it. Root cause analysis is an integral part of domain knowledge. That, paired with predictive maintenance algorithms, creates an effective predictive maintenance program. If the algorithm part of the equation is a new and intimidating undertaking, you can take steps to reduce the learning curve.
Firstly, it is important to define upfront what your goals are (e.g. earlier identification of failures, longer cycles, decreased downtime). You should then think about how the predictive maintenance algorithm will affect these goals. Building a framework that can test an algorithm and estimate its performance relative to your goals will enable faster design iterations. It will no longer be up for debate whether a new algorithm is better than the previous state, but rather it will be clear if a new algorithm is better based on the agreed-upon goals.
But, if the understanding of what is causing the failures is there, then the right amount of domain knowledge is present. Make sure you understand the features and factors that affect the performance of the system and build a predictive maintenance algorithm. Once you and your team are comfortable building the algorithms for a simple problem, you can apply that knowledge to more complex systems.
When predictive maintenance algorithms begin to show promising results, use current and historical data to test and validate models before moving to production. Use the domain knowledge within your team to tune models to predict different outcomes based on the cost/severity of those outcomes and to further validate models, add generated failure data similar to known historical conditions and test the system.
4. New predictive maintenance models come with financial risk
Every new technology requires investment that must be justified, and the time required to realise ROI should be as short as possible. But this is difficult when there are uncertainties about how quickly you and your team can learn to use these new tools. If machine learning has only recently been introduced, it is only natural to see what might be considered an advanced application of it as a risk. However, there are steps to take that can minimise that risk and get up and running with a working predictive maintenance model as quickly as possible:
- Work with tools that your engineers already know and work with.
- Access sensor data gathered from multiple sources, such as databases, spreadsheets, or web archives.
- Pre-process data by adjusting noise filtering or outlier settings or comparing the effect of different filtering on overall algorithm performance.
- Instead of feeding sensor data directly into machine learning models, extract features from the sensor data. These features capture higher-level information in the sensor data.
- Train the model by classifying the data at the outset and creating a comprehensive list of failure scenarios to predict, choose classification methods, and simulate models.
- Generate code and deploy models as an application on hardware.
Predictive maintenance is an achievable goal with the right tools, guidance, and motivation. Find the features, models, and methods that work for your business and iterate until you get it right – and remember you do not have to do it alone.
Jos Martin is Senior Engineering Manager at MathWorks, the leading developer of mathematical computing software. MATLAB, the language of technical computing, is a programming environment for algorithm development, data analysis, visualisation, and numeric computation. Simulink is a graphical environment for simulation and Model-Based Design for multidomain dynamic and embedded systems. Engineers and scientists worldwide rely on these product families to accelerate the pace of discovery, innovation, and development in automotive, aerospace, electronics, financial services, biotech-pharmaceutical, and other industries.