What is Predictive Maintenance and what are its benefits?
In the field of maintenance, transformation plans encompassing Operational Excellence powered by digital and human dynamics can deliver substantial tangible benefits.
The growing availability of both operational data and digital tools unlocks important opportunities for operational improvements. It will help you build your industrial future with data being the cornerstone of Industry 4.0 and the latest technological innovations. One key application is Predictive Maintenance. Especially in industries such as chemicals or aerospace where operations are more complex and specialized, Predictive Maintenance can bring tremendous business value.
In this article we will focus on defining what is Predictive Maintenance and exploring its benefits. We will then share with you the main success factors when embarking on a Predictive Maintenance improvement journey.
What is Predictive Maintenance?
The goal of Predictive Maintenance approach is to predict, as good as possible, the optimal moment when maintenance should be planned and executed on your equipment. In contrast to the more classic notion of Preventive Maintenance, where maintenance checks are planned routinely, Predictive Maintenance allows you to perform maintenance when it is necessary. Thus, it saves time, energy and money.
One important remark to make before we continue: Predictive Maintenance is an approach with clear benefits (on which we will zoom in further in this article), but this approach should always be seen within the bigger picture of the overall maintenance journey. In this regard, certain base conditions need to be met before Predictive Maintenance starts to make sense. This can include a proper performance management structure, clear priorization processes, … For more on this, please refer to the more general topic on maintenance excellence.
An old trade with a fast-growing field of applications
You might be tempted to think that Predictive Maintenance is yet another new approach coming to the surface following the recent boom in importance of Big Data and Data Analytics.
However, Predictive Maintenance techniques have been around since the 90s. Originally, these techniques were very time-consuming and expensive investments. Therefore, they were only used in highly capital-intensive environments and situations, such as the maintenance of jet engines in airplanes.
A couple of factors are now leading to the fact that these techniques become more widely accessible and more easily applicable. The cost of sensors approximately decreased by half in the last 10 years. Next, the decrease in bandwidth costs has led to increased connectivity, as well as increased data storage capacity. Not surprisingly, these are the same factors which are also supporting the general trend towards Big Data and Data Analytics. In short, applying the right technologies and digital expertise to the more classic Predictive Maintenance approach, allows us to unlock further operational improvements.
What are the benefits of Predictive Maintenance?
We briefly mentioned above that Predictive Maintenance allows everyone to save time, energy and money. Let us dive a level deeper into what these benefits can specifically entail for whom. It is important to realize that the benefits of Predictive Maintenance are very broad, apply to different impacted groups and should be considered holistically.
First, from a production standpoint, the main positive impact of Predictive Maintenance is on the unplanned downtime. There will be fewer unplanned breakdowns and interventions, directly leading to a higher Overall Equipment Effectiveness (OEE). In short, Predictive Maintenance boosts your productivity.
Second, we can assess the viewpoint of the actual maintenance teams. From their perspective, Predictive Maintenance allows these teams to schedule their activities more efficiently and more effectively. This reduces wrench time (i.e., actual time performing the maintenance work).
Most importantly, since maintenance activities can be planned more effectively, the number of spare parts needed to perform unforeseen interventions is reduced. In this way, you avoid replacing (expensive) parts that are still in good condition. Next, this also reduces the necessary inventory of spare parts to be held. Finally, Predictive Maintenance will ensure the right interventions are done at the right moment, directly ensuring a longer equipment lifespan. All of this translates to the same general output: Predictive Maintenance leads to less capital employed and less expenditures.
In conclusion, looking at the possible benefits from a holistic perspective, Predictive Maintenance allows you to boost your productivity and reduce both the time performing maintenance activities as your maintenance expenditures.
Different types of maintenance strategies for different situations
Up to now, we have been discussing Predictive Maintenance as one general approach or technique. Its goal being trying to predict the optimal time to perform maintenance on your equipment. In practice, however, we can establish four maintenance strategies. The difference between these maintenance strategies is based on the complexity and criticality of your equipment:
- Prescriptive Maintenance: Let us start with the situation where the equipment is complex and the impact of a possible failure is high. In this case, it might be interesting to choose an approach called Prescriptive Maintenance. This technique goes further than Predictive Maintenance. Not only are failure modes predicted, but possible counteractions are proposed based on the data and the model.
- Predictive Maintenance: Predictive Maintenance will allow us to predict when equipment will (most likely) be in bad condition and schedule our maintenance tasks accordingly.
- Anomaly Detection: For equipment which is slightly less complex or critical, there is Anomaly Detection. Using this approach, the purpose is no longer to predict when a failure will occur. Rather, using pattern recognition, a continuous monitoring of conditions allows for rapid interventions in the event of unnormal patterns or equipment behavior. In contrast to the above-mentioned Predictive Maintenance techniques, this is a more reactive approach.
- Condition Monitoring: A technique which might be applicable to the least impactful equipment is Condition Monitoring. As the name suggest, the condition of the equipment is monitored by following up specific criteria within limit values.
How to approach a successful Predictive Maintenance Journey?
So far, we have established an understanding of Predictive Maintenance and why the digital side of the equation is crucial, its evolution over the past decades and its possible benefits. Before discussing an actual Client case study, it is worth to briefly explain the typical steps to take on a Predictive Maintenance journey. A typical Predictive Maintenance journey includes 3 steps: ‘Explore’, ‘Prototype’ and ‘Industrialize’.
1) Strategic Exploration: What are your options?
A possible challenge many organizations face when considering Predictive Maintenance is to clarify the scope, the expected benefits and the starting point of the journey.
A ‘Proof of Value’ approach can fix this. The first step is to explore open-formatting in a number of areas and to build your own company insights into this. This will also allow you to establish a target business case very early on.
An important step to take is to classify your own equipment. Some of your equipment may be less critical, less complex, brand new or already in use since a while. Focus is a good first approach, since completely working out a full scope may not be worth the time nor money. In the end it should be your objectives and your equipment that determine your Predictive Maintenance strategy and the related business case.
2) Prototype: Make or buy?
Building further on your previous insights, the next step is to create a working prototype. This model will need to be validated against reality, which in turn will allow you to truly test the business case as well. Note that when you get at this stage, there will be a vast array of possible vendors having some sort of software or service to offer.
Whether to develop your own Predictive Maintenance program or to buy an off-the-shelf solution is a decision that should be made after having built some experience during the prototype phase as, during this phase, you will also sharpen your knowledge on the functionalities you need and the data that are required to run the models.
3) Industrialize: How do you scale-up your Predictive Maintenance?
The most integral success factor which will determine the further scaling of your Predictive Maintenance has already been done in the previous steps. We recommend being very focused on a single site and even a single equipment, namely the one with the most potential for savings, to start your journey. Ideally, the savings you will be able to generate can serve as funding for the next equipment(s) and the next site(s).
Two very crucial points need to be made in this regard as closing points:
- First, it does not make sense to keep optimizing your model. At a certain point, the marginal returns of a more optimized model simply do not outweigh the extra development costs anymore. The Predictive Maintenance journey for a single equipment, therefore, is completed.
- Second (to balance out the first point), as an organization as a whole, your Predictive Maintenance journey is not finite. This is not a project you start and close, but rather it should be an ongoing program. Your equipment will still need maintenance year after year. Hence, you will need to keep refining your Predictive Maintenance.
Predictive maintenance applied at a chemicals company
Unplanned breakdowns and inefficient use of maintenance resources led to significant (maintenance) costs. The objective was to proof the value of applying predictive maintenance technology, by developing a system, together with the Client, that predicts failures for a certain equipment group.
Our approach started by developing a clear business case. It is very important to align with the stakeholders about the possible outcomes of predictive maintenance. Given the Client’s situation, the decision was to focus more on reducing maintenance costs than on improving OEE per se. This was crucial in managing expectations.
To start off, the symptoms and failure modes of the equipments were identified, together with the Client. It was possible to rely on historical events, FMEAs, equipment drawings and the actual experience of operators.
Then, through several workshops, ‘variables to be measured’ were identified and the availability of necessary data on these variables was assessed. A gap was identified that needed to be filled by mostly sensor data, initiated by IoT sensor vendor selection. It is worth mentioning that it is common for (old) equipment to lack the data that is required for Predictive Maintenance.
Once these steps have been taken, historical (event) data was assessed and used to build predictive models. This work is iterative in nature, as it requires adding data types and validation of its workings.
When the model was successful in predicting, the last steps were to automate the processing of sensor data, develop the right visualizations and to take it into production.
Together with our Client, we delivered the following concrete benefits:
- A data architecture for automated processing of sensor data and generating predictions on equipment failures was established
- Dashboards were set up for exploration of sensor data, output of the prediction models and machine learning model performance
- An estimated €200K p.a. hard maintenance cost savings and multiple qualitative benefits (such as sustainability, spare parts, OEE) were achieved
- We delivered a business case template and implementation process for the development and rollout of predictive maintenance to other sites and improved internal capabilities to run future PdM projects.
Key learnings: how to achieve success in your own Predictive Maintenance project
Now that we have explored an overall view on predictive maintenance, together with a general approach and an actual Client case, let’s dive into the critical factors that determine success. After all, if predictive maintenance were just about following a route with simple steps, every company would be doing it successfully.
First success factor: Balance model complexity & business case
Moderation is key for Predictive Maintenance. As can be seen in the graph below, an optimum can be found. In this visual this optimum is set somewhere in the middle of the x-axis of ‘model complexity’, that point where the difference between operating costs and avoided costs is the highest. The goal is to reap the benefits of the avoided costs while maintaining the operating costs, that act as a lever, to a minimum.
So, what can companies do to achieve that optimum?
A first best practice is to use acceptance criteria for model acceptance to balance true positives vs. false positives. Setting a clear criterium helps to avoid discussion and helps to define a common understanding of the success of a model, for all stakeholders involved.
Another best practice is linked to the art of scoping, as that helps to draw a similar looking graph as the one illustrated above, but then with real-life data, or accurate approximations thereof. Only the right equipment should be in scope for rolling out Predictive Maintenance, and this is determined by:
- Data availability and quality. A clear advantage can be gained whenever IoT sensors are installed months prior to implementation.
- Potential of KPI improvement. Stated differently: equipment families with benefit-cost-ratios > 1.
- Good instrumentation and scalability across the sites
Second success factor: Leverage domain knowledge – process driven & technology enabled
Domain knowledge is key for the quality of the Predictive Maintenance model. The following best practices are defined as to extract the knowledge in the optimal way.
- Project setup: projects have different phases. It is crucial to know what parties should be involved in which phase, e.g. excellence centers & IT departments should be given a larger role in an implementation phase than in the Proof of Value phase.
- Project team: make sure that all the right resources are involved, as described in the visual below. The required resources will, for some companies, also coincide with a new way of working especially whenever the roles described below are new in a company.
- Training: all involved stakeholders should be aware of the possibilities of Predictive Maintenance, the route to implementation and the set-up of the project.
Third success factor: Manage expectations
The last success factor may not be the most selling or appealing success factor. Perhaps, therefore, this factor is even more important to mention. The rewards of Predictive Maintenance can be very significant, but it simply requires patience. The time-consuming aspect arises from the following elements:
- The time-consuming and iterative nature of the required activities related to Predictive Maintenance, e.g. sensor placement, model building.
- Scoping is crucial and doing this right requires a lot of data cleaning to make the data available and usable.
- Lastly, reaping the benefits of reduced maintenance/diagnostic costs usually requires some sort of organizational or structural changes. Companies should have a plan in place as to capitalize on the obtained time.
Where should you start?
By now, we hope to have made it clear that Predictive Maintenance is not something you can expect to quickly set up within your organization.
Predictive Maintenance can be an integral part to your overall operational excellence program. It is through the deliberate application of digital technologies that you will be able to unlock even more operational improvements. Remain result-driven and focus on key equipment to generate impactful and tangible results in the short term. Finally, to guarantee sustainable results: take care of including several dimensions, from processes to technologies and human dynamics into a pragmatic Predictive Maintenance journey.
The author: Max Vermeeren, Senior Consultant, EFESO Consulting Belgium