Why Predictive Maintenance Isn’t Always the Right Solution

The world of maintenance strategies has evolved significantly over the years, with Condition-Based Maintenance (CBM), and Predictive Maintenance (PdM) emerging as prominent approaches. While Predictive Maintenance has gained attention due to its reliance on Artificial Intelligence (AI) and Machine Learning (ML), it’s important to recognize that it might not always be the ideal solution for every situation. In this article, we explore the nuances of Predictive Maintenance and shed light on why it might not always be the preferred strategy.

Differentiating CBM and Predictive Maintenance

Condition-Based Maintenance and Predictive Maintenance are often used interchangeably, but there’s a fundamental difference between the two. CBM involves setting measurement thresholds based on human experience, which operators adjust over time to improve accuracy. On the other hand, PdM involves initially defining breakdown scenarios and thresholds by human operators and then harnessing AI algorithms and ML to detect patterns in data streams to predict potential failures.

The P-F Curve and P-F Interval

As we discussed in our previous blog on CBM, the P-F curve, is a graphical representation showing the points of Potential Failure (P) and Functional Failure (F) of an asset. This highlights the significance of early detection of potential failures. It’s crucial that the P-F interval should be long enough to plan and conduct maintenance effectively. The P-F interval for Predictive Maintenance is even longer than that for CBM. This raises the question of why Predictive Maintenance isn’t always the preferred choice, especially since both strategies leverage sensors, data streams, and software.

Balancing Predictive Maintenance Considerations

Predictive Maintenance can be a suitable solution for critical assets where downtime and repair costs are substantial. Its effectiveness lies in the ability to plan maintenance over a longer P-F interval, allowing for proactive measures. Additionally, PdM shines in situations where changes in equipment condition are hard to detect through other means.

Nonetheless, it’s important to acknowledge that Predictive Maintenance isn’t a one-size-fits-all solution. Here are 5 reasons why PdM might not be the right solution:

1. Costs of storing and processing data: Predictive Maintenance’s Achilles’ heel is its dependence on processing and storing vast amounts of data, which comes at a cost. We’re aware of a case within industry where the data costs to monitor frequency changes in electric motors were so high that it proved cheaper to simply have a spare motor on standby.
2. Data Acquisition Hurdles: Implementing Predictive Maintenance can pose data acquisition challenges, particularly for smaller operations or industries with constrained data collection capabilities. Gathering the essential information for precise predictions might require specialized sensors, which could necessitate a significant initial investment. Consequently, this factor can significantly influence the feasibility of achieving a satisfactory return on investment.
3. The P-F interval still isn’t long enough: Despite being longer than when using CBM, the P-F interval may not necessarily be long enough. For it to be useful, the P-F interval needs to be long enough to allow a response which prevents or drastically reduces downtime and repair costs. If the P-F interval isn’t long enough to allow this, then Predictive Maintenance isn’t worth adopting for that specific case.
4. Changing Environments: Production lines and machinery often get changed over time, and this can render previously identified failure patterns obsolete.
5. Expertise Requirements: Just like other methods, implementing Predictive Maintenance requires an understanding of machinery and processes to identify suitable breakdown scenarios and know what to measure and analyse. This requires an initial commitment of personnel resources during implementation. The time saving advantages of AI and ML will only be felt in the long run.

Realizing the Potential: A Step-by-Step Approach

At Innius we advocate for a step-by-step approach to asset management. This should start by identifying reasons for downtime and prioritizing those scenarios with the worst impact. The innius Insight app, includes a Downtime Report which does this using a Pareto analysis. A balanced assessment of the data availability, costs, and operational needs can then be conducted to determine the best strategy for each scenario. If a digital approach is appropriate, it will usually make sense to start with CBM. Again, the existing innius apps can facilitate this with sensor thresholds, push notifications and assigned tasks. Assuming the P-F interval is long enough, this approach yields quicker results and faster ROI than Predictive Maintenance. Then as operations mature and data accumulates, integrating Predictive Maintenance powered by AI and ML can become a natural progression; particularly if an even longer P-F interval is desirable.

Conclusion: Striking the Maintenance Balance

In the ever-evolving landscape of maintenance strategies, the choice between Predictive Maintenance and CBM requires careful consideration. While Predictive Maintenance dazzles with its AI and ML capabilities and the promise of prolonged P-F intervals, it’s not always the golden solution. Balancing the advantages of early detection and proactive planning with the challenges of data costs, acquisition hurdles, changing environments, and expertise requirements is crucial. Rather than succumbing to the allure of the latest technologies, a pragmatic step-by-step approach emerges as a wise course of action. By understanding the unique needs of each scenario, embracing CBM for quick gains, and transitioning to PdM when warranted, organizations can ensure efficient operations and maximize their return on investment while keeping the machines running and the downtime minimal. At Innius, we champion this balanced approach, realizing that each maintenance strategy has its place in the grand orchestra of asset management.