Periodic inspection, continuous condition monitoring or APM 4.0
Selecting the right asset reliability strategy is a question that keeps many of our customers awake at night. Currently, on social media, at conferences, in email marketing, a lot of attention is on the newest technologies like machine learning, predictive analytics and anomaly detection. It seems like APM4.0 or Predictive maintenance is the only valid strategy. But is that true? Is there a one-size-fits-all approach to asset reliability strategy? Although this might be the best approach in some cases, in many cases, this strategy's business case isn't proven and achievable.
Is a proactive strategy like condition-based maintenance feasible and effective? Alternatively, the asset may run to failure or time-based preventive maintenance is applicable. Is there a silver bullet? Is one strategy better than the other? In this article, I will try to answer these questions.
Improving asset performance is not something to be solved by the maintenance department only. Operations is just as much involved! Studies show that operations has just as much impact as maintenance when it comes to asset performance. Therefore eliminating organizational silo's and enhance cooperation between operations, maintenance and engineering towards the same objectives is crucial. In my opinion, implementing an Asset Management framework with a clear Policy and Strategy is fundamental for the desired business outcome.
Before we start looking at condition monitoring techniques, let's first look at the problem we're trying to solve by looking at the main types of failure patterns (see figure below).
First, we have a situation that failure happens mostly at the start of operation after installation and commissioning or after a maintenance intervention. This is often referred to as infant mortality. Think here of misalignment, bolts not torqued right, unclear maintenance instructions, or non performed quality checks.
Then we have the other extreme failure pattern: late-life or basic wear and tear causing failures.
And between these phases, we have what is referred to as a "constant random failure" pattern. Many things can cause this, including operating outside the operating window, process inputs, changing quality of raw material or changing weather conditions… and sometimes you simply don't know why equipment fails even when you have a preventive maintenance plan in place. Failure just "happens".
Many of us assume that most failures are age-related. Based on what studies by Nowlan & Heap and John Moubray (RCM2 Practioner) prove, we can postulate that 11% of failure is age-related, while 89% of failure is not age-related.
So in many cases, time-based preventive maintenance is ineffective and very costly. We continue to increase the frequency of an intervention, but the assets continue to fail. Random failure patterns are the ideal candidates for condition-based maintenance as it increases the accuracy of the probability of failure.
In short, condition-based maintenance helps to predict an upcoming failure that can avoid unnecessary plant down-time and helps eliminate unnecessary maintenance tasks. Improved scheduling of maintenance interventions will require fewer people at the site, which is a welcoming additional benefit amidst COVID-19.
Condition monitoring techniques
What are the different techniques that condition-based maintenance can provide? The main objective is to capture a potential failure before it actually occurs – before point F (figure 2). So the key question becomes: "how can we move left to increase the lead time to properly schedule a maintenance intervention before point F or even before point P?".
Whenever it is, a longer lead time enhances better planning and ensures to have the spare parts in time. It also enables us to assign the right people to the job who will be better prepared and can work more safely. So let's take a closer look at the different condition-based maintenance techniques we have: periodic inspections, online condition monitoring and APM 4.0 or Predictive maintenance.
Let's first look at periodic inspection. Periodic inspection has many manifestations:
- Periodic visual inspection based on human senses
- Periodic infra-red camera inspection
- Periodic oil analysis
- Periodic wall thickness measurements
- Periodic vibration measurements
- And several more techniques where someone approaches a piece of equipment and performs a measurement.
Of course, the key element is making sure that the inspection interval is shorter than the PF interval; otherwise, you might miss spotting the degradation and, as a result, miss "F", which could potentially result in unplanned downtime.
The advantages of this approach are quite straight forward:
- Relatively low cost to perform
- A high degree of flexibility
- High precision
The downside of periodic inspections is that the process of deterioration is already quite far advanced. Also, this technique relies heavily on the observer's experience and state of mind. It has a relatively short lead-time to plan an intervention, and it is cost-effective if experienced people are already on-site.
Online condition monitoring
Online condition monitoring is designed to detect potential failure effects, such as changes in vibration, temperature changes, leaks, and energy consumption changes. It uses technological observation methods to collect data regarding the condition and health of individual assets in real-time. When the measured parameters reach an unacceptable level, a maintenance intervention can be performed when needed. Your asset tells you it needs maintenance.
Usually, condition monitoring devices monitor only one condition, so one type of functional failure. When getting started with online condition monitoring, it is important to determine the appropriate device or sensor to be installed, how the data will be transmitted and stored, determine the thresholds or bandwidth of the parameters and monitor the equipment.
With the increasing volume of data from different data sources, the necessity to more accurately predict failures and the requirement for a longer lead-time to schedule maintenance interventions, predictive maintenance is the ideal condition-based maintenance technique. New technologies like anomaly detection, machine learning using data from different data sources like dark data (text, images, sensors, devices) and artificial intelligence are essential enablers for predictive maintenance, also known as APM4.0.
The best condition-based maintenance technique
I will illustrate how to choose the best condition-based maintenance technique for your situation by outlining the approach we use at Stork AMT. After criticality analysis based on the corporate risk matrix, we determine which equipment is highly critical and is subject to perform a Failure Mode Effect Analysis (FMEA). This FMEA is our foundation to start with. To define the most suitable condition-based technique for a customer, we have developed a questionnaire that we complete together with our customers. The company objectives are our guiding principle. Then we take into consideration the impact on the asset, the availability and quality of the data, and the complexity of the physical and technical situation.
To illustrate the selection process for the best suitable monitoring technique, an example: the chemical industry seems an ideal candidate for APM4.0 due to the high level of automation, instrumentation and stable operating conditions. In reality, we have noticed that the potential of APM4.0 is not always proven and realizable for several reasons:
- In many cases, there is too little data as unplanned downtime is typically concentrated in a small number of large events; thus, not enough failure data to train the predictive model;
- There is too little lead-time as the production process can't be interrupted immediately;
- Some maintenance interventions demand long lead-times to prepare and order spares;
- There is too little impact due to redundancy;
- Savings are small as, in some cases, optimizing planned shutdowns delivers more value than optimizing unplanned downtime.
Instead, improving condition monitoring through improved remote sensing can cut the mean-time-to-repair significantly reducing the impact of equipment failures. This solution provides more benefits in optimizing the intervention and thus mitigating the impact of failure.
In this article, I tried to show that developing a condition-based maintenance strategy is not a one size fits all approach. The use of internal knowledge, technology and thorough business cases are key enablers to make the best choice. In summary, all asset reliability strategies have their merits. There isn't a single silver bullet. Which strategy to select depends on your specific business, your goals and your typical situation.