Transformation of traditional maintenance to predictive maintenance
Maintenance is the process of ensuring that physical assets continue to perform their intended tasks. Maintenance decision-making entails evaluating and selecting the most efficient maintenance strategy, such as reactive, preventive, and predictive maintenance (which we covered in Section 1 of the white paper Predictive Maintenance (Smart Maintenance and Repair); Click here for more information). Adopting an inefficient maintenance policy will harm the organisation far beyond the direct costs of equipment failure. It is crucial that companies choose the best suitable maintenance strategy to save money and slash downtime. This blog discusses transforming traditional maintenance into Condition-Based Maintenance (CdM) and from CdM to Predictive Maintenance (PdM).
Traditional maintenance and the move towards Condition Monitoring Maintenance (CdM):
The traditional maintenance practice is based on the use of scheduled maintenance programs. They consist of regularly scheduled maintenance and reactive maintenance as needed. This implies that when components fail or wear out, the equipment is removed from the production cycle until it can be fixed. The concept of scheduled maintenance includes the view that every part of a complex machine has a “right age”. The right age is when a specific component must be replaced for a continuance of reliable and safe operations. Scheduled maintenance is interlinked with functional equipment failures, failure consequences, failure modes, and failure effects.
Traditional maintenance strategies are insufficient for satisfying modern industrial requirements. The concept of maintenance has evolved from a reactive attitude (maintenance intervention post-failure) to a predictive attitude (maintenance intervention to prevent the fault). Strategies and concepts such as Conditioned based Maintenance (CbM) have progressed to support this outcome.
The CbM is different from schedule-based maintenance. The difference is that the maintenance depends on the machine’s actual condition and not on any pre-set schedule. For example, a typical schedule-centric maintenance strategy demands automobile owners change the engine oil every 3,000 – 5000 miles. The actual condition of the vehicle or functional performance of the oil is ignored. Conversely, if the driver has some method of knowing the oil lubrication and actual state of the vehicle, then they can choose to postpone the oil change or extend the vehicle usage. The change of oil may also be preponed if the need arises.
Figure 1 illustrates the overall maintenance strategy. The supporting programs are included in the picture. The strategy is a blend of preventive and reactive maintenance programs. (Traditional maintenance approach).
Figure1: Flow Chart of Maintenance Strategy.
Condition Monitoring Maintenance (CdM):
This maintenance strategy involves sensor use. These sensors measure the status of an asset over time whilst the asset is in operation. Mechanical and operational conditions are monitored as part of this strategy, either in real-time (aka “real-time condition monitoring”) or periodically. Figure 2 depicts CdM, which can be performed using various approaches. Periodic monitoring is conducted at pre-set intervals, such as every alternate hour, with the help of portable indicators such as hand-held measurement instruments and acoustic emission units. In on-line (or real-time) monitoring, a machine is continuously monitored, and a warning alarm is triggered whenever an error is detected. Such monitoring is done via sensors and process data.
Figure2: Different approaches to Condition Monitoring
Each monitored variable is given an alarm threshold. The alarm is sent when a variable value jumps its threshold. The problem part is identified and duly scheduled for maintenance. Continuous CdM techniques can be used on various equipment, including compressors, pumps, spindles, and motors. These can also be applied to detect partial discharge on vacuum leaks or machines.
The assumption underlying schedule-based maintenance is that machine failures are directly linked to the operating machine age, which may not always be the case. Failures may not always be linear in nature. Studies indicate that 89% of the problems are random with no direct relation to the operating age. Table1 showcases some known failure patterns along with their conditional probability (Y-axis) with respect to Time (X-axis).
Table1: Failure Conditional Probability Curves
Complex items routinely demonstrate some infant mortality. For those parts that survive the initial period, the probability of failure either remains constant or increases gradually. Parts will wear out due to age. If this fact is considered, then there is a greater chance of a potential failure regardless of any schedule-based maintenance. It is highly probable that the system may fail immediately after a scheduled maintenance. Thus, preventive maintenance imposes additional repair costs. The CdM reduces such additional costs by scheduling maintenance that occurs if and only when a probable breakdown symptom is identified.
Figure3: Difference between Preventive and Condition Monitoring
Three major technology enhancements enable a Condition-Monitoring Maintenance Management (CBMM) solution: Remote Sensor Monitoring & Data Capturing, Real-time Stream Processing of Sensor Data, and Predictive Analytics over a traditional maintenance solution.Three major technology enhancements enable a Condition-Monitoring Maintenance Management (CBMM) solution: Remote Sensor Monitoring & Data Capturing, Real-time Stream Processing of Sensor Data, and Predictive Analytics over a traditional maintenance solution.
CdM is an economical approach, although sophisticated CdM approaches can be expensive. It is an excellent solution when known indicators provide a reliable warning of impending failure, and the cost of failure can moderately impact the organisation's finances. Thus, CdM is necessary for any industry that uses simple to complex machinery. So, why is it being replaced by PdM?
PdM predicts failures. To do this, it tries to learn from machine performance. PdM uses data collated through CdM and then applies analysis or AI/Machine Learning to discover patterns to predict failures before they occur. This technique is a suitable solution for costly potential failures and critical assets. The main difference between PdM and CdM is the timing. CdM concentrates on real-time conditions, while PdM focuses on detecting defects in their nascent stage (60 or 90 days in advance).
How to move towards Predictive Maintenance (PdM) from Condition Monitoring Maintenance (CdM):
You must evaluate your organisation's assets' unique attributes before deciding to move from Condition Monitoring Maintenance (CdM) to Predictive Maintenance (PdM). Consider all actual and potential trade-offs of the cost of prevention (detection of possible failure) and the cost of repair/failure before settling on the best approach. Advanced techniques such as CdM and predictive analytics are best when the cost of repair or failure is high. Figure 4 shows how to select a maintenance strategy based on machine asset type.
Figure4: Selecting a maintenance strategy based on machine asset type
CdM was first adopted in the 1980s and improved gradually with different CdM versions making their way across industries until the process transformed into PdM, as explained in the following sections.
Condition Monitoring 1.0:
In an industrial milieu, Legacy CdM (the1980s and early) includes lagging indicators like low lube oil pressure, high temperature, and irregular pump discharge pressure. An alert condition on such measurements implies an ongoing failure, or the failure has already happened. Thus, an alert demands a timely response. The indicator will not give you sufficient time to plan. Production lacks planning time, and maintenance lacks adequate time to assemble the right parts, skills, and tools.
Condition Monitoring 2.0:
The CdM 2.0 (the 1990s to 2000s) noticeably improved defect identification. The deployment of variable speed drives to boost electrical energy consumption efficiencies increased power, motor current, and speed. Such improvement is possible due to cost reductions, enhancements in IO systems infrastructure, better reliability, and easy magnetic mounting of sensors. The CdM 2.0 spans motor current, bearing temperature, speed, power, and Overall Vibration. A variation in any of the measurements mentioned above indicates an impending failure of the concerned pump or pumping system.
You can use these measurements to diagnose problems successfully. However, challenges appear when you try to set the alert thresholds for automated alerting. Nuisance alarms occur with annoying regularity due to the varying nature of the product recipe, process, or season. The clarity and simplicity of the approach are thus compromised, necessitating in-house or third-party expertise for success. The variation of motor current, pressure, and flow with process conditions needs intelligence or human analysis to identify an anomaly or fault in the continuously varying measurements.
You can identify pre-existing conditions if you deploy Overall-Vibration with knowledge of ISO 10186 alert standards. Overall, vibration can detect failure modes. A clarification of the failure modes can help explain the misses that have occurred.
Overall-Vibration is a direct measurement for detecting and monitoring imbalance, looseness, and misalignment of a rotating asset. The units denoted for overall vibration are inches per second – peak, a velocity measure. You can use an accelerator to calculate overall vibration from an acceleration reading. The ISO 10816 defines how to measure and also sets the alert thresholds.
It is debatable whether, overall, it is predictive. Since the defect or problem already exists, overall vibration is a lagging indicator. Yet, operations and maintenance have sufficient time to fix the situation in the emergent stage. Such early repairs involve minimum costs, whilst there can be additional collateral damage- and higher repair costs- if you wait for a more extended period.
Condition Monitoring 3.0 as Predictive maintenance:
CdM 3.0 is the version of PdM adopted in the 2010s. With the introduction of Industry 4.0, the Industrial Internet of Things (IIoT) measurements for PdM are similar to 'leading versus lagging indicators. The monitoring can be good but still lagging or condition-based. For some, PdM identifies with technologies like infrared thermography (IR), ultrasonic, vibration spectrum analysis, partial discharge testing, and monthly vibration routes by trained and experienced professionals.
The introduction of IIoT has led to the emergence of intelligent processing, communication, storage, alerting, and translating. This new intelligence targets failure modes such as 60- and 90-day advance detection of lubrication defects, pump seal failure, bearing faults, and cavitation.
These failure modes are important as multiple industry studies concede that lubrication causes failure on as high as 80 per cent of rotating assets. In CdM 3.0, overall vibration in combination with ultrasonic or high frequency offers the opportunity to realize "predictive" maintenance. This scenario involves a fault condition identified 60 or 90 days in advance, permitting maintenance and operations to plan and schedule a repair with the correct parts, skills, and tools at the right time.
How Predictive Maintenance Works:
Baseline setting for Measurement: The maintenance team finalises the thresholds for the acceptable performance of assets.
Installing Interconnected Sensors: (IoT Devices): )The assets selected for monitoring need high-sensitivity sensors. These sensors can transmit real-time data for CdM. The IoT devices and the computerised maintenance management system (CMMS) are interconnected.
Data Processing and Analysis: The CMMS, when enabled with machine learning (ML), can predict potential issues, analyse collected data, and study patterns.
Manual or Automated Maintenance Scheduling: The CMMS can include maintenance decisions. Such decisions can be manually or automated. The automated maintenance tasks initiate the instant with the prediction of an issue, creating an automatic and seamless maintenance cycle.
Cost savings arising from PdM are 8-12 percent higher compared to condition maintenance and 30-40 percent higher when compared to reactive maintenance. It is clear that PdM enables industries to predict potential issues to perform proactive maintenance, and the ROI on PdM is positive.
Farnell element14 provides a one-stop platform for the key components needed to provide intelligent, real- monitoring with partners to provide the key additional products necessary for complete solutions. Some of the products categories of Predictive Maintenanceare are Test equipment,Development kits,ESD protection products,Electrical Testing Equipment, Thermal Imaging, and Machine Safety;these products, as well as the associated ecosystem of its partners, enable Predictive Maintenance in a wide range of applications in smart industries and facilities for improved reliability and operations.
