1.
What is Prescriptive Maintenance?
1.1.
Prescriptive Maintenance vs Predictive Maintenance: Key Differences
2.
4 Feasible Use Cases of Prescriptive Maintenance
2.1.
Advanced Inspection Data Processing
2.2.
Early Equipment Failure Detection
2.3.
Asset Lifespan Optimization
2.4.
Prevention of Unplanned Downtimes
3.
Final Thoughts
When it comes to critical assets, a reactive, ‘break and fix’ maintenance approach won’t cut it as it leaves businesses exposed to prolonged downtime and regulatory action. Preventive maintenance strategies, aimed at predicting future issues, get more frequently applied.
But what if you could not only predict asset degradation and failures but also receive tailored advice for avoiding the problem altogether? That is the exact promise behind the emerging generation of prescriptive maintenance systems.
Prescriptive maintenance (abbreviated RxM) is a strategy that leverages conditional, operational, and environmental asset data to anticipate asset failures and provides recommendations for preventing degradation and optimizing performance.
By analyzing vast data reserves, prescriptive maintenance systems deliver recommendations for preventing unexpected downtimes, optimizing maintenance schedules, adjusting operating conditions, and taking other corrective actions. This way, you ensure high reliability, availability, maintainability, and safety of assets. Prescriptive maintenance systems calculate the best course of action by combining predictive analytics with operational modeling, optimization and simulation algorithms, and quantitative methods for decision-making.
Such a system can go beyond individual asset failure analysis and look at broader organizational impacts. For example, it might evaluate how a malfunction in one upstream asset affects a related downstream asset. It can also show faulty habits and inefficiencies in company-wide asset management programs. Among RxM adopters in the mining industry, 9 out of 10 say its positive impact on asset management and downtime is significant.
Generally, a prescriptive maintenance system features the following tech components:
Prescriptive maintenance (RxM) is the next step from predictive maintenance (PdM). While the PdM uncovers what might happen, RxM guides you on what you should do. The two strategies are connected at the hip, although they have notable differences.
The first distinction is the scope of analysis. Predictive maintenance is all about forecasting future risks and malfunctions. Prescriptive maintenance couples predictions with data-backed corrective advice. Such systems go through multiple “what if” failure scenarios, model their outcome, and offer tailored prevention tactics.
Therefore, PdM aims to minimize unplanned downtime by predicting when equipment might fail. RxM enhances operational efficiency by prescribing actions that can prolong equipment life and optimize its performance.
Technology requirements are the second important difference. Both predictive and prescriptive maintenance systems require access to ample condition monitoring data, machinery running records, and servicing data. The difference lies in subsequent data analysis.
Predictive maintenance systems primarily rely on supervised machine learning methods and statistical approaches like linear regression or survival analysis. Prescriptive systems utilize statistical, optimization, and simulation algorithms, as well as reinforcement learning.
Here’s a summary table, comparing the most-used analytics approaches for predictive and prescriptive maintenance.
| Maintenance type | Technique category | Specific techniques | Use cases |
| Predictive maintenance | Machine learning | Linear regression Decision trees Random forests Support vector machines (SVM) K-nearest neighbors (KNN) | Indirect failure predictions Anomaly detection Remaining useful life estimates |
| Deep learning | Recurrent neural networks (RNNs) Long short-term memory (LSTM) Convolutional neural networks (CNNs) | Time-series forecasting of asset degradation Advanced visual inspection and anomaly detection | |
| Prescriptive maintenance | Machine learning | Reinforcement learning (RL) Bayesian networks Gradient boosting (XGBoost) | Optimizing maintenance schedules Recommending optimal corrective interventions |
| Deep learning | Deep Q-networks (DQN) Policy gradient methods Generative adversarial networks (GANs) | Failure and performance scenario modeling Complex system behavior modeling |
Prescriptive maintenance aims to streamline some ‘last mile’ operational tasks like work order generation, schedule coordination, and spare part ordering among others. The rapid advancement in AI, also presents opportunities for autonomous decision-making.
In theory, prescriptive systems can apply ‘self-healing’ actions, based on the estimated interventions by sending commands to on-premises edge devices or connected equipment to continuously adjust its performance. In practice, however, few operators feel confident to deploy such a degree of autonomy.
Prescriptive maintenance is still a developing field. ML and DL models require extensive validation in industrial settings to ensure accurate performance under varying conditions. Regulations also reign in some of the enthusiasm.
Nevertheless, many companies gradually introduce prescriptive practices in their maintenance process with a high degree of success. Here are four scenarios where RxM already makes a tangible difference.
Companies have an abundance of condition monitoring data. IoT sensors track asset temperature, vibrations, motion, pressure, humidity, and the presence of corrosive chemicals.
Non-destructive testing equipment provides extra insights into surface defects and material degradation patterns. For example, Voliro T has a selection of six NDT payloads for early detection of asset damage. The UT payload does A-scans for measuring thickness loss up to 2-150 mm/0.08 – 5.9 in, which is indicative of pending cracking and corrosion. Our PEC payload detects corrosion under insulation (CUI) with up to 100 mm of insulation thickness.
The problem is that with conventional strategies, most of this info goes under-analyzed and used only for corrective maintenance after the damage becomes apparent. Prescriptive systems can help you transform scattered inspection reports and standalone maintenance records into a snapshot view of your assets’ health. Algorithms can link anomalies with potential failure mechanisms and direct toward possible triggers. Then suggest the optimal approach to servicing or performance optimization.
SesnseHawk developed a digital observability platform for solar farm operators. A solar engineering, procurement, and construction (EPC) company used SesnseHawk observability platform to identify thermal defects in the final stages of construction. Using proprietary algorithms, OBSERVE analyses images collected by drones, classifies inverter damage by its cause or severity, and overlays a thermal map on assets’ digital models. Thanks to this data, EPC could speed up issue resolution and enhance expected energy yields.
Using large sets of historical and real-time data, prescriptive systems can identify defect signatures and changes in asset performance that might indicate imminent failures.
For instance, operators can receive alerts, based on acoustic wave signals that indicate stress corrosion cracking or detect temperature spikes, suggesting tensile stress. Early problem identification gives your teams a leg way to prepare: acquire necessary spare parts, schedule field trips, and notify affected teams or customers about planned downtime.
That’s the type of proactive response Pan American Energy has adopted. The company’s maintenance system by Aspen Mtell, delivered an alert about an imminent axial displacement failure in one steam turbine, due to cracks in the blower motor. With 60 days’ notice, PAE engineers had enough time to plan, schedule, and replace an on-stream 60 km3/hr main air blower at its refinery.
Another case serves as more of a warning tale. A European refinery experienced several consequent seal and bearing breakdowns in their vacuum bottom pumps. Having analyzed historical data with failure signatures, Aspen Mtell estimated timelines for the next incidents: 28-31 for seal failures and 10-28 – for bearing-related ones. Unfortunately, the team didn’t act upon the prescriptive insights and their equipment failed right on the dot.
Regular asset integrity inspections, paired with continuous condition-based monitoring and regular corrective maintenance, prolong the asset’s service life. If you do everything by the book—apply protective coatings on schedule, lubricate components, and replace worn-off parts—the asset will stay in use for longer. The wrinkle, however, is that all of these ‘best practices’ have to be carried out in alignment with other factors like asset use rate, environmental exposures, history of use, etc.
Prescriptive analytics systems produce maintenance suggestions based on a combination oflifespan factors, including manufacturer specifications, industry standards, past experiences with similar equipment, as well as current asset condition. Effectively, you get a customized upkeep schedule for each asset, continuously updated as new information becomes available.
Preston Johnson of Cutsforth, Inc. showcased how such a system works for digital pump maintenance. The company installed sensors to monitor vibration levels, motor current, lubrication, temperature, pressures, and flows. Data analytics helps define change rates, average mean variations, and correlations with pump sensor performance. Then, the system classifies anomalies by severity and urgency and recommends optimal servicing windows. With this system, the company can accurately estimate the remaining life of digital pumps and minimize disruptions to operational processes.
With prescriptive maintenance tools, asset managers can simulate multiple ‘what-if’ scenarios to predict failures, estimate their precise onset, and select remediation strategies.
This way, you address the root cause of the issue, not just the symptoms. Depending on the asset stressor, you can either limit it without halting operations or eliminate it during pre-planned servicing sessions. Either way, the result – is a prevention of unplanned downtime.
Augury developed an AI-powered platform for machine health assessments. Once connected to the data source, the algorithm will detect performance anomalies, explain their causality, and suggest the next mitigation steps. The system boasts high accuracy because it compares obtained data against a database of over 100K previously monitored performance records for similar assets. The recommendations are based on the experience of 20+ Cat III & IV vibration analysts the company employs.
In one case, Augury’s solution detected faults in the client’s equipment at the early stage and pre-calculated potential downtime duration – 3 weeks. Augury’s proactive warning gave maintenance personnel time to order spares and make replacements. As a result, the client saved almost $6.2M by eliminating potential downtimes.
Companies with mature predictive maintenance programs eye prescriptive maintenance as the next frontier. Although many RxM use cases are already algorithmically feasible, data availability is a major roadblock.
Industrial inspection drones can compensate for the limitations of hand-held NDT equipment and sensor network deployments. Voliro T is a compact, maneuverable drone, built for close-to-structure inspections. We have designed six interchangeable NDT payloads for taking precise readings on the fly (literary and figuratively). A flare stack inspection can be completed in just 2 hours (instead of 72 hours when using cranes). Full external column and piping inspections at large plants can be completed in under 3 days instead of 8 when using traditional methods.
Voliro helps detect defects as early as possible, locating wall thinning as small as 2 mm – 0.08 in and coating thinning – up to 0-1.5 mm / 0-60 mils. By combining drone-collected data with prescriptive analytics, you can better understand material degradation patterns, predict servicing needs, and apply the right corrective strategies.
