Learn how we helped a large energy producer drive a more proactive maintenance approach and quantify future risk and costs by simulating thinning and vibration scenarios through a QRO pilot.


A 1.5 MMbd gas-oil separation and stabilization facility wanted a more quantitative reliability model for its stabilization and stripping section to forecast availability and optimize its maintenance costs.


In three months, Pinnacle piloted Quantitative Reliability Optimization (QRO) to model failure degradation on 56 of the facility’s fixed and non-fixed assets. LVC models were used to predict failure and simulate process condition changes for thinning and vibration scenarios.


The model calculated a forecasted baseline availability of 94.22% for the 56 assets. The facility is able to drive a more proactive approach to maintenance by dynamically monitoring the impact of real-time facility data to understand its impact on availability, risk, and cost.


A global super-major with an integrated portfolio of upstream and downstream assets wanted to better quantify its decision-making processes to maximize reliability across its fixed and non-fixed assets at one of its gas-oil separation facilities. The facility, which went into service about ten years ago, was an early adopter of technology and looked to remain at the forefront of the industry in reliability and maintenance digital transformation.

The Challenge

While the facility had implemented a risk-based inspection (RBI) program, the facility’s reliability and integrity programs operated in data silos. This siloed approach made it difficult to quantify maintenance priorities and left the leadership team with a qualitative approach to managing integrated downtime events and annual maintenance plans. This static, siloed program, coupled with the facility’s varying feed composition and rates, caused the maintenance teams to be in a constant reactive state and spend a significant amount of its budget on unplanned costs.

While management trusted their data, they were unable to quantify the future failure dates of the facility’s assets or predict how future equipment corrosion and damage across fixed equipment, rotating machinery, and instrumentation would affect the availability of the facility. Facility leadership recognized that to be at the forefront of the industry, they would need to adopt a more proactive and quantitative approach to managing assets and needed a solution that would provide three primary elements:

  • Scenario Comparisons:

    While the risk of failure was relatively low for the facility’s assets, leadership wanted the ability to predict failure dates and visualize the impact of future corrosion and machinery issues on the facility’s overall risk.

  • Risk Assessment:

    Facility leadership wanted to have a more comprehensive risk assessment of its assets and visualize the impact of its data over time.

  • Availability Forecast:

    Facility leadership could not predict its assets’ future availability.

Pinnacle’s Solution

The facility decided to pilot Quantitative Reliability Optimization (QRO) for its capability to visualize the impact of data over time. QRO is an evolution in reliability modeling that combines the best traditional reliability methods, data science, and subject matter expertise into a hybrid model. The QRO approach provided a framework for this facility to achieve its goal of bringing the best of digital transformation to its reliability and maintenance work processes, satisfying the three solution elements above.

The QRO pilot included a select group of 56 assets within an oil train unit. This group of selected assets was critical to the success of crude oil production and included compressors, pressure vessels, heat exchangers, and pumps. The QRO pilot was completed over three months and had the following scope of work:

  • Data Collection:

    The Pinnacle team extracted data from the facility’s Process Flow Diagrams (PFDs), Piping and Instrumentation Diagrams (P&IDs), Corrosion Control Documents (CCDs), Corrosion Loop Drawings, SAP asset list, and Process Description Handbook. Additional data exports included an RBI program export, work order history export, and vibration monitoring. Because the facility has only been in service for about ten years, little to no corrosion issues were found for the selected group of assets. However, facility leadership still wanted to visualize the impact of future corrosion issues on facility performance despite this insight to mitigate risk proactively.

  • Data Organization and Analysis:

    The data collected during the first step of the process was then loaded into Newton™, a software that facilitates the QRO methodology. This data included the asset register and hierarchy, asset attribute data (material of construction and operating conditions), work order history, inspection and monitoring data (vibration and thickness readings), Probability of Failure (PoF) data (functions, failure modes, and failure mechanisms), and Consequence of Failure (CoF) data (representative fluids, volume, and production impact). During this step, Pinnacle conducted a work order history review to categorize each event into predictive, corrective, or preventive maintenance activities. This step was critical for identifying and understanding bad actors and their associated failure modes and frequencies.

  • System Model and Asset Risk Analysis (ARA):

    All assets were linked within the Newton™ system model. The system model is used for availability forecast calculations and shows how data from each asset impacts overall facility performance. After all available data was loaded into Newton™, an Asset Risk Analysis (ARA) was created. An ARA analyzes an asset’s function and specific performance requirements and creates causal links between the asset’s failure modes and potential failure mechanisms to calculate risk. This calculation also uses data science models to create Lifetime Variability Curves (LVCs) for each vibration and thickness monitoring point, which are used to accurately predict component failure dates. An LVC is a dynamic model that predicts the probability of failure (PoF) over time by applying data science principles, subject matter expertise, and historical plant data. These curves show how the likelihood of failure for each component changed over time and was used for each vibration and thickness monitoring point to predict component failure dates more accurately.

  • Calculated Availability:

    The Pinnacle team calculated the availability forecast for the oil train by leveraging the asset interdependencies from the system model with the dynamic LVC and PoF curves.

  • “What If” Scenario Comparison:

    The facility was able to run various “What If” scenarios for the oil train through the Scenario Comparison module in Newton™ to simulate the impact various failure modes have on an asset’s failure date. During this pilot, the facility modeled a severe thinning scenario on a dehydrator and an excessive vibration scenario on a pump.

  • Final Results:

    At the conclusion of the pilot, the facility received an availability forecast, risk results, and scenario comparisons, described in more detail below.

With QRO, the facility was able to model two “What If” scenarios based on current data and visualize the impact that specific measurements have on the long-term risk of the asset. Two of these “What If” scenarios have been detailed below, including a severe thinning scenario for a dehydrator and an excessive vibration scenario for a pump. With both these examples, management was able to quantitatively predict equipment damage and associated risk to enable better maintenance and inspection planning.

Scenario 1: Thinning

The first “What If” scenario that QRO modeled was a severe thinning scenario. Figure 1 is a thickness LVC for a dehydrator developed using the facility’s thickness measurements. Because the facility had low measured corrosion rates, the initial predicted failure date of the dehydrator was May 2100, and the PoF for this asset was 0%.

Figure 1: Thickness LVC for a dehydrator

However, the facility wanted to visualize what would happen if the dehydrator were to experience severe corrosion. To model this scenario, the Pinnacle team replaced the most recent thickness measurement with a lower value. Figure 2 shows the updated LVC with a new predicted failure date of March 2025 and an updated band of uncertainty, displayed in blue.

Figure 2: Modeled Severe Corrosion Scenario

To demonstrate the decision-making and planning functionality of Newton™, a repair task was planned for the dehydrator. The results of this plan can be seen in Figure 3, which displays the current and mitigated risk curves.  With the addition of a planned repair task, the “risk with plan” curve (dashed grey line) drops to zero.

Figure 3: Dehydrator Risk vs. Mitigated Risk Curve for Severe Corrosion Scenario

Scenario 2: Vibration

The second “What If” scenario that QRO modeled was a severe vibration scenario. Figure 4 is a vibration LVC for a pump bearing within the crude train created with the facility’s vibration data. The analysis did not show any immediate onset of damage with a predicted failure date of December 24, 2030.  However, due to the nature of the failure mode, there is a high degree of uncertainty displayed on the LVC by the blue band.

Figure 4: Vibration LVC for a Pump Bearing

The Pinnacle team added exponentially increasing vibration measurements to the model to visualize the impact vibration will have on the pump bearing. Figure 5 shows the updated LVC with a new predicted failure date of early 2022.

Figure 5: Excessive Vibration LVC

The Pinnacle team added exponentially increasing vibration measurements to the model to visualize the impact vibration will have on the pump bearing. Figure 5 shows the updated LVC with a new predicted failure date of early 2022.


As a result of piloting QRO, the facility was able to visualize the impact of its data on failure and risk over time. The deliverables of the pilot included:

  • Scenario Comparisons:

    The most valuable result for the facility was its ability to model potential thinning and corrosion scenarios on all equipment types. By simulating various “What If” scenarios, facility leadership will be able to identify predicted failure dates and the associated PoF curves for both fixed and non-fixed assets within the platform.

  • Risk Assessment:

    The facility received a risk profile over time for each asset, allowing them to visualize the impact of historical and future activities and identify when assets cross various thresholds.

  • Availability Forecast:

    QRO calculated a forecasted availability of 94.22% for the selected group of 56 assets within the oil train, as shown in Figure 6. Previously, the facility was unable to forecast availability, but with QRO, the facility can now proactively predict availability and failure dates to better manage its assets. The primary drivers for the loss in availability were a lack of measured thickness data in a set of heat exchanger tube bundles and a lack of recorded maintenance history in a group of compressor seals. The lack of eddy current reports for the heat exchanger tube bundles resulted in an over-conservative estimated corrosion rate that increased the PoF.

Figure 6: Availability Forecast for 56 Critical Assets within the Oil Train


After completing the pilot for the oil train, the facility is moving forward with implementing QRO on the remaining assets within the oil train and other gas trains at the facility, which includes 2,000 fixed and non-fixed assets. The expansion of the QRO implementation is expected to result in a 14% total inspection spend reduction, 6% total maintenance spend reduction, and 2-3% total utilization improvement through reduction of loss of containment events and downtime.

At the conclusion of the expanded QRO pilot, the facility will receive:

  • 10-year availability forecast, which can be drilled down into PoF prediction for any equipment component

  • Top 10 contributors to unavailability and cost, which is updated as new data is uploaded into Newton™

More resources like this