How did asset utilization modeling save Sadara $1 B in capital cost?
Sadara is a JV between Saudi Aramco and Dow Chemical.
Sadara is a JV between Saudi Aramco and Dow Chemical. The two companies invested $20 B to build the world’s largest petrochemical complex ever constructed in one phase (FIG. 1). This project included new process units that had never been built before in Saudi Arabia, or in any of the countries of the Gulf Cooperation Council (GCC). In addition, some of the complex’s units process risky and challenging chemicals, whereas others operate at world-scale rates. With such a large investment in a complicated, highly integrated site, it was important that the money be well spent to earn the rate of return investors expect.
FIG. 1. A view of the Sadara petrochemical complex. Photo courtesy of Sadara.
As most engineering and construction companies are more interested in minimizing the initial investment, operating companies such as Sadara have a different objective: They want to balance the initial capital investment against the production and the conversion costs necessary to make their products over the life of the assets.
During the design phase, it is important to factor in or discern how different equipment and process units will affect availability and asset utilization, so that the plant can meet business production targets within cost constraints. This action is more difficult than it sounds, since even an experienced engineer has difficulty predicting actual production in a large, highly integrated site, given different unit availabilities, complex process unit configurations, equipment failure and repair, seasonal effects, fluctuations in supply and demand, and intermediate storage capacity. This is a classic optimization problem that requires the right methodology and tools. For that reason, Sadara used reliability, availability and maintainability (RAM) modeling as a key part of its design approach.
Overview of asset utilization modeling
RAM modeling is a Monte Carlo mathematical methodology that allows a user to evaluate different design scenarios, assuming a given process/equipment configuration, pipe flow, tank management and equipment failure/repair characteristics. Model outputs are asset utilization, availability and the costs to achieve that asset utilization over a given time period. It is important to note that, since Monte Carlo modeling is a probabilistic mathematical technique, model outputs are also probabilistic (e.g., outputs are probability distributions, rather than specific discrete values).
RAM models can be built to any degree of detail. However, in practice, these models should be built only to the detail required to answer the specific business/technical questions under consideration. Sadara uses three different “levels” of RAM models: RAM1, RAM2 and RAM3.
RAM1 site model
RAM1 is the highest-level model and is a single model built for the entire site. Constructing the initial model required incorporating process unit configurations, tankage sizes, import/export capabilities and turnaround schedule estimates, as well as power, utilities and infrastructure information. As opposed to the RAM2 models used at the plant level, the RAM1 model uses simple uptime/downtime estimates for each major process unit.
Initial RAM1 model performance data was based on historical data from Sadara’s parent companies, and from worldwide refining and petrochemical best practices. Over time, these initial data estimates will be replaced with actual Sadara operating data.
Process units, tanks and other major assets are modeled as simple blocks with little underlying supporting detail. What is important in these models are the interconnections and dependencies between the units.
In the initial phases of the Sadara program, RAM1 models were crucial for determining which process units would be built and the ultimate storage configuration. Production revenue determined from the RAM1 model was then balanced against the costs to build, operate and maintain these units, as determined from engineering estimates. However, the RAM1 model remains evergreen, and is updated as new business input becomes available.
Primary outputs of the RAM1 model are “utilized hours,” along with storage and import/export capabilities over a 20-yr time frame. The term “utilized hours” is best understood as the percentage of time that the unit must be available at full production. Simply put, it becomes the yearly asset mechanical reliability (AMR) percentage over that 20-yr time horizon.
FIG. 2. The output of the RAM1 model sets the target availabilities for each of the underlying 26 process unit models.
Overview of asset utilization modeling
The output of the RAM1 model sets the target availabilities for each of the underlying 26 process unit models (FIG. 2).
RAM2 plant models
Whereas there is only one RAM1 model, there are RAM2 models built for each of the 26 process units and utilities. RAM2 models are much more detailed than the RAM1 model. Blocks are added for each major equipment item, using typical failure and repair characteristics for that equipment. The model is then run, and adjustments are made that add or subtract redundancy, using more/less reliable equipment (obviously with cost differences) until the RAM1 and RAM2 model outputs converge. Outputs from the RAM2 models include the asset utilization of the plant, the total number of failures for all equipment and the percentage of production losses that each equipment caused, in addition to tank full/empty events.
Of importance is the Pareto chart of equipment vs. asset utilization loss for every piece of equipment in the unit—often called the “sensitivity” level. These charts can often be used to determine what equipment performance could be improved to increase overall asset utilization. They are frequently used by reliability engineers to identify potential projects that would increase asset utilization for that unit. Tanks’ empty/full events suggest whether the tanks are properly sized, or if they need to be increased to improve asset utilization or decreased to save capital cost.
RAM3 models
RAM3 models are smaller, and usually more specific, models that are often built to answer a specific question. They have frequently been used for spare parts optimization or case studies. The information needed to conduct a spare parts optimization analysis includes failure and repair characteristics, replenishment time, the cost of spare parts and spare parts production sensitivity. Model outputs are the minimum/maximum stocking levels that guarantee production is never lost because of spare parts unavailability, and that spare parts are not overstocked.
RAM modeling benefits
Sadara has received RAM modeling benefits from numerous sources. The following are some of the more common benefit areas, such as the ability to determine:
- When redundancy is needed or when the system is over-designed
- How reliable a given piece of equipment should be
- When, where and how much tankage is needed
- Which equipment improvement projects make the most sense economically
- Optimal quantities of spare parts for given equipment
- Effectiveness of preventive and predictive maintenance (PPM) and turnaround frequency/duration
- Load-shedding strategies when unplanned events occur.
One benefit often received is the determination of whether redundancy is needed or not. Sometimes added capital cost can save a future loss of production, resulting in huge benefits. Conversely, there may be times when redundancy is not needed and, in these cases, redundancy can result in capital spent unwisely, along with possibly a lifetime of unnecessary maintenance costs. At Sadara, in two separate instances, entire trains were eliminated.
Determining the reliability of a piece of equipment or a component within a given piece of equipment is not a trivial matter. RAM modeling makes it possible to accurately determine the cost/benefit of improved reliability or to select between vendors. This was used extensively during the RAM modeling effort.
Tankage is usually desired for operational flexibility reasons. However, it often cannot be justified economically. RAM modeling allows operators to pinpoint precisely how much tankage is needed—if any at all—between process units, raw materials storage or finished product storage. If RAM modeling had not been used, the Sadara complex would have been built with more, and larger, tankage than necessary.
RAM modeling is also very good at determining the sensitivity of equipment, which leads to asset utilization losses. Once known, the cost for projects to eliminate or reduce those losses can be calculated to determine which projects make the most sense financially. This aspect of RAM modeling remains an important part of Sadara’s reliability program on an ongoing basis.
Spare parts optimization has become an important part of Sadara’s asset management effort. In the Middle East, this is especially important in the Gulf region, since lead times are much longer than along the US Gulf Coast. Spare parts optimization strikes a balance whereby no production is lost due to the unavailability of spare parts or the overstocking of spares.
PPM and turnaround schedules are often set to maintain certain equipment to keep them in running condition and to prevent unplanned outages. Once a PPM or turnaround has been completed, the model resets the age of these maintained equipment parts to a specific value, as determined by equipment experts. This way, credit can be given to the renewed parts by adjusting their age. This process results in a more accurate availability and asset utilization estimate for future model runs.
Setting load-shedding strategies is of great importance to handle process interruptions. It helps maintain focus on assets based on their importance to safety, production and asset performance.
Takeaways
By utilizing RAM modeling, Sadara saved more than $1 B in capital spending, while simultaneously increasing production capabilities and achieving business objectives. These benefits accrued from the elimination of process trains, decreased capital equipment spend, a more robust electrical grid and supporting components, the addition or elimination of redundancy where appropriate, adjustment of tank sizes and spare parts optimization.
The real validation that RAM modeling was beneficial has been the operational results. After the plant had been fully commissioned and all units were operating for several months, the overall plant availability was slightly greater than 97%. No doubt this was due to many factors. However, it is unlikely that such a high availability could have been achieved without the extensive use of RAM modeling.
While it played a very important role during the design phase, RAM modeling continues to be an essential component of Sadara’s asset management program. For example, reliability engineers use RAM2 models as a source of ideas on how a process unit’s availability can be improved by targeting equipment projects based on sensitivity results.
In 2018, Sadara successfully completed its performance test, the creditors reliability test (CRT), an exercise required by the company’s financial lenders to demonstrate the company’s ability to operate its market-facing plants and supply chain at 90% of design capacity or better for a period of 60 consecutive days. From October 19 through December 17, Sadara engaged in its first official run of the CRT, monitored by the lenders’ representative. All plants and supply chain facilities passed their tests safely on the first official try. New industrial companies very rarely pass lenders’ reliability tests on the first official attempt, and Sadara was the first Saudi Aramco JV ever to do so. This is a validation of the overall asset management approach that was taken on the project, and there is no doubt that RAM modeling played an important role in achieving that important milestone.
At the very least, asset utilization modeling should be an essential part of every capital project. Models created during the design phase can then be leveraged to answer future operational and maintenance issues based on a rigorous analysis rather than on a gut feeling. Used this way, asset utilization models complement the process simulation models that process engineers already maintain for each unit. HP
The Authors
Hayek, F. - Sadara Chemical Co., Jubail Industrial City, Saudi Arabia
Fareed Hayek is a Senior Reliability Simulation Engineer at Sadara Chemical Co., specializing in RAM modeling and spare parts optimization. Previously, Mr. Hayek was a reliability and improvement engineer at Dow Chemical in Freeport, Texas. Prior to that, he was a senior reliability and maintainability engineer at NASA for the International Space Station (ISS) and the Space Shuttle programs. He received the Space Flight Awareness Honoree award for reliability and maintainability support of the ISS. He holds ASQ CRE and CMRP certifications. Mr. Hayek earned a BS degree in mechanical engineering from the University of Jordan, an MS degree in computer engineering from the University of Houston, and is a PhD candidate in industrial engineering at Mississippi State University.
Moran, M. - Sadara Chemical Co., Jubail Industrial City, Saudi Arabia
Marty Moran is a Senior Reliability Engineer at Sadara Chemical Co. He is trained as a chemical engineer and has more than 35 yr of experience in process industries. He has concentrated his career on using advanced computer technology in the areas of asset management/reliability, advanced process control and plant optimization. Mr. Moran holds a US patent for multivariable control. Prior to joining Sadara, he worked for AspenTech, Meridium, Continental Controls and Setpoint.
Related Articles
From the Archive