January 2020

Process Optimization

RAM analysis for refinery process design optimization

RAM refers to reliability, availability and maintainability analysis. The RAM model uses discrete event simulation (DES) software and provides a quantitative assessment of the performance of an industrial plant. This article discusses the use of RAM analysis on petroleum refinery projects to select an optimum process configuration and associated storage requirements.

Gandhi, S., Kortnicki, H., Nangia, K., Fluor Enterprises, Inc.

RAM refers to reliability, availability and maintainability analysis. The RAM model uses discrete event simulation (DES) software and provides a quantitative assessment of the performance of an industrial plant. This article discusses the use of RAM analysis on petroleum refinery projects to select an optimum process configuration and associated storage requirements.

The RAM model for an overall refinery includes all major process units, intermediate storage tanks, interconnect routings, and operating logic defining normal operation and actions upon process unit failures. The model calculates individual process unit onstream factors, assesses the intermediate tankage requirements, and provides an estimate of the expected overall refinery onstream factor (OSF). The objective is to establish a cost-effective design to maximize refinery production.

Refinery block flow diagram

The starting point for the RAM model is the overall refinery block flow diagram (BFD). The diagram shows the major process units in the refinery, proposed unit design capacities, crude throughput, product yields across each unit and the flowrates of the main process streams at normal operation. FIG. 1 shows a simplified schematic of an overall refinery BFD.

FIG.1. Overall refinery block flow diagram (BFD).
FIG.1. Overall refinery block flow diagram (BFD).

Availability block flow diagram

The next step is to prepare an availability block flow diagram (ABD), as shown in FIG. 2. ABDs are pictorial representations of the series and parallel steps needed to perform a given function. The availability values for each stream (represented as hexagons) are shown as a percentage of the overall refinery production. The process unit capacity values (represented as squares or blocks) are shown as a percentage of the unit design capacity to the nominal operating rate. The size and location of the intermediate tankage are also shown on ABDs. The tanks are typically bypassed for energy conservation, but are available for use during failure or maintenance outages of the downstream unit.

FIG. 2. Availability block flow diagram (ABD).
FIG. 2. Availability block flow diagram (ABD).

Unit mechanical reliability data

An important input to the RAM model is the mechanical reliability data for the different process units. The reliability data determines the failure frequency and restoration times for these units. TABLE 1 lists process reliability data for typical refinery process units based on historical industry data for similar units, augmented with information compiled by the authors’ company based on past projects. Reliability data is shown as a function of three different tier levels that are based on a refinery’s operating and maintenance practices and account for factors such as the level of trained maintenance personnel onsite, the availability of spare parts, etc.

Process unit failure frequency and restoration time are entered into the model to match the reliability factors. It is important to properly characterize unit downtime parameters since the shutdown duration directly affects intermediate storage tank inventory levels. The model incorporates proprietary data for mean time between failure (MTBF) and mean time to restore (MTTR) specific to each unit type of refinery process unit [crude distillation unit (CDU), coker, hydrotreater, etc.]. The MTTR consists of repair time and other restoration activities, such as detection, preparation, drainage, cooldown and return to service.

The authors’ company database categorizes the MTTR into different groups, such as short-term, medium-term, medium-long and long-term duration outages. For example, the following MTTR may apply:

  • Short-term restoration MTTR, 0 hr–18 hr
  • Medium-term restoration MTTR, 24 hr–42 hr
  • Medium-long-term restoration MTTR, 54 hr–64 hr
  • Long-term restoration MTTR, 120 hr–220 hr.

Depending upon unit type, the failure frequency for each of the above four outage duration groups is a fixed percentage of the overall failure frequency for the group. For example, the hydrotreater failure frequency for each outage duration group may be:

  • Short-term restoration = 40% of total unit failure frequency
  • Medium-term restoration = 30% of total unit failure frequency
  • Medium-long-term restoration = 20% of total unit failure frequency
  • Long-term restoration = 10% of total unit failure frequency.

Intermediate tankage

Tanks designated for storing intermediate process fluids are entered into the RAM model and can have a significant impact on refinery availability, particularly for failures with short- and medium-term restoration times. Optimization requires selecting the proper balance between tankage volume and unit processing capacity.

The service, number and capacity of the storage tanks are defined prior to the RAM analysis. Some tanks may be existing tanks at the refinery, while others may be new tanks proposed as part of the project expansion.

Support units

Support units such as the sulfur recovery, amine regeneration, sour water stripping, hydrogen production and utilities systems can also be included in the RAM model. The design capacity, as well as the number of trains for these support units, have impacts on overall refinery availability. As the number of trains increases, availability of the overall refinery improves; however, capital costs for the support unit also increase. Based on RAM analysis, an optimum selection can be made. If the reliability of support units is high compared to the process units, the support units can be excluded from the overall RAM model.



The model uses discrete event simulation software with programming language to permit the user to simulate complex operating logic necessary to respond to and recover from process unit failures, to address intermediate tankage reaching full or empty conditions, and to evaluate different “what-if” scenarios.

To determine the overall refinery onstream factor (OSF), the model simulates the following expected scenarios:

  • The normal steady-state operating condition under which all process units are operating at their nominal rates, with the refinery producing product streams at their nominal production rates.
  • Upset conditions caused by failure of one or more process units, in which part or all of the intermediate stream is rerouted to other process units or to available intermediate tanks. This permits the upstream process units to continue operating.
  • Recovery conditions following failure events or maintenance activities in which some process units may be temporarily operating at full design capacity to empty intermediate tanks that have filled during the previous upset condition.

To simulate the randomness associated with equipment failures and restoration cycles, the model uses probability distribution functions (PDFs) and Monte Carlo techniques to sample the PDFs as the model moves through time to calculate the final products.

The capacity loss associated with process unit failures is logged and accumulated over an extended period (referred to as “mission time”). Each mission is typically assumed to have a duration equal to 30 yr. The result is an estimated overall refinery OSF for one “mission.” Due to the randomness associated with unit failure and restoration data, the model calculates the OSF for multiple missions. The results from these multiple runs are statistically analyzed to establish the mean OSF, along with the associated standard deviation.


The authors’ company has applied RAM analysis on a variety of projects, including petroleum refineries, gas processing facilities, upgraders, gasification units and chemical complexes. The analysis has been used to finalize the process unit capacities, optimize tankage, evaluate the impact of different process configurations on the OSF, and provide the required technical basis to support key project decisions. Subsets of the RAM model have also been used for turnaround planning to evaluate the availability of portions of the refinery that are not in turnaround mode. Some typical examples from refinery projects are provided here.

Impact of unit design capacity

The first example is from a RAM analysis conducted for a grassroots refinery project to evaluate the impact of increasing process unit design capacities on the overall refinery availability. TABLE 2 shows the overall refinery availability as a function of process unit design capacities.

For this specific refinery, the planned activities, such as scheduled turnarounds and catalyst change-outs, were estimated to reduce OSF by an additional 2%. Therefore, the OSF for this refinery, after accounting for both unplanned and planned shutdowns, was estimated to be between 90.1% and 93.9%, depending on the design capacities selected.


FIG. 3 is a composite plot of the OSF vs. the percentage of the total number of missions. The plot is used to calculate the mean and the standard deviation for the OSF. For this specific example, the mean OSF is 98.6% with a standard deviation of +/– 0.2%.

FIG. 3. Onstream factor histogram.
FIG. 3. Onstream factor histogram.

Process unit operating log

The RAM model develops several operating logs in Excel format to assist in the review of the results. TABLE 3 is an example of a snapshot of the process unit utilization log generated by the model. The example is taken from a project on an existing refinery that was being upgraded with the addition of a new delayed coker unit and several other new process units.

The process unit utilization log shows the “% time operating” (same as unit onstream factor) and accounts for unavailability due to “% time failed” and “% time idle.” The “% time failed” is due to failure of that specific process unit and includes the time to repair and restart the unit. The “% time idle” is due to a shutdown from a lack of feed from the upstream unit, or due to the inability to route the products to the downstream unit, or to the intermediate tankage.

The process unit utilization log also records each downtime, including the time to restore the unit. It also records the idle times, which specific tank caused the process unit to enter the idle state, and the duration, for which the process unit was idle.

Tankage utilization log

The RAM model also generates a tankage utilization log that shows the average level of each tank over the duration of the run. The tankage log can be used to optimize the allocation of the existing tanks or to specify new tankage requirements.


The RAM model simulates the overall refinery operation over an extended period of time. The authors’ company has applied the model on numerous projects to optimize the overall refinery process design during the engineering phase of the project. The model has been used to finalize the individual process unit design capacities, evaluate different train configurations, select the intermediate tankage requirements, and estimate the overall refinery OSF for use in economic analyses of different design options. HP

The Authors

Related Articles

From the Archive



{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}