Figure 1. The process design basis should cover a wide range of areas.
A typical PDB pulls together information on a variety of project parameters (Figure 1). Each of these may have multiple subsections. Depending upon the specific job, the subsections may move around. For example, a plant modification to improve safety will treat the safety enhancement as the objective. In another case, say, when the goal is to increase capacity, the safety section may appear within constraints.Let’s look at what the parts of a PDB should cover:Objective
. What should the plant modification achieve? Is the goal increased safety, higher capacity, changed products, greater energy efficiency, improved reliability, lower pollution or something else? The objective should be clearly defined. For projects with multiple aims, this section should spell out the relative priority of each as well as methods for evaluating tradeoffs among them.Inputs
. What goes into the plant? What is the feed? How does it vary? Often overlooked is that various feeds may have different properties or that feed rate, conditions and compositions may change over time.Outputs
. What does the plant make? What are the product specifications? What ranges of specifications might be required at different times? This section should include specifications that can be quantified and more-general requirements such as odor, haziness, taste, etc.Constraints
. What limits the plant today? What constraints are imposed by regulation, capital availability, schedule or other factors that the engineering team can’t modify? These may include restrictions on the size of the unit, availability of utilities, type of cooling and other general process requirements.Boundaries
. What are the limits on temperature and pressure of all streams entering and leaving the unit?General Configuration and Special Requirements
. Must the engineering team stick with a particular process choice? For example, must a specific option —such as pressure swing absorption versus membranes for hydrogen purification — be used or avoided? What level of reliability is required, e.g., how many days a year must the plant be available or what production target must be met?Uncertainty
. What is unknown or might vary? What risks require mitigation? How are the risks to be evaluated? PDB documents frequently fail to address these issues. Most often, no information is given on what the uncertainties are or how to handle risk evaluation. Risk here specifically refers to the process risk of not meeting a production or economic target. Defined regulatory criteria usually cover environmental and safety risks, so those risks are best thought of as constraints on the engineering team.The most common approach to uncertainty is to assign a set of ranges to specific plant objectives and limits. The engineering team then checks the capability of the future plant to handle all the possible combinations of results. This approach attempts to ensure that all objectives can be met across an entire operating envelope. The benefit is a highly flexible plant. The cost is that meeting certain extreme cases may require large amounts of capital. A better way to use capital may be to sacrifice some capability, capacity for example, to handle extreme, but rare, situations.Available tools can provide better insight into economic risk. However, because of tight schedules, they rarely are used. (Not surprisingly, projects with overly shortened schedules often suffer from poor capital allocation.)Chief among these tools to reduce economic risk are Monte Carlo analysis and carefully conducted test runs. Monte Carlo analysis applies statistical distributions to possible values of selected inputs, outputs and constraints. Multiple evaluations are run using a sample of statistical values for the selected factors. On generally well-understood processes, the analysis can provide impressive insight on how to save capital or reduce economic risk.Thorough and accurate test runs yield detailed understanding of plant constraints and capabilities. But such runs rarely are done. What most plants call a test run involves gathering barely more than usual operating data plus a few extra process measurements and samples. High-quality test runs include detailed data analysis to provide for independent confirmation of every measurement that affects the likely investment decisions.The Monte Carlo analysis enables evaluating uncertainty. The plant test run provides information on the constraints of the process. Both are key parts of defining how to handle economic risk in the PDB.
ANDREW SLOLEY is a Chemical Processing contributing editor. You can e-mail him at [email protected]