Tornado diagrams are one of my favorite tools for visually representing the results of a techno-economic model. They highlight the most influential input parameters in a way that is clear and easy for anyone to understand. In chemical technology development and evaluation, they can be used to identify and target the process parameters with the greatest effect on profitability. They can also help identify risks, like a strong dependence on raw material prices.
Technically speaking, the tornado diagram is a form of deterministic sensitivity analysis; that is, it compares the results of a number of discrete cases. To build a tornado diagram, you begin by listing the parameters that you want to examine. These might include process parameters (e.g. conversion, temperature, product specifications), engineering parameters (e.g. production capacity, storage capacity, materials of construction), and market parameters (e.g. raw material prices, product prices, discount rate).
For each parameter, you then specify three values: worst case, expected case, and best case. Sometimes these cases are referred to as P90, P50, and P10, to represent the 90th, 50th, and 10th percentiles of the probability distribution for the particular variable. For process parameters, these values can be specified based on input from the scientists and engineers working on the technology. For market parameters, they can be specified based on historical data.
The ‘tornado diagram’ is centered on the result value when all input parameters are set to their ‘expected case’ values. Each bar deviating to the left or right from this central axis represents the result of changing the corresponding parameter from its ‘expected case’ value to either its ‘worst case’ or ‘best case’ value. A longer bar represents greater sensitivity to that parameter. The parameters are then arranged in order of decreasing impact, and this gives the diagram its tornado shape.
So, let’s look at the tornado diagram below, which shows the effect of eleven different parameters on internal rate of return for a chemical process. What does it tell us? For one, it tells us that single-pass conversion is by far the most influential process parameter. It also tells us that reactor pressure, catalyst price, and catalyst lifetime are relatively unimportant within the ranges specified. Research and development efforts should therefore focus on increasing single-pass conversion, perhaps even if it comes at the expense of operating at a lower pressure or decreasing catalyst lifetime. After a few months of work, an updated tornado diagram might show that single-pass conversion has been improved to a point where it’s no longer the most important parameter, and efforts should be shifted elsewhere.
A market parameter, Feedstock 1 price, is also very high on the diagram. This indicates a potential threat, an external factor that has a strong influence on the profitability of the process.
In the video below I show how tornado diagrams can be used to quickly view and understand the results of a techno-economic model from different angles.