Optimization Guidance Steps

The process of gas detector optimizations can be daunting for even experienced engineers. There are many steps to consider, each having several layers of decisions to be made. We hope to help users overcome this hurdle with the below guidance steps. Feel free to contact info@insightnumerics.com with any questions or comments. 

 

  1. CAD Model Prep - when performing several 1000s of CFD cases, any time taken to simply geometry can significantly reduce computation time of the entire project. Consider having a MASTER file with the full CAD model and then dividing up the study by module or unit - keeping original CAD for the module needing analysis and replacing far-field CAD items with porous regions. Ventilation sensitivity studies can be performed on the MASTER file to determine volume porosity values needed to accurately model your site.

  2. Porous Region Definition - Grated vs plated decks have a large influence on gas dispersion results. CAD models typically define decks as plated when they are in fact grated. Porous regions should be defined using Grated Deck as the Type and an Area Porosity near z=0.9. Porous regions representing far-field CAD objects or modules should use volume based porosity (Type as Porous Region), using porosity values based on sensitivity studies.

  3. Gas Composition Definition - multi-component gas compositions should be defined using the available options in the DIPPR Database. Note that in:Flux only simulates components in the gas phase - liquid streams will need to be "flashed" to determine input parameters.

  4. Leak Definition - the risk manager's leak frequencies tab can be used to define leaks individually or groups at a time. Use the Properties editor at the bottom of the window to edit leaks in bulk or make copies of sets of leaks (e.g. duplicating a set of leaks to have a larger hole size). Frequency values can be entered for each leak individually or divided across a set using the Cumulative Frequency option. Time should be taken here to accurately define each leak location and direction. 

  5. Wind Rose Definition - wind direction and speed probability values can be entered for up to 12 wind directions. However, each wind direction and wind speed defined will exponentially increase the number of simulations run. It can be beneficial to perform some initial dispersion cases to determine sizes of gas clouds produced for various wind conditions. This information can then be used to justify a smaller set of wind directions and speeds for the project. Doing so can save hours or days of computation time. For example, a project with 20 gas leaks and 12 wind directions and 2 wind speeds would result in 480 dispersion simulations, if each took 30 minutes to run that results in 240 hours of computing time. A project with the same number of leaks and only 4 wind directions and 2 wind speeds would result in 160 dispersion cases results in just 80 hours of computing time (a third of the earlier amount).

  6. Simulation Summary Review - Once leaks and wind cases are defined they combine to make the dispersion cases of the project and are listed in the simulation summary. Exporting this table can be helpful in double checking all the entered information. Each column can also be sorted in this table so you may determine that only cases about a certain frequency should be simulated (this is only possible if frequency information is entered for each defined leak). Use Ctrl+Shift to highlight the cases to be added to the project or click the All checkbox so select all cases. It is suggested to have the Run Now checkbox unselected so that risk data sets and monitor regions can be defined before calculation starts.

  7. Risk Data Set and Monitor Region Definition - Risk data sets are used to combine all the simulated cases for one variable, most commonly this would be the Flammable Gas Volume Fraction, %LFL. The gas detector optimization will only be performed on the information generated within this region. The risk contours will also be limited to this region. Having large or multiple risk data sets in the file will significantly increase the file size of the project (>10GBs). It is recommended to only have one or two risk data sets per file and for it to be contained around the region of the assessment. When risk data sets are defined users are prompted if a monitor region should be made with the same dimensions, click yes as this will be used to generate consequence values later. It is important to perform this step BEFORE calculations start as if they are defined after simulations are complete it will take several hours to extract the needed information.

  8. Create Distributed Solver Files and run with ifx:Solve - rather than have all simulations calculated on one machine, solver files and the ifx:Solve software can be used to distribute the computation of the project to other machines to be run in parallel. This greatly expedites computation times. Using cloud based virtual machines, 2000 dispersion cases can be completed in under 48hours. Determine the available computation power you have available before deciding on the number of solver files to create. It is recommended to use 3-4 ifx:Solve instances per machine. Once solver files have been created and moved to their perspective machines to calculate, ifx:Solve can be automated with .bat files to assist in starting up the calculations.

  9. Updating the main in:Flux project - Upon completion the solver files will be on the range of 500MB to 2GB each depending on your project. Move them back to the directory which they were created overwriting the original files. In the main in:Flux project Update All the solver files, this will take some time. Save the project when this has completed.

  10. Applying Consequence Values - with the solver files updated gas cloud volumes for each case will exist. Check the Gas Cloud Analysis window to ensure they have been updated before opening the Risk Manager to the Simulation Summary tab. Here you can choose the consequence method deemed most appropriate for your project. One option would be setting the Method to By Volume (LFL-UFL) and the Scaling to Cube Root (Normalized), although several other options exist. 

  11. Monitor Point, Line and Group Definition - This step may be interchanged with Step 12 below. Define Monitor Points (point gas detectors), Monitor Lines (open path gas detectors) and Monitor Groups in the project ensuring that their Purpose is set to Detector Optimization. These are locations that the optimization will choose from for an optimal layout. Monitor groups can be selected or removed in the optimization controller, which can be useful for comparisons of existing layouts to the optimal ones. The more monitors which are defined the better the optimization will be. 

  12. Performing Gas Detector Optimizations - Select the alarm settings for the point and open path devices as well as the alarm and control targets. Optimizations are quick to calculate so you can compare different targets or alarm values. It may be useful to set the Algorithm to Optimal Layout which calculates the mathematically optimal layout of point gas detectors without limiting itself to the locations defined by the user. This is useful as a “first pass” to get an understanding of how many devices may be needed for a module before more detailed analysis is undertaken. Additionally, once a layout has been generated by the optimization algorithm, a new risk data set can be made and risk contour viewed for the new risk data set which will indicate regions with undetected gas so new monitors can be added. This process then can be repeated multiple times to eventually converge on an optimal layout.