Merge branch 'master' of https://github.com/SlaybaughLab/Coeus

Author: jamesbevins

UseCases.md

@@ -9,7 +9,6 @@ The Coeus inputs allow for the geometry to be fully specified in 2D for this cas
 The contraints considered are weight and number of fissions (reactions).  These are set through the user inputs.  The Read_MCNP_Output function gets the weight and reactions from a set tally structure.  The weight is implemented as a hard constraint, and the # of fissions is incorporated as a soft contraint in the objective function calculation.  
 ###Objective Function
 The objective function is calculated in the RelativeLeastSquares function.  It is a weighted sum of the squares of the differences between the two spectra.  
-###Other Considerations
 
 ##Boron Neutron Capture Therapy(BNCT)
 ###Overview
@@ -31,4 +30,74 @@ The best way to do this is to use case #2.  The current approach of a "diagnosti
 ###Objective Function
 The objective function is to maximize the dose to tumor to dose to patient ratio.  
 The current code can do this by taking the known spectrum that accomplishes this minimization as the objective and using the current spectra comparing objective functions. 
-The better approach would be to use case #2.  The current approach of a "diagnostic volume" could be generalized to allow for a patient's head to be specified.  Then, an ability to have user specified tallies needs to be added.  The objective function would then look to maximize the ratio of the two doses that were computed in the tallies.    
+The better approach would be to use case #2.  The current approach of a "diagnostic volume" could be generalized to allow for a patient's head to be specified.  Then, an ability to have user specified tallies needs to be added.  The objective function would then look to maximize the ratio of the two doses that were computed in the tallies.   
+###Challenges
+1) Implement a flexible user geometry input that allows for parts of the problem to be fixed and parts to be variable. 
+
+##Radiation Shields and Converters
+###Overview
+There are many applications covered under this scope that might be useful. Here we describe two of the more interesting/challenging cases.  These numbers are carried forth throughout this section for clarity.
+1) Space shielding design: For satelites, space reactors, and future space travel, light weight, efficient shields are required.  This is one area in nuclear engineering where more advanced optimization techniques have been applied in the past, but many of these problems are still solved by a parametric study.
+2) Radiation converter design: It is sometimes useful to convert one type of radiation into another.  
+###Geometry
+The geometries for these applications tend to be fairly straighforward in the traditional layered concept.  However, matrices of material have been introduced and show some promise for advancing each.  The matrices can be handled explicitly by modeling 3D repeating or stochastic lattices or implicity by modleing homogenous material with a varying composition fractions.  Either is available under Coeus, but which is more appropriate depends on the actual material.  
+###Constraints
+Constraints for both tend to be mass and thickness.  There can be cost factors as well.  Given the simple geometries, this weight is easy to check before running MCNP, which could enhance efficiency by tossing constrained designs and generating a new design before run time.
+###Objective Function
+The objective function for 
+1) Either a dose threshold or a dose reduction percentage.  This will require tally a flux and implementing a flux to dose conversion tally.  This is all on the user input side.
+2) This typically is either a target conversion rate (or percentage) or a generated particle type rate. These are both straighforward tallies and can be implemented within the imput from the user.
+###Challenges
+1) Ensure that the input reader can handle MCNP universes.
+2) May require the use of alternative codes. 
+
+##Medical Isotope Production
+###Overview
+Medical isotope production is a fairly big business, and it appears that the target designs are optimized by experience, back of the envelope calcuulations, parametric studies.  The basic idea here is that you want to generate as much of isotope X as posssible while avoiding other isotopes that are difficult to seperate from the irradiated sample.  Based on the relative cross-section of the two or more reactions, there will be an optimal beam energy that is obtained by a) tuning the incident neutron beam  or b) degraders.  Since beam is expensive and limited, often a stackup of irradiations is performed at once to generate multiple medical isotopes in sequence as the beam downshifts in energy through the irradiated sample and degraders.  These beams can be either charged particle or neutrons.  
+###Geometry
+Typically a stackup of layered materials.  It maybe possible that a 2D matrix would also be desirable, primarily for neutron beams.  
+###Constraints
+Typically there are not difficult constraints here outside of physics.  Space an mass can be a limit depending on the facility.  This may be more of a consideration for neutron beams than charged particles.  You could incorporate maximum bad/good isotope ratios here if the goal is to produce the maximum number of atoms up to some contamination percentage. 
+###Objective Function
+Objective functions can be the sum of the number of atoms created for each isotope (with or without preference weighting) ot the ratio of bad to good isotopes. It is possible that both of these will be the objective function of interest and need to be combined.  Each calculation is possible to be done inside of a tally, but the combination and weighting needs to be handled elsewhere in post-processing.
+###Challenges
+1) The ability to run other codes. However, MCNP is probably capable of most of this work and would be the most liekly to be used.
+2) The ability to have user specified objective functions made from a combination of a multiple tallies. 
+
+##Fusion Blanket Design
+###Overview
+Generating sustainable fusion is only half of the problem.  Once that happens, to make it economically viable, the energy must be harvested and new tritium must be bred.  This is typically done in a blanket that surrounds the core.  
+###Geometry
+These typically are in some layered sphereical or eliptical geometry.  There is typically some level of symmetrry, and the problem can be simplified to 1D or perhaps 2D, to accelerate the calculations.  
+###Constraints
+There is typically a minimum tritium breeding ratio and energy capture rate.  There could also be a maximum tritium breeding ratio. There may also be some wieght/size constraint, but these tend to be rather large.  
+###Objective Function
+While there may be a lower bounds for the minimum tritium breeding ratio and energy capture rate, these are typically what you want to maximize.  This can be treated as a multi-objective function, or they could be combined into a single objective (either weighted or unweighted).  
+###Challenges
+1) Implement materials that can varying weight/atomic fractions that follow some requirements/relationships dictated by materials science. 
+2) Multi-objective optimization
+
+##Inverse Problems
+###Overview
+There are many applications covered under this scope that might be useful.  The general concept is that you have a measurement, and you want to know what the source causing that measurement was.  This could be complicated by an intervening environment that is either known or variable.  
+###Geometry
+Here the geometry can be anything.  It can also be a fixed value or have variable components.  The source can itself be a variable (energy, distribution, size, location, etc). 
+###Constraints
+The constraints will vary widely depending on the specific application, but some to consider are masses of materials, ordering of materials, types of sources, and source intensities, among others. 
+###Objective Function
+This can be formulated many ways but in essence is a comparison to data measurement(s)/observation(s).  This will likely require post-processing of multiple tallies.  It is possible that multi-ojective optimization would be an improvement, but it is believed that these can be solved with single objective optimization.
+###Challenges
+1) It is likely that a different code will be desired for this application, so that interface needs to be easy to move from MCNP to something else.
+2) Ensure variable sources are addressed by the inputs
+3) Multi-objective optimization
+4) The ability to have user specified objective functions made from a combination of a multiple tallies. 
+
+##Reactor Design
+###Overview
+These range from assembly level to full core calculations.  
+###Geometry
+###Constraints
+###Objective Function
+###Challenges
+1) The big question here is if the cost of the calculation can be brought in-line with the number of function evaluations required to optimize. This is not just a question of radiation transport efficiency however.  The choice of the number of variables, variable domain, weight windows, and code will go a long ways to determining the run time.  This is outside of Gnowee/Coues development, but something to consider.
+2) It is likely that a different code will be desired for this application, so that interface needs to be easy to move from MCNP to something else.