Multi-objective engineering design optimization using sequential adaptive sampling in the pareto optimal region
Systems and methods of selecting sampling points (product designs) in a multi-objective engineering design optimization of a product are disclosed. The method comprises (a) receiving a description of the product to be optimized, (b) selecting an initial set of sampling points in a design variable space of the product, (c) obtaining numerically-simulated structural responses of each of the current set, (d) deriving a set of approximate POPs from optimization using metamodels constructed from numerically-simulated structural responses, (e) establishing subregions around POF kernels that are selected from approximate POPs using “Piercing” procedure, (f) creating a set of Diversity Basis Points by populating the subregions with a space filling criterion, (g) selecting another set of sampling points from a combined group of the Diversity Basis Points and POF kernels using “Piercing” procedure, (h) reducing the subregion size, and (i) repeating (c)-(h) until a termination condition has been reached.
1. A method of selecting sampling points in a sequential multi-objective engineering design optimization of a product based on the product's durability in response to an impact event, said method comprising:
(a) receiving a description of the product to be optimized;
(b) selecting an initial set of sampling points in a design variable space configured for evaluating durability of the product in response to an impact event;
(c) obtaining numerically-simulated structural responses by conducting a computer simulation of the impact event for each of the currently-selected set of the sampling points;
(d) constructing metamodels using the numerically-simulated structural responses and deriving one or more approximate Pareto Optimal Points (POPs) from said metamodels;
(e) selecting one or more Pareto Optimal Frontier (POF) kernels from said one or more approximate POPs with a first point selection procedure that maximizes minimum distance amongst the selected POF kernels;
(f) establishing one or more subregions encasing the respective POF kernels, each subregion being associated with a size;
(g) creating a set of Diversity Basis Points by populating said subregions according to a space filling criterion;
(h) selecting another set of sampling points from a combined group of the set of Diversity Basis Points and the POF kernels using a second point selection procedure that maximizes minimum distance amongst the currently-selected set of sampling points; and
(i) reducing the size of said subregions and repeating (c)-(h), until a termination condition has been reached and the currently-selected set of sampling points represent optimal designs of the product.
2. The method of claim 1 , wherein the product comprises an automobile.
3. The method of claim 2 , wherein said durability in response to the impact event comprises crashworthiness of the automobile.
4. The method of claim 1 , wherein said computer simulation comprises a time-marching simulation using explicit finite element analysis.
5. The method of claim 1 , wherein said termination condition comprises a maximum number of iterations.
6. The method of claim 1 , wherein said termination condition comprises a minimum size of the subregions.
7. The method of claim 1 , further comprises merging the currently-selected set of sampling points into the previous set of sampling points.
8. The method of claim 1 , wherein said one or more subregions are overlapped at respective borders.
9. A system for selecting sampling points in a sequential multi-objective engineering design optimization of a product based on the product's durability in response to an impact event, said system comprising:
a main memory for storing computer readable code for an application module;
at least one processor coupled to the main memory, said at least one processor executing the computer readable code in the main memory to cause the application module to perform operations by a method of:
(a) receiving a description of the product to be optimized;
(b) selecting an initial set of sampling points in a design variable space configured for evaluating durability of the product in response to an impact event;
(c) obtaining numerically-simulated structural responses by conducting a computer simulation of the impact event for each of the currently-selected set of the sampling points;
(d) constructing metamodels using the numerically-simulated structural responses and deriving one or more approximate Pareto Optimal Points (POPs) from said metamodels;
(e) selecting one or more Pareto Optimal Frontier (POF) kernels from said one or more approximate POPs with a first point selection procedure that maximizes minimum distance amongst the selected POF kernels;
(f) establishing one or more subregions encasing the respective POF kernels, each subregion being associated with a size;
(g) creating a set of Diversity Basis Points by populating said subregions according to a space filling criterion;
(h) selecting another set of sampling points from a combined group of the set of Diversity Basis Points and the POF kernels using a second point selection procedure that maximizes minimum distance amongst the currently-selected set of sampling points; and
(i) reducing the size of said subregions and repeating (c)-(h), until a termination condition has been reached and the currently-selected set of sampling points represent optimal designs of the product.
10. The system of claim 9 , wherein said one or more subregions are overlapped at respective borders.
11. A non-transitory computer readable storage medium containing instructions, when executed in a computer system, for selecting sampling points in a sequential multi-objective engineering design optimization of a product based on the product's durability in response to an impact event by a method comprising:
(a) receiving a description of the product to be optimized;
(b) selecting an initial set of sampling points in a design variable space configured for evaluating durability of the product in response to an impact event;
(c) obtaining numerically-simulated structural responses by conducting a computer simulation of the impact event for each of the currently-selected set of the sampling points;
(d) constructing metamodels using the numerically-simulated structural responses and deriving one or more approximate Pareto Optimal Points (POPs) from said metamodels;
(e) selecting one or more Pareto Optimal Frontier (POF) kernels from said one or more approximate POPs with a first point selection procedure that maximizes minimum distance amongst the selected POF kernels;
(f) establishing one or more subregions encasing the respective POF kernels, each subregion being associated with a size;
(g) creating a set of Diversity Basis Points by populating said subregions according to a space filling criterion;
(h) selecting another set of sampling points from a combined group of the set of Diversity Basis Points and the POF kernels using a second point selection procedure that maximizes minimum distance amongst the currently-selected set of sampling points; and
(i) reducing the size of said subregions and repeating (c)-(h), until a termination condition has been reached and the currently-selected set of sampling points represent optimal designs of the product.
12. The non-transitory computer readable storage medium of claim 11 , wherein said one or more subregions are overlapped at respective borders.