ModelCenter® Explore drives innovation and improves product quality by enabling users to thoroughly explore and understand the design space, make better decisions, and find optimal solutions.
ModelCenter Explore allows users to:
- Run powerful algorithms and trade study tools
- Search, investigate and understand the design space
- Incorporate multiple variables (cost, performance, risk)
- Visualize results and the impact of design changes
- Find optimum solutions
Simulation and trade-study results can be used to discover important trends and tradeoffs.
Quickly generate many design alternatives
Once a repeatable engineering analysis process has been created, engineers can repeatedly execute the process (using parallel computing resources if available), with each execution corresponding to a different set of inputs. This allows engineers to explore and quantify the performance, cost, reliability, and risk of a large number of different design alternatives in a relatively short period of time.
Identify important variables
Sensitivity analyses allows users to analyze their DOE data and understand which of their input variables has the most impact on their output variables.
Video – Sensitivity analyses allows users to understand which of their input variables has the most impact on their output variables.
Advanced visualization plots allow users to visualize their design space, understand variable relationships, investigate the effects of constraints on their solution, and perform “manual” optimization (Figure 2). Visualization of “Pareto plots” allow engineers and stakeholders to understand the relationship between key metrics such as performance, cost, risk, and schedule (Figure 1).
Approximate Long-running analyses
ModelCenter Explore allows users to create fast-running approximations (Response Surface Models) for long-running analysis tools. In essence, Response Surface Models are general-purpose multi-dimensional “curve fits”. Given a set of data points (a set of inputs and outputs), ModelCenter can generate a quickly executing mathematical model (the response surface) that approximates this data. Several different approximation techniques can be used to find a mathematical model that works best for your data). Charts and statistics are generated that help users assess the quality of the approximation and improve it as necessary. Once a response surface is created, it can then be added back into your ModelCenter workflow where it can be run just like any other component in the model. The user can utilize the response surface along with ModelCenter’s ‘If-Then’ workflow statements to switch back-and-forth between the original long running component and the quickly running (but approximate) response surface model).
With ModelCenter Explore, engineers can take advantage of advanced mathematical optimization algorithms to help them search for better designs.
Specify your goals
The engineer starts by specifying their goal (objective) – the workflow variable or variables that they would like to either maximize or minimize. Next, they specify which of their input variables (design variables) they will allow the optimization algorithm to modify as it searches designs that best meet that goal. Optionally, engineers can also specify a set of requirements (or constraints) for their problem. For example, the user may wish to maximize the performance of their design, but constrain the solution so that manufacturing costs remain below a specified amount.
Choose an algorithm
Once the objective, goals, and constraints are specified, the engineer can choose any of the built-in optimization algorithms to solve their problem. ModelCenter Explore includes over 25 different algorithms, each with their own strengths and weaknesses. Because most engineers are not optimization experts, an Algorithm Wizard is included to help users choose the algorithms best suited to solving their problem.
Add your own algorithm
Engineers can also add their own custom algorithms to ModelCenter. Programming is required (.NET or Java), but the optimization SDK provides a complete set of instructions and examples. Once added, custom algorithms will be available in the ModelCenter Explore alongside the default algorithms.
When the algorithm runs, it will repeatedly execute the workflow (changing the design variables each time) in an attempt to find one or more designs that maximize the user’s goals while also satisfying all of the constraints. Engineers can utilize the charts and reports generated by the optimization run to gain insight into the nature of their design problem and ultimately to find an optimal design.
How reliable is my design?
Reliability analysis tools can be used to help users access and understand the impact of uncertainties on their engineering analysis results (Figure 3). For example, a designer may wish to better understand the effect of manufacturing tolerances on the performance and/or failure probability of a product. ModelCenter Explore includes a Monte Carlo analysis tool that will help engineers to make these assessments. When used in conjunction with an optimization algorithm, the Monte Carlo tool allows users to perform robust and reliability based design.
How long will it take?
Accurate Monte-Carlo analysis can requires many thousands of workflow executions. If this is not feasible for a given workflow, there are two alternatives:
- Create a response surface model for any long-running components, and then run a Monte-Carlo analysis using the response surface models instead of the actual components. Be careful – make sure the response surface models are accurate!
- Utilize one of ModelCenter’s Advanced Reliability Algorithms.