Simulation Experiments (Part 2)
In my last blog post (Simulation Experiments (Part 1)) I introduced the topic of creating and running simulation experiments. As I mentioned, experiments are a way to ask system performance questions and get corresponding answers – nothing new here, since system modeling and simulation were developed specifically for this purpose. What is new, however, are methodologies developed to ask the questions, calculate the answers, and analyze the results.
Defining an experiment starts with deciding what you want to know about your system – in other words, the questions you want to ask. For systems, you usually want to quantify one or more performance metrics like system gain, risetime, overshoot, delay, etc. With questions in hand, the next step is determining which design parameters contribute most to performance variations. You might argue that a key reason for running design experiments is figuring out which parameters to tweak in order to get a desired performance – and you can certainly create experiments to do just that. But knowing ahead of time the list of key design parameters improves your experiment efficiency. But how do you determine which parameters most affect a performance metric? Two words: sensitivity analysis (see my blog post “How Sensitive is Your System?”). Knowing the performance metrics you want to focus on, and the parameters that most affect the metric, defines both the analysis and measurements you need to run. You are now ready to setup your experiments.
Experiment Manager, SystemVision’s application for defining and running simulation experiments, is a GUI-driven integration connecting SystemVision and Microsoft Excel. Using this integration, Excel can access the SystemVision design database as well as send simulation and analysis commands to the simulator. After starting Experiment Manager from the SystemVision menus and naming your experiment, you define your experiment in 4 easy steps:  select the parameters you want to control during the experiment — each parameter defines a unique column in the experiment spreadsheet;  choose the analysis command file, which controls the analysis type and parameters;  select your performance measurements, including any minimum and maximum limits – each measurement defines a unique spreadsheet column;  set parameter values in the experiment spreadsheet. Once your experiment is defined, simply click the Simulate button in Excel to run the experiment. Simulation and analysis commands are sent to SystemVision; experiment results are displayed graphically in the SystemVision waveform analyzer, and in tabular format in Excel. In the waveform analyzer you can further analyze experiment results using both graphical measurements and transforms. In Excel you can use built-in number crunching capabilities to further analyze numeric data.
Experiment Manager simplifies the creation, execution, and management of simulation experiments. In my next blog post I will show you an example of using Experiment Manager to define and run experiments on a mechatronic system.