Technical systems must be continuously improved so that they can remain competitive on the market. Also, the time-to-market is an important factor for the success of a product. To achieve this goal, new methods and processes are needed. Especially the testing and calibration are important phases in the development process.
This paper introduces a method, which helps to reduce the time effort while increasing the quality of the calibration process. The basic idea is to use measured test data to parameterize a physical (or mostly physical) model structure to create adequate models for the optimization. The main advantage of the method is the reduction of test effort because the number of variations of the design parameter is one, or extremely decreased (depending on the system). Another advantage is that the uncertainty and the limit of the model can be quantified more accurately compared to common approaches based on non-physical model structures. These normally use artificial neuronal networks (ANN) or polynomial approaches for the test-based optimization.
This contribution illustrates the method by using the example of the calibration process of a double clutch gearbox (DCT) regarding energy efficiency and drivability on a roller test bench. First step is the test planning and test execution. In this step the method calculates the optimal execution order of the measuring points. In this example 81% timesaving can be achieved compared to the equivalent on the test track. The second step is the automated generation of the simulation model. In this step the unknown parameters of the model structure are calculated. The contribution shows different approaches for the identification of non-linear systems. In the last step the model is used to perform the optimization of the design parameters.
