46 T. Kamper et al. be acquired and processed. During the step of data acquisition, the user must configure settings for the frequency response functions (FRFs) like the time to be recorded, block size, windowing of excitation and/or response. Sensor ranges must be adapted to induce a good SNR without overloading. Depending on the structure, there is a bad reproducibility at some measurement points (i.e. non-linear behavior) and in the case of impact testing, the user might be confronted with double hits. Once the data acquisition is complete, the modal parameters can be extracted in the next step wherein the curve fitting method must be selected. The next experience driven decision is to further parameterize the curve fitting method itself. The method outputs pole candidates which then must be grouped, selected and extracted by the user. Finally, the modal parameters consisting of mode shapes with corresponding eigenfrequencies, and damping values form the modal model which can be evaluated in the postprocessing step. Fig. 1 Process of experimental modal analysis and steps that require user interaction, affecting result completeness and quality. This paper focuses on the validation of solutions for two different steps in the process of modal analysis: The initial step of test design for the modal analysis and the step of parameter extraction where the user must provide input parameters for the curve fitting algorithm. Both steps and the corresponding decisions to be made by the user will be described in the following sections. Moreover, the suggested solution approaches based on AI will be discussed. To validate the suggested solution approaches a study with participants that have different experience levels regarding modal analysis was performed. The study demonstrated the advantages of the suggested approaches and will be discussed in the last section of this paper. Identified Problems and Solution Approaches Test design: Reference DOFs identification In the step of test design, the user must decide among other on the correct number and positions of reference DOFs. The correct number and positions of reference DOFs is important to make sure that every eigenfrequency present in the frequency range of interest can be detected. This is particularly critical because a missing eigenfrequency, and thus an incomplete modal model, can hardly be detected by the user. The importance of the correct choice of reference DOFs can be illustrated by the simplified examples given in Figure 2, where the first modes of two simple oscillating structures are depicted. If the reference DOF is placed at the position marked in red, the respective mode cannot be detected because the reference DOF is placed in a node of the mode shape. Placing the reference DOF at or near a node of the mode shape is not suitable for introducing energy into the structure to excite the specific frequency if used as an excitation DOF, nor for detecting a vibrational amplitude if used as a measurement DOF. Using simple examples like they are given in Figure 2 the decision on the number and positions of reference DOF is simple and straight forward. But for real, complex structures with a lot of modes this decision is not easy and cannot be made without any previous thoughts and analyses. For the identification of the number and positions of reference DOFs different methods exist. They can be divided into two groups by the fact whether a numerical model is needed to get insights into the dynamic properties of the structure or not. • Methods based on numerical simulations: Based on the modal deformation derived from numerical simulations it is possible to decide on the number and positions of reference DOFs [1, 2]. In [3] a method was introduced which works based on a non-validated numerical model in a fully automated manner.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==