*3.1. Experimental Setting*

As shown in Figure 2, the four-layer frame model is composed of plexiglass plates and aluminum columns, which are connected by bolts (due to the universality of the application of a frame beam in practical production, the frame structure is taken as an example. The non-destructive testing algorithm is used to evaluate and monitor the strength, stiffness, damping and other technical parameters of the frame beam, which is of great practical significance to eliminate potential safety hazards and ensure the safety of the frame structure [29]). The structure sits on orbits and is only allowed to move in the x direction. Each layer of the structure consists of four aluminum columns (25 × 25 × 1 mm) connected with two up-and-down plexiglass plates (900 × 450 × 2.5 mm), respectively, forming a four-degree-of-freedom system.

**Figure 2.** Frame structure and the sensor arrangement.

In Figure 2, an electromagnetic exciter (SA-JZ-50, Wuxi Shiao Technology Co., Ltd, Wuxi, China) is used to apply the lateral excitation to the bottom plate along the center line of the structure. The structure and the exciters are installed together on the base plate. A force sensor (BK-2Y, China Academy of Aerospace Aerodynamics, Beijing, China) (sensitivity 2.23 mV/N) is installed between the rod of the exciter and the structure to measure the excitation force. Four acceleration sensors (AD100T, Qinhuangdao Xinhua Technology Co., Ltd, Qinhuangdao, China) (sensitivity 100 mV/g) are installed in the positions shown in Figure 2 to measure the acceleration response of each layer. The excitation signal is the random excitation with a bandwidth of 5~50 Hz and the excitation level is 2.6 V as shown in Figure 3.

**Figure 3.** Measured random excitation signal.

The sampling frequency and sampling time of the experimental test system were 160 Hz and 25.6 s respectively, and 4096 points were sampled. The experiments were tested on 18 different working conditions, including increasing the mass of a certain layer, reducing the stiffness of a certain column, and introducing nonlinear damage into a buffer gap. This paper only analyzed the identification effectiveness of the algorithm on the stiffness reduction of Column 2 in Figure 2. We only selected four representative working conditions for research in this paper, as shown in Table 1.


**Table 1.** Physical experiment condition of the frame structure.
