52 Y. Reuland et al. Hall boundaries (walls above slab) Structural walls below slab Non-structural walls below slab Sensor Locations 1 2 3 4 Step Location Walking pattern for Human detection 5.70 m 5.90 m 13.90 m 18.20 m Fig. 6.1 Schematic view of the tested slab with sensor locations and walking pattern −2 0 2 0 6 12 18 24 30 36 42 48 Time (s) Filtered Signal (Bandpass 18-24 Hz) Detection Thresholds Acceleration (mg) Fig. 6.2 Successful human detection from sensor 1 The feasibility of human detection is shown for a tested walking pattern (see Fig. 6.1) on the slab. For the filtered signal of sensor 1 (see Fig. 6.2), footsteps that are close enough to the sensor, and that are not damped out by a structural wall underneath the slab, can be successfully detected. This example shows that even for a stiff slab having short spans, an accelerometer is able to measure the impact of a human step (person of approximately 85 kg) in a radius of more than 4 m. It is also interesting to note that steps that are done in different directions do not have the same amplitude. This might be linked to path-length dependent gates and is a promising feature for user behavior studies. As shown in Fig. 6.3, all four sensors (see Fig. 6.1 for sensor locations) showed the single footstep that is subsequently used for localization. Sensor 1 is closest to the footstep and logically recorded the highest acceleration. The amplitude of filtered acceleration for sensor 2 is comparable due to the absence of walls supporting the slab between the two sensors. Sensors 3 and 4 show a slight exceedance of the triggering thresholds, which shows that steps can be detected even if supporting walls are between the step source and the sensor. However, this observation is not true for all the steps that were analyzed and therefore, the sensor configuration that is necessary for robust human detection may need to take into account the walls and other types of supports underneath floor slabs. Localizing the presence of a human walking on the floor is treated as an inverse problem. The recorded vibration data is used to update knowledge about location of the occupant on a slab using EDMF. In this application, the only model parameter that is intended to be identified using EDMF is the location of an occupant. For application of EDMF, a finite element model of the hall is developed using ANSYS. Using the location of an occupant as a variable, a unit load is applied along a grid on the model and the vertical displacement at sensor locations is recorded. The model response is subjected to uncertainties from many sources, such as modulus of elasticity of the concrete deck, locations and stiffness of boundary conditions. Also, uncertainties arise from processing the measured vibration data using numerical integration in order to obtain displacements, as well as from the weight of the occupant and dynamic amplification
RkJQdWJsaXNoZXIy MTMzNzEzMQ==