Vision-based vibration measurement using speckle pattern and system identification using autoregressive model

Abstract

Non-contact sensing techniques are a currently active and vibrant research field on structural health monitoring community because they are cost efficient and simple to use in test field. The non-contact sensors reduce expenses and installation process because they enhance the accessibility of a harsh structural spot that are difficult to approach. Many researchers have studied non-contact sensing techniques to solve civil engineering problems, such as: a non-contact laser ultrasonic scanning system visualizing the cracks at a distance, and a vision-based measurement to measure the displacement signal at a remote place. This study introduces a new non-contact sensing technique for vision-based displacement measurement. Some researchers have already focused on vision based displacement measurement. They utilized target panels or robust objects as the reference point and their device successfully detected the movement of either target panels or robust objects. The displacement signal obtained by their vision-based measurement gave satisfactory resolution to the dynamic properties of the target structure. The new method proposed in this study doesn’t require a physical reference point but it utilizes laser speckle pattern to create the optical reference point. The speckle pattern occurs when a coherent red light is diffusely reflected on a rough surface. The set of scattered photons from the specular surface is called the speckle pattern. In this study, the laser beam is spread over a target structure surface. Under the constant incident ray, a slight tilt of the target surface changes an incidence angle. The diffusely reflected rays, speckle patterns, intensively react against the small movement of the target. A camera outside of the target structure records the sensitive response due to the vibrations in real time. The movement of the target due to ambient vibrations is estimated by image processing algorithms, such as: the laser speckle contrast imaging for detecting the movement, and k-means clustering for estimating virtual nodal points. The two consecutive algorithms estimate the actual displacement. The laser speckle contrast analysis (LASCA), which is popular methods to visualize blood flow in biomedical community, quantifies the blurring extent of each raw speckle image. Because the blurring represents the motion during a camera exposure, the movements of the target in stationary images is digitalized by LASCA. With speckle contrast images in chronological sequence, tracking a single pixel in different images is infeasible. This study utilizes the k-means clustering algorithm. The displacement signal is estimated by linking the location of k clusters centroids in each interval. The computed displacement signal in time domain is decomposed to frequency domain by frequency domain transformation techniques such as: frequency domain decomposition, or stochastic subspace identification. This study utilizes both eigendecomposition for system identification and frequency domain decomposition. To validate the method two lab scale experiments were done; one was done with a shaker and a steel plate and the other was done with a small steel beam model. In the first experiment, the steel plate on the shaker was vibrating under assigned frequency. The laser beam was spread over the steel plate and the camera recorded the speckle pattern created. The purpose of the experiment was to prove that the speckle pattern recognition algorithms exactly reflect the dynamic property of the vibrating target. In the second experiment with a small steel beam model, the laser was shot on the center of the beam. The purpose of this experiment was to check if the speckle pattern could detect and reflect ambient vibration. These two experiments prove that the speckle pattern recognition method can identify the vibrations of the targets.