Chi-Lun Wu

Micro Cracks Recognition of Concrete Bridge Surface Based on Digital Image

keywords: inspection of the bridge, micro cracks, image processing, image measurement, artificial intelligence, preventive maintenance 

In recent years, more and more attention has been drawn to the damage accumulated overtime on large structures such as bridges for transportation. For the safety of road users, the structures require preventive maintenance and the following reinforcement to extend their service life and prevent them from being damaged again by natural disasters. As a result, the practical implementation of structural health monitoring has become more and more important these days. Currently, the DER&U evaluation method is commonly adopted for monitoring the bridges in Taiwan. Conventionally, the evaluators will perform direct visual inspections on the bridges. However, as the bridges have been constructed higher and higher, evaluators find it more difficult to directly perform visual inspections nowadays without the adoption of other indirect visual inspection strategies. 

This study aims to overcome the difficulties in direct visual inspection with modern technologies. After obtaining the digital images through wireless devices, we then introduced artificial intelligence (AI) into image processing to identify the microcrack properties on the surface of concrete bridges. The width value detected by the bridge inspection vehicles will then be used to validate the accuracy of the analysis data, aiming to provide the relevant information about the cracks on the concrete bridges for the evaluators' references when identifying the health condition of the bridges. It is obvious from the experiment results that the method proposed in this study greatly decreases the time and labor needed in comparison with the current evaluating method. Also, this method is able to detect most of the microcracks on the concrete bridge surfaces and offer complete information about the microcracks. 

Identification Machining State and Specimen Quality by Using Digital Image Correlation

Keywords: Digital Image Correlation, Noncontact measurement, Stability Lobes Diagram, ASTM B557M, Decision Tree, Solidworks CAM, Milling Health Monitoring System. 

This study composes a machine monitoring system (MMS) for the milling vibration by parameter design, sensor fusion, feature extraction, and machine learning. Experiment uses a 3-axis accelerometer and dynamometer to prove the Digital Image Correlation (DIC) results. Decision tree is built by extracting the features of DIC, and the stability lobes diagram (SLD) is established as the standard of machining parameters. Selecting machining parameters of the SLD collocate computer-aided design (CAD) and computer-aided manufacturing (CAM) to create ASTM B557M specimen G-codes for CNC milling machines. DIC decision tree classifies results and uses the Keyence 3D microscope measuring the roughness to discuss different parameter specimens. Validating parameter and identification system of the CNC machine via standard CAD/CAM is uncomplicated and is cost-effective. The MMS is applicable for predicting chatter in general circumstances, enhancing the productivity of smart manufacturing. 

Research on Active Noise Resistant System of Speaker

Keywords: adaptive filter, sound pressure level, precision measurement.

This paper hopes to establish a system for measuring and modeling single characteristics. By cooperating with Foxconn tech, using the measuring instrument developed by KLIPPEL, the same monomer performance, such as Transfer Function, Peak to Peak Signal, sound The Sound Pressure Level (SPL) is used to verify the accuracy of the measured data and the processed program. Then use the measured input signal and output signal to calculate the mathematical model of the adaptive filter, which is used to predict the output signal in the system. Then use the adaptive filter establish a control model for the pre-processing of the input signal so that the output signal can be closer to the original signal to eliminate distortion. This study will be divided into two parts, the speaker system setup and the filter design. The first part is to establish a system that is effectively available, mainly using scale signals to measure the characteristics of the monomer. It is then verified by a commercially available measurement system. The second part is to use the complex frequency signal to excite the monomer characteristics, and use its characteristics to train the mathematical model of the adaptive filter. Finally, a method to eliminate delay distortion is proposed

https://hdl.handle.net/11296/pdwnje

Development of High-precision Full-field Stereo Digital Image Correlation based on Structured Light

Keywords: digital image correlation, non-contact measurement, stereo digital image correlation, tensile test, compression test, strain, full-filed measurement, three-dimensional deformation, structured light

With a series of advantages compared to the traditional sensors, Stereo DIC has been widely applied in experimental mechanics recently for its properties of non-contact, full-field and three-dimensional deformation measurement. Base on the Stereo DIC measurement system developed by the laboratory itself, this study combined with the principle of grating-projection encoding to quantify the image district for convenience of measurement, reducing the waste of computational operation, increasing the range of measurement as well as the precision. First of all, operate the measurement experiment extending the specimen made of polydimethylsiloxane (PDMS) based on the standard model of ASTM D638, its property of hyper-elastic would also cause a huge-scale displacement in image normal vector. Through the full-field analysis by Stereo DIC, obtain the data of deformation, biaxial strain and Poisson’s ratio, then deeply discuss for the partial area and prove the result with single-point measurement by laser displacement sensor. At the result, apply the improved DIC system to another experiment that compressing the jounce bumper repeatedly, taking advantage of the district encoding to certify the capability for measurement with large-curvature and rugged surface, being able to analyze the local strain of abnormal specimen and mutually verified against the finite element method (FEM) simultaneously.

https://hdl.handle.net/11296/djgb49

Development of a Pressure Sensing System with Intelligent Piezoelectric Film and its Application to the Life Evaluation of Punching Mold

keywords: piezoelectric films, support vector machine, graphene, punching state, piezoresponse force microscopy, isotropic cantilever plate experiment, real-time monitoring, preventative maintenance, early warning

In this paper, the PVDF piezoelectric film used to measure the stamping signal of the stamping press. We embedded the film within the mold to obtain direct measurements of the reaction forces imposed by the punch. Next, we used Support Vector Machine algorithm to train a mathematic model by multi-dimensional feature vector such as frequency domain、correlation coefficient, etc. The model can classify the current state of health of the mold. We also used graphene and Polyvinylidene Fluoride (PVDF) to make a new type of piezoelectric film. The composition and structure of the material observed by Scanning Electron Microscope and Piezoresponse Force Microscopy or isotropic cantilever plate used to verify the sensing ability of the film. In the end, we established the pressure sensing system with the intelligent piezoelectric film. The system can achieve real-time monitoring and preventive maintenance or early warning. In addition, the data related to the mold uploaded to the cloud server, which can establish a data log for the user to query. After measuring the workpieces through the stereo microscope, data statistics can prove that the accuracy of the mold failure classification of this system can be as high as 99.2%.

https://hdl.handle.net/11296/agn489

Developing Core Algorithms for 2D/3D Image Correlation and The Precision Measurements of Strain and Poisson s Ratio for ASTM Specimen

keywords: large deformation, full-filed measurement, stress and strain, tensile test, non-contact measurement, digital image correlation

In this paper. Fixing the DIC algorithm to overcome that DIC can’t track large deformation and using high-resolution image to measure displacement field and strain field of Standard specimen of ASTM B557M. The experiment result are consistent with strain gauge that express DIC has the ability to measure strain in tensile test. Combine DIC and MTS to get Stress-Strain curve, further get Youngs modulus of elasticity, Yielding Stress and Ultimate stress. Also Using Stereo DIC to measure True stress-strain of specimen.

https://hdl.handle.net/11296/xjfwz3

Multi-scale Measurement and Vibration Inspection for Reinforced Concreate Structure, Coordinate Measuring Machine and Industrial Motor by using Digital Image Correlation

keywords: coordinate measuring machine, large structures, multi-scale, digital image correlation

This thesis uses the digital image correlation method to measure the multi-scale experimental measurement. DIC method solves the problem that traditional sensors are difficult to measure quantitatively by non-contact and full-filed measurement characteristics. In the first chapter, the digital image correlation method is used to quickly and easily detect the inter-layer drift, drift angle and column strain response of large structures under seismic vibration, and verify the accuracy of DIC method with linear displacement sensor and accelerometer. The second experiment was combining the DIC method with the horizontal correction technique to measure the movement track. Verify the feasibility of the experimental method by measuring the horizontal movement of the Coordinate Measuring Machine, and compare the DIC value without the horizontal correction. Finally, the vibration mode of the industrial motor is detected by the DIC method. And use the DIC method combined with vibration amplification technology to let the naked eye see the tiny vibration of the motor. Compared with the commercially digital image correlation method suite, the algorithm technology of this paper has a high flexibility application and can perform multi-scale measurement experiments.
https://hdl.handle.net/11296/nmdp7d

Vibration Measurement of Induction Motor and the Development of Bionic Eardrum by using Piezoelectric Film

keywords: 3D printing, induction motor, FFT, STFT, PVDF

 The isotropic cantilever plates as the PVDF (polyvinylidene fluoride) piezoelectric film sensor were analytically, numerically, and experimentally investigated. The theoretical analysis of the cantilever plate and finite element method simulation were verified by plate tapping experiment. Experimental vibration signal of the motor analyses were carried out using the Fourier algorithm to extract the frequency response functions and to search for the feature. Besides, we applied 3D scanning technology, 3D printing technology, and PVDF film sensor to make bionic eardrum. The PVDF is internally excited and the vibration signal is converted by the Fourier algorithm to verify the signal receiving sensing capability. The PVDF is internally excited and the vibration signal is converted by the Fourier algorithm to verify the signal receiving sensing capability. 

https://hdl.handle.net/11296/9c5x3x

Measurement and Analysis Systems for Characteristic Frequency and Surface Roughness

Keywords: point cloud, KEYENCE, MATLAB, surface roughness, 3D measurements, sweep system, vibration, theory of plate

 In this paper,we have developed automatic measurement systems for increasing convenience and efficiency of the experiment. There is divided into two parts in the paper﹐first part is analyzing vibration of cantilever plates by theoretical analysis﹑finite element method and impact experiment. Finally﹐we measured characteristic frequency by using self-developed automatic sweep system. That is goal to verify function and result of automatic sweep system. This system can automatic adjusts signal frequency﹑remote controls instruments and keeps away from the noise of the experimental environment. It can also returns and saves data of sensor to user computer. Second part, we developed a MATLAB program, surface rough analyzer (SRA), for retrieving raw data from KEYENCE VR3100. The SRA provides command mode for quantitatively selecting region of interest, and gives high flexibility for a batch of measurement of magnetic connectors. The in-plane resolution and depth resolution is 1 and 0.1 , respectively. In comparison of conventional morphology, the SRA simultaneously gets RGB and depth data from 3D measurement and presents isometric viewpoint of point clouds. The program is able to export XYZ data for Solidworks, and directly compare the mechanical error of 3D model created in computer design and those obtained from practical manufacture. 

https://hdl.handle.net/11296/8szgte