As tunnel construction proceeds ever more rapidly, the efficiency of seepage detection by engineers with expert knowledge is facing unprecedented challenges. Moreover, it suffers from strong subjectivity. In recent years, deep learning, as an algorithm of machine learning, has achieved state-of-the-art performance in pattern recognition. In this paper, we address such a problem by building convolutional neural networks that operate on conventional graphics processing units. Within the project, the data is obtained by an infrared thermal imager since there exist different characteristics of temperature between the area of seepage and non-seepage. Considering the difficulty of collecting many images, generative adversarial nets and other data augmentation skills are applicable to enlarge data sets. We design several novel architectures where the attention mechanism is plugged into various traditional models, considered as VGG16 network with Attention Module and RestNet34 with Attention Module, and the overall identification accuracy achieved is more than 97%. The codes of this project can be found at https://github.com/Scotter-Qian/cnn .

This paper presents the design and optimization of a novel MEMS tuning fork gyroscope microstructure. In order to improve the mechanical sensitivity of the gyroscope, much research has been carried out in areas such as mode matching, improving the quality factor, etc. This paper focuses on the analysis of mode shape, and effectively optimizes the decoupling structure and size of the gyroscope. In terms of structural design, the vibration performance of the proposed structure was compared with other typical structures. It was found that slotting in the middle of the base improved the transmission efficiency of Coriolis vibration, and opening arc slots between the tines reduced the working modal order and frequency. In terms of size optimization, the Taguchi method was used to optimize the relevant feature sizes of the gyroscope. Compared with the initial structure, the transmission efficiency of Coriolis vibration of the optimized gyroscope was improved by about 18%, and the working modal frequency was reduced by about 2.7 kHz. Improvement of these two indicators will further improve the mechanical sensitivity of the gyroscope.

No abstract

No abstract

The camera resolution and electronic noise limit the accuracy of the projection speckle three-dimensional digital image correlation (3D-DIC) which is a non-invasive method to detect the reliability of electronic packaging structure. In this study, a measurement method of super-resolution (SR) reconstruction coupled with projection speckle DIC is proposed. The algorithm based on the maximum a posteriori (MAP) model for DIC measurement systems was also optimized, and a speckle-specific bimodal prior was proposed to adapt to speckle images. By using optimized SR technology as an image pre-processing technique to enhance the resolution of captured images, the accuracy of measurements is improved. Full-field displacement measurement experiments show that, with suitable magnification and speckle size, the use of SR technology reduces the range of displacement errors from 8 μm to 2 μm. Experiments on step block topography measurements show that the use of SR technology reduces the error between DIC measurements and Moiré interferometry from 5 μm to within 2 μm. Therefore, SR technology can be effectively paired with projection speckle DIC measurements to adapt to various measurement scenarios in the field of electronic packaging reliability testing.