http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Optimization of a piezoelectric wind energy harvester with a stepped beam
Jiantao Zhang,Dong Qu,Zhou Fang,Chang Shu 대한기계학회 2020 JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY Vol.34 No.11
A galloping-based piezoelectric energy harvester using the stepped cantilever beam is proposed and investigated. Transverse galloping is induced with the square cross sectioned bluff body. When the wind speed exceeds the critical wind speed, the self-excited oscillation of the harvester occurs and more output power is generated. To obtain the optimal design of the energy harvester, the sequential quadratic programming (SQP) and the evolution strategy (ES) are employed to determine the optimal solution. The finite element method is used to calculate the output voltage of the harvester. After optimization, the output voltage of the optimal harvester is significantly improved in comparison with that of the initial one. Two prototype harvesters based on the initial and optimal dimensions were fabricated and measured experimentally. An open-circuit rms voltage of 36 V and an output power of 0.52 mW were obtained at the wind speed of 14 m/s for the optimal harvester. They are about 8.3 times and 4.73 times of that of the initial harvester. The validity of the optimal design is verified with the experimental results.
Qibin Wang,Bo Zhao,Hongbo Ma,Jiantao Chang,Gang Mao 대한기계학회 2019 JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY Vol.33 No.6
Real-time monitoring and rapid evaluation of bearing operating conditions, especially for the reliability evaluation and remaining useful life (RUL) prediction, are major challenges in the rotating machinery field. A two-dimensional (2-D) deep convolution neural network (CNN) is proposed for rapidly evaluating reliability and predicting RUL, in which a signal conversion method is proposed for converting the one-dimensional signal into the 2-D image to satisfy the input requirements of the 2-D CNN. Different activation functions are employed to implement the conversion of the input data to the output data for each layer of the network, and dropout is only adopted in the hidden layer to change the network structure to prevent overfitting. The maximum correlation entropy with regular terms is employed as the loss function of the model to obtain better training performance compared with the mean square error (MSE). Then, the rolling bearing degradation vibration data is applied to the proposed model to verify the accuracy and rapidity. The results show that the proposed method has good accuracy and fast calculation ability in bearing reliability evaluation and RUL prediction, especially, its time consumption is shorter than that by other deep learning networks.