| With the wide application of wireless sensor networks,the sustainable power supply and battery life of sensor network nodes have gradually become the fundamental problems restricting their development.The flow-induced vibration piezoelectric energy harvesting device can collect energy from the environment where the sensor is located,so as to realize the sustainable self power supply of the sensor equipment.However,the traditional flow-induced vibration piezoelectric energy harvesting device has the disadvantages of narrow effective working bandwidth and low output power,which limits its development and application.Based on the above background,this paper proposes to introduce the downstream interference rectangular plate or introduce the upstream ash deposition bluff body to improve the performance of the flow-induced vibration piezoelectric energy harvester.The vibration characteristics and energy harvesting performance of the energy harvester under different working conditions are studied through wind tunnel experiments,and the optimal design of energy harvesting is determined.In addition,considering that the traditional wind tunnel experiment and numerical simulation research methods are time-consuming,labor-intensive and high cost,this paper proposes to use machine learning technology to predict the voltage output and displacement response of tandem double bluff body wake galloping piezoelectric energy harvesting.Three machine learning algorithms,decision regression tree(DTR),random forest(RF)and gradient boosting regression tree(GBRT),are used to train the model,and the prediction performance of different machine learning models is compared.The main conclusions of this paper are as follows:(1)In the flow-induced vibration piezoelectric energy harvesting system under the influence of tandem rectangular plates,changing the height of the rectangular plate and the spacing between the upstream and downstream bluff bodies can significantly affect the vibration response mode and performance output of the energy harvester.For a specific height of the downstream interference rectangular plate,galloping-type response(i.e.,full interference between galloping and vortex-induced vibration),partial interference between galloping and vortex-induced vibration,and vortexinduced vibration-type response are observed successively with increasing the spacing between the cylinder and the plate.The optimal design scheme is to place a 2 D high interference rectangular plate at 0.2 D ～ 0.4 D downstream of the cylinder.This scheme significantly expands the effective wind speed bandwidth of the traditional vortexinduced vibration energy harvester,and the maximum voltage output is increased by more than 102%.(2)In the wake galloping piezoelectric energy harvesting system under the influence of tandem ash deposition bluff body,the upstream bell-shaped ash deposition bluff body is more conducive to the energy harvesting of wake galloping piezoelectric energy harvesting than the horn-shaped ash deposition bluff body.When the spacing ratio L/D = 1.5,the energy harvesting performance is the optimal.Compared with the traditional vortex-induced vibration energy harvester,the effective wind speed bandwidth is significantly expanded,the threshold wind speed is reduced by 12%,and the maximum voltage output is increased by more than 111%.In addition,both types of ash deposition bluff bodies will lead to the quenching phenomenon of piezoelectric energy harvesting in the region with spacing ratio L/D = 2.5 ～ 3.0,which greatly reduces the performance of piezoelectric energy harvesting.Therefore,it is necessary to pay attention to and avoid energy harvesting in this region.(3)Using machine learning technology,the voltage output and displacement response of the tandem double bluff body wake galloping piezoelectric energy harvester can be accurately predicted.The research shows that the four input parameters: the cross-sectional shape of the upstream bluff body S*,the diameter ratioη,the spacing ratio L* and the reduced wind speed U* have important effects on the performance output of the wake galloping piezoelectric energy harvester.Among the three machine learning models,the GBRT model showed the optimal prediction performance,and could accurately predict the voltage and displacement of the energy harvester within the variation range of the four studied parameters S*,η,L* and U*.Machine learning techniques can provide a fast and accurate evaluation of the design of flow-induced vibration piezoelectric energy harvesters.It can be used as a supplement to the traditional wind tunnel experiment and numerical simulation methods. |
