Design And Implementation Of Wireless Sensor Network Based On Vibration Energy Harvesting

It is vital to improve the safety index of metro train operation by taking advantage of wireless sensor network to monitor the running state of wheel bearings in real time,however,wireless sensor nodes are powered by external battery,whose energy is limited.This paper can solve the energy limitation problem of wirelsess sensor nodes by utilizing vibration energy harvesting technology to convert mechanical energy occurring on wheel bearings into electrical energy,however,the power supply mode of vibration energy harvesting is full of uncertainty,there is a mismatch between the energy supply of vibration energy harvesting and the energy consumption of wireless sensor nodes.To ensure the reliable transmission of vibration acceleration data,the data transmission scheme of wireless sensor network must be coordinated with the instability of power supply source and the dynamic characteristics of network topology.The difficulty and innovation of this paper lie in how to design the fault diagnosis model of wheel bearings to capture the temporal correlation and spatial correlation of vibration acceleration data.Aim at the above problems,combined with the real industrial scene,this paper describes the design and implementation of wireless sensor network based on vibration energy harvesting,as shown in the following three aspects.Firstly,the vibration energy harvesting scheme is designed and implemented.The vibration acceleration data is collected on the wheel bearings of metro trains and its vibration index is analyzed,which is used as the selection standard of vibration energy conversion element and accelerometer;design wireless sensor node and optimize its power consumption;test the energy conversion efficiency of the vibration energy conversion element under the experimental platform which restores the vibration scene;aiming at piezoelectric energy harvesting and electromagnetic energy harvesting respectively,design the energy management strategy of harvesting without consumption and harvesting with consumption,and design the corresponding energy management circuit.Secondly,the data transmission scheme of wireless sensor network is designed and implemented.The routing algorithm decomposes the long-span and long-line network topology into cluster network,which limits the randomness of these situations that wireless sensor nodes join the network when they get energy and wireless sensor nodes leave the network when they lose energy within the cluster network;to ensure the realtime ability and reliability of data transmission,the MAC protocol adopts the centralized scheduling control protocol TDMA,furthermore,this paper designs the assignment of superframe time slot,time synchronization scheme,transmitting and receiving slot timing scheme and network packet;taking electromagnetic energy harvesting as the power supply source of wireless sensor nodes,this paper designs the network operation process.Thirdly,the fault diagnosis method applied to wheel bearings of metro trains is designed and implemented.To ensure that the proposed model can be migrated to the real industrial scene,this paper collects the vibration acceleration data on wheel bearings of metro trains and process the data by the way of data augmentation.The proposed fault diagnosis model Metro Net consists of encoder module and decoder module,where encoder module can capture the temporal correlation and spatial correlation of vibration acceleration data.This paper tests and analyzes the feasibility of the above three aspects under the experimental platform.In the electromagnetic energy harvesting system,the time required for wireless sensor nodes to power on is short,and the wireless sensor nodes can work continuously;the data transmission system of wireless sensor network can effectively deal with the uncertainty of energy supply that wireless sensor nodes get,the instability of vibration energy harvesting and clock drift caused by crystal error;the classification accuracy of Metro Net model is up to 97.20% and AUC index is up to 0.9951,which can be deployed in real industrial scenarios.