Research On Key Technology Of Wireless Sensor Network For Medium Or Large Scale Distributed Testing System

With the rapid development of equipments in recent years,such as vehicle,aircraft and ship,a large number of distributed tests are required for equipment in all stages of research and development,setting,acceptance,and use.With the increasing complexity of test projects,the types and quantities of parameters are exponentially increased under different conditions and loads.Traditional test methods are difficult to meet distributed testing needs because of the difficulties in wiring with poor flexibility.Wireless sensor networks are suitable for large-scale distributed testing because of some excellent features,such as flexible deployment,expandability and resistance to damage.However,currently,wireless sensor networks still have some problems: the limitation of network scalability and damage resistance by synchronization model and algorithm,and the limitation of the network's ability to handle large data volumes.In view of the above limitation,this paper carries out relevant theoretical analysis from the multi-scale firefly-inspired synchronization model,algorithm and verification technology,hardware acceleration processing and efficient transmission technology,data compression technology and packet loss repair technology.The specific research works are as follows:Aiming at the problem that the traditional master-slave synchronization model has a strong dependence on the topology structure which leads to a poor scalability and antidestruction ability,while the traditional firefly-inspired synchronization model can be implemented poorly,a mathematical model based on multi-scale firefly-inspired synchronization is proposed.Inheriting the advantages of the traditional firefly synchronization model,the phase in traditional integration and coupling model is discretized into a single-scale integration and coupling model firstly,which is suitable for hardware platform implementation.Then,the single-scale phase model is converted to a multi-scale phase model.The network environment model,communication model,and node model,that may affect the synchronization performance,are analyzed.Finally,the comparison simulation verifies the validity and advancement of this model that it has better engineering practicality,synchronization speed and accuracy than the traditional firefly synchronization model.While the traditional master-slave synchronization algorithm has poor scalability and anti-destruction ability,the traditional distributed synchronization algorithm has a slow convergence,a poor stability and the terrible channel congestion.Thus,based on the proposed multi-scale firefly synchronization mathematical model,a multi-scale fireflyinspired synchronization algorithm is proposed.In order to achieve a fast and stable synchronization,five simultaneous tasks are designed: Self-increasing the state vector,sending the synchronization packet at a random moment,delay and frequency drift compensation,synchronization packet processing and reachback state vector multi-scale adjustment.This algorithm has advantages,such as easy expansion and anti-destruction,as the traditional distributed synchronization algorithm,while improving the synchronization convergence speed and accuracy,and reducing network channel congestion.A hardware verification platform is established based on the hardware and software design,and the proposed firefly synchronous mathematical model and algorithm are implemented using this hardware verification platform in the laboratory environment for proving the validity of our model and algorithm.In order to remove the redundancy of data in wireless sensor testing system,a signal decomposition compression algorithm based on compressed sensing was proposed.The compression algorithm decomposes the original signal in accordance with the sparseness in the same or different dictionaries.Then,encode the decomposed signal by projection on a Bernoulli matrix,and generates a dictionary mask during the compression process.The compressed signal and mask are sent back to the terminal.The terminal restores the data step by step according to the encoding information,the sparse dictionary,and the dictionary mask.This data compression method has stronger robustness than the traditional compression techniques,and is insensitive to packet loss.Compared with the original compressed sensing algorithm,it can greatly reduce the amount of encoded data and improve the real-time information acquisition.Furthermore,the encoded data packets have priority attributes,which is easy for network management.Aiming at the problem that the traditional transmission strategies and protocol stacks of wireless sensor testing system are weak when dealing with sudden large-amount data transmission tasks,a multi-channel pipelined transmission strategy was proposed for engineering practicality,and the internal data stream was designed.The hardware accelerated processing design realizes the parallel processing,backup and transmission of high-speed data streams.Utilizing the heterogeneity of network control and data stream,a dual protocol stack for processing control and data stream separately was designed.Thus,the control flow is stable and reliable while the data flow is efficiently transmitted at the same time.Aiming at the packet loss problem caused by the transmission process in WSN,using the time correlation of the data and the ability of the convolutional neural network to understand the signal prior,an encoder-decoder architecture deep learning neural network algorithm is proposed to repair one-dimensional damaged signal.This algorithm performs data repair processing on one-dimensional signals without any prior knowledge of the signal.By preparing data sets,introducing window functions and regular penalty terms in loss function,defining two iterative stopping conditions,the stability and convergence of the algorithm are enhanced and the generalization ability of the network is improved.Comparison experiments show that this algorithm has more excellent data recovery performance than traditional interpolation and compression sensing algorithms.This paper analyzes and studies the key technologies in wireless sensor network for medium or large scale distributed testing systems.From the perspective of theory and engineering practicality,a multi-scale firefly synchronization model,algorithm and a corresponding hardware verification platform are proposed.The corresponding strategies and algorithms are developed for data preprocessing,transmission and post processing in large data volume transmission technology.Through the research of this paper,it has important theoretical significance and application value to improve the test performance of wireless sensor network test system.