Improvement Of Data Aggregation Of Wireless Sensor Networks Using Fuzzy Neural Networks

A wireless sensor network consists of a large number of sensors which deployed over a geographical area for monitoring physical phenomena. Sensors can collect information in their coverage area and sent it to staff computer. However, each sensor has a limited energy supply and they are often deployed in remote areas or even areas without accessibility, which makes the replacement battery for additional power become unrealistic. Moreover, due to the aging of their internal devices or by the influence from outside, noise is recognized as an essential issue. A good data fusion method can greatly reduce the impact of noise on fusion results. Data fusion aims at reducing the network redundancy data, and the network data traffic, saving node energy, improving information collection for the purpose of accuracy. It is one of the important research areas on wireless sensor network.To save energy for wireless sensor networks (WSNs), NBPNA, a new data aggregation algorithm based on back-propagation networks was proposed, which integrates a three-layer BP neural network with clustering routing protocol. We use it for data fusion in WSNs, and then send the weight and threshold rather than the raw data monitored from sensors to the sink, at the same time, using the weight and threshold in the last fitting as the input of the new fitting, the number of Neural Network training steps can be reduced greatly. Simulation results show that the proposed algorithm can be effectively reduce data transmissions, so as to achieve energy efficiency in WSNs, and the lifetime of the network is prolonged. At the same time, this algorithm is also verified the feasibility and effectiveness of environmental monitoring, etc.Last, a new data aggregation algorithm based on fuzzy neural networks was proposed, which integrates a Takagi-Sugeno fuzzy neural network with clustering routing protocol. We use it for data fusion in WSN, and then using the Network training parameters and the parameters of membership functions in the last fitting as the input of the new fitting, the number of Neural Network training steps can be reduced greatly. Simulation results show that the proposed algorithm can be effectively reduce data transmissions, so as to achieve energy efficiency in WSN, and the lifetime of the network is prolonged. At the same time, this algorithm is also verified the feasibility and effectiveness of environmental monitoring, etc.