Research On The Remote Elevator Fault Diagnosis System Based On GPRS

In this paper the remote elevator fault diagnosis system based on GPRS (General Packet Radio Service) is thoroughly researched.The structure of the remote elevator diagnosis system is studied first. By designing the system structure reasonably, the remote maintaining service center can collect the data of elevators, which are in a wide area, in real time through Internet. Then the data can be processed in the center. After lucubrated the principle of GPRS used in remote data transmission, embedded elevator data collection board is deigned. Based on it along with the RTOS(Real Time Operation System)- μC/OS-II, thoroughly discusses how to establish a development platform of embedded software. And embedded TCP/IP protocol is proposed based on the analysis of the standard TCP/IP protocol. Then discusses the realization of embedded GPRS protocol. Discusses how to use the platform to design application software, so the data collection board can connect Internet thorough GPRS to transmit the elevator data. When the elevator is in fault state, the data collection board can realize many kinds of alarm mode, such as making a phone call, sending out a short message or playing voice record to the maintaining people in time. Practice proves that the development of embedded system based on embedded software development platform can increase the development efficiency, reduce the repetition labor and increases the system's reliability, stability and real time character widely.Thoroughly researches the principle of Fuzzy Petri Net (FPN) and Neural Network (NN) that are used in fault diagnosis. Based on the deep research of their advantage and disadvantage, a NNFPN (FPN based on NN) model is proposed.. NNFPN synthesizes the advantage of both systems and can be used in fault diagnosis system. It provides a transparent modeling and analyzing capability and is not any more a "black box". NNFPN representing a fuzzy fault diagnosis expert system based on knowledge can be used to analyze the different inference states step-by-step and the meaning is clear and definite. At the same time because of the learning unit of NNFPN is not very big, the computation speed is obviously quicker than that of complex NN. More important, NNFPN has learning ability, that is to say, the parameters including membership values, weights and certainty factors etc. of fuzzy production rules, which are modeled by NNFPN, could be tuned. So the NNFPN model is no longer fixed and can learn new expert experience knowledge constantly. The learning algorithm of NNFPN is proposed. And give a practical example to demonstrate the correctness and validity of the algorithm. Discusses how to establish the NNFPN model of the elevator fault diagnosis system and how to use the model to do fault diagnosis reasoning. At last proposes that to elevators, which are in different condition, should use different NNFPN model to do fault diagnosis. Thus the reasoning result could be more accurate and even mach actual circumstance.