Development Of Internet-Based Sensor-Network Node And Related Signal Detection Technique

In modern industrial production, unexpected equipment-failures may lead to the "chain reaction", resulting in the paralysis of the whole production system, causing huge economic losses and disastrous accidents. And the importance of equipment-condition monitoring and fault diagnosis technology is being generally recognized nowadays.For the theme of equipment-condition monitoring and fault diagnosis system development, this master thesis mostly focuses on the network sensor-node design for the data collection in the front-end, and the mechanical-failure signal processing technique in the back-end, both with in-depth analysis and research. In data-collection field, as concerns on the cost, technical implementation and maintenance, a new widely-accepted standard, to solve the interface between the sensors and network connectivity, becomes more and more urgent nowadays. Based on that, the Ethernet and embedded technology are applied to the sensor-node design in this paper, which has greatly utilized the cost advantages of the Internet for industrial monitoring systems. The customized TCP / IP protocol stack is specifically discussed. And finally a low-cost network sensor-node prototype for remote monitoring by Internet is developed, and also a remote client application program is provided for communications test with the sensor-node. This work paves the way for building large-scale Internet-based remote monitoring system.In the back end, signal processing for mechanical fault diagnosis is critical for the ultimate purpose. How to extract the fault information effectively has become one of the key issues in this field. A novel signal processing method, integrated with time-domain synchronous average and the continuous wavelet transform, is studied for detecting cyclical impact of high-frequency oscillations ---- Synchronous average continuous wavelet transform array method. The result of original signal's continuous wavelet transform is viewed as a group of time-domain signals corresponding to different frequency bands, and the signals array is processed by synchronous average scanning, after that, each energy distribution of signals will be depicted in the form of two-dimensional gray level image. The period of the impulse train and the frequency of the oscillation can be further estimated from the location of the peak-launched spot in that image. The method has achieved good results for weak fault signal detection.