Sensor Fault Detection and Signal Compression for Underwater Surveillance Systems

Abstract

With the development of modern science, the importance of maritime territorial defense is emphasized due to interest in marine resources and the increase in maritime trade between countries, and the interests of countries to strengthen their dominance in the sea are sharply opposed. Under these circumstances, underwater acoustic engineering plays a major role in various fields such as marine exploration and underwater surveillance. In particular, research on sonar technology for underwater surveillance has been ongoing since World War II. Sonar refers to a technology or equipment that uses underwater sound to determine the existence of an underwater object and find out its location and characteristics. Recently, military sonar technology using underwater sensors and commercial and It is being actively developed in the field of scientific research. Sonar, which is generally used for military purposes, is mainly used for detection, location calculation and tracking, identification of targets such as submarines, torpedoes, and mines, or for underwater communication. In academic or commercial fields, it is mainly used for fish detection and underwater exploration. Sonar can be classified into two types, active sonar and passive sonar, according to the operation method. Both types use acoustic sensors called projectors and hydrophones, respectively, for the underwater sound source and receiver. In most active sonar systems, the same transducer can be used as a projector and hydrophone. On the other hand, passive sonar systems for search, surveillance, or monitoring of various noises use only hydrophones. Active sonar is a method that transmits an acoustic signal from the projector and detects the echo sound reflected from the target, and passive sonar is a method of detecting and receiving noise radiated from the target, i.e., mechanical noise, flow noise, and propeller noise. An underwater acoustic sensor used in an active sonar and a passive sonar detects a pressure change in an acoustic signal and underwater noise, and generates a sensor signal corresponding to an output voltage proportional to the pressure. In order to detect an underwater target signal that is weak compared to the surrounding noise, an array sensor composed of multiple sensors rather than a single sensor is mainly used to minimize the influence of the background noise and to estimate the position of the underwater target. Such an array sensor operates a sensor array such as a linear, circular, cylindrical, or planar type composed of tens or hundreds of channels. Through this, an array signal processing technique for improving the signal-to-noise ratio or predicting the direction of arrival of a target signal source is applied using multi-channel sensor signals input simultaneously. When designing a sonar system, it is important to design the sensor array shape and use a acoustic sensor with high sensitivity along with the determination of the operating frequency band and the number of sensor channels. The technologies that need to be applied in operating such a sonar system are as follows. First, it is a fault detection technique for the acoustic sensors constituting the array. Due to the characteristics of a system that needs to detect a weak signal using multiple sensors, if a fault or failure of the sensors constituting the array occurs, target detection performance may be degraded or false alarms may occur due to the generation of a distorted signal. In particular, if there is a fault in the sensor of the sonar system for real-time monitoring, it may cause confusion and tactical damage in the occurrence of false alarms. Therefore, it is essential to secure the reliability of the signal output from the sensor, and a study to detect a sensor fault that can determine the fault of the sensor is required. Second, it is a compression technique for underwater acoustic sensor signals. In the case of the sensor signal of the passive sonar system, the required sampling rate is low because it is mainly designed to detect the noise of the ship machinery existing within a few kHz, but it is necessary to process data of several tens or hundreds of channels. In addition, in the case of a sensor signal of a high-frequency active sonar system excluding baseband shift, a sampling rate of 100 kHz or higher is required, so the amount of data of the sensor signal to be processed is large. In addition, in the case of a specific platform-based sonar system, it is necessary to transmit/receive sensor signals using wireless communication in a marine environment or to operate in a hardware environment with limited storage capacity and computational amount. To this end, it is necessary to apply a compression technique to the sensor signal. In particular, in order to detect a weak target signal, it is necessary to apply a compression technique in which there is no signal distortion or loss of information or is minimized. In this dissertation, we propose a sensor fault detection method and sensor signal compression method that can determine the presence or absence of a sensor fault required to operate an actual sonar system. First, the proposed sensor fault detection technique measures the root-mean-square value of the acoustic signal for each channel input through the sensor, then determines whether there is a failure through the threshold value analysis method, and performs the secondary determination by analyzing the value of root-mean-square crossing rate. composed of structures. In order to verify the performance of the proposed method, we used an array sensor signal that is actually operating off the coast of Korea, and it was observed that the proposed method showed excellent fault detection performance. Next, we propose two compression techniques for underwater acoustic sensor signals. The first technique is an MPEG-4 ALS-based lossless compression technique. The proposed technique is combined with a sensor defect detection technique. After determining whether the multi-channel sensor input signal is normal or defective, if it is determined that the sensor has a defect, the corresponding signal is excluded from compression before signal compression, or coding efficiency can be improved through pre-processing. Then, the multi-channel sensor signal is encoded by an MPEG-4 audio lossless coding encoder, where the index of the defective sensor is also sent to the decoder, which applies a post-processing technique to the lossless decoded signal by the MPEG-4 ALS decoder. apply For this technique, performance evaluation was also performed using the array sensor signals in actual operation. As a result of the evaluation, it was confirmed that the defect detection of the proposed method worked correctly for the experiment, and the compression performance was further improved in the defect sensor signal transmission and non-transmission mode compared to MPEG-4 ALS. The second technique is a subband analysis-based near-lossless sensor signal compression technique. The proposed compression method further improves the data compression rate and minimizes power consumption for applying the compression technique. The proposed compression method divides the input signal into a low-frequency band and a high-frequency band in order to increase the data compression rate and minimize power consumption, and each subband signal is encoded independently based on the analysis of the characteristics of underwater noise for each subband. In the design of the proposed method, the concepts of subband division and linear prediction coding based on Quadrature Mirror Filter (QMF) were adopted, and an entropy coding technique suitable for underwater sensor signals was proposed. Through the performance experiment, the proposed method showed a higher compression ratio than the methods used as the comparative method. In terms of processing time, the performance was similar to that of the comparison techniques, and it was observed that the signal distortion was negligible.