Sensor networking node abnormity compensation and correction method

Abstract

The invention relates to a sensor network node abnormity compensation and correction method, which comprises the following steps of: 1) arranging a plurality of sensors on the surface of a detection structure to form a sensor network, dividing a detected area into a plurality of grids, and detecting each grid through four sensors; 2) judging a grid where an abnormal sound source is located through energy; 3) judging the approximate range of the signal, and if the judgment cannot be carried out, proving that the signal-to-noise ratio of the signal is too poor and the sensor is abnormal; 4) establishing a TDOA curve according to signals acquired by the four sensors in the quadrant when the four sensors are normal, and taking an average value of all intersection points as a final positioning result; if the sensors are abnormal, similar sensors are selected to form a positioning range until the four sensors are normal, and a TDOA curve is established. The abnormal sensors can be eliminated by analyzing the signals of the sensors, the method can be accurately applied to various abnormal sound sources capable of being monitored under sensor networking, and the method has important significance on structural health monitoring under different working conditions.

Description

Technical Field

[0001] This invention belongs to the field of sensor network integrity detection and structural health monitoring technology, specifically relating to a compensation and correction method for abnormal sensor network nodes. Background Technology

[0002] In daily production, damage is inevitable in large mechanical structures. These losses may be caused by impacts from foreign objects, or the damage may further generate abnormal sound sources. Timely and accurate location of abnormal sound sources is of great importance. Due to its advantages such as high system integration, high sensitivity, convenient device, and fast detection speed, acoustic emission method is often used to locate abnormal sound sources in actual working conditions, such as high-pressure pipeline and storage tank leak detection, spacecraft launch and on-orbit monitoring, and bridge and building structure monitoring.

[0003] For some large or complex structures, a single-digit number of sensors is insufficient to provide adequate positioning resolution. Therefore, a sensor network needs to be deployed on the surface of the structure being measured. By identifying the grid within the network where the abnormal sound source is located, the sensor array within that grid is used to locate the sound source, achieving accurate and real-time monitoring of abnormal sound sources. However, increasing the number of sensors also introduces the risk of sensor malfunctions. Therefore, timely and effective compensation and correction are required for abnormal nodes in the sensor network. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a compensation and correction method for abnormal sensor network nodes. By finding abnormal nodes in the sensor network and combining them with the network configuration, the method can still collect and locate abnormal signals even when nodes are damaged.

[0005] The technical problem solved by this invention is achieved through the following technical solution:

[0006] A method for compensating and correcting anomalies in sensor network nodes, the method comprising the following steps:

[0007] S1. Measurement System Setup: The sensor network node anomaly compensation and correction system consists of a measurement object, an abnormal sound source excitation device, a preamplifier, acoustic emission sensors, a power supply signal splitter, a signal acquisition card, and a computer. The abnormal sound source excitation device acts on the measurement object to simulate the sound signal of a foreign object impacting in real working conditions. The acoustic emission sensors are attached to the surface of the measured object with a coupling agent to achieve sensor network for signal capture and omnidirectional detection of the measured structure. The acoustic emission sensors amplify the acoustic emission signals through the preamplifier. The preamplifier is connected to the power supply signal splitter for independent power supply. The power supply signal splitter is connected to the signal acquisition card, which acquires the amplified signals. The signal acquisition card is connected to the computer, which processes and calculates the signals acquired by the signal acquisition card and locates the acoustic emission source. The acoustic emission sensors provide comprehensive and uniform coverage of the measured structure, ensuring that when an abnormal sound source occurs at any location in the measured structure, the sensors can accurately detect it and ensure that some sensor signals have a good signal-to-noise ratio.

[0008] S2. Sensor Grid Selection: Number all acoustic emission sensors and group the n sensors into m grid systems. Each grid consists of four sensors, and the four sensors in each grid form a distributed sensing network. Calculate the energy of all sensor signals for the same length. Let the energy of sensor i be E. i (i = 1, 2, 3…n), the energy set is:

[0009] E = {E1, E2, ..., E} n};

[0010] Because the energy of acoustic emission signals attenuates significantly after propagating over long distances within the structure, the four sensors with the highest energy were selected: max 4 {E i If the four sensors can form a square distributed sensing grid, the sound source can be located and analyzed.

[0011] S3. Eliminate abnormal sensors: The following abnormalities may exist in the sensor: (a) abnormal sensor connection, no signal; (b) poor contact between the sensor and the measured surface, weak signal; (c) signal bias; (d) sensor itself has a problem, weak signal; abnormality (c) can be corrected by filtering, while other abnormalities are judged by adaptive threshold method.

[0012] S4. Analyze the possible damage to the faulty sensor and handle it according to the specific situation:

[0013] S5. Arrival Time Calculation: Calculate the specific arrival time of the sensor signal in mode S0. Use STFT to perform short-time windowing on the signal, dividing the non-stationary signal into multiple approximately stationary short time intervals. Then perform Fourier transform on each time interval to obtain the spectrum changes over time. Set a signal judgment threshold by evaluating the noise level, retain peaks with peak values ​​greater than the threshold, and take the peak with the earliest arrival time as the signal arrival time value.

[0014] S6. TDOA Calculation: The TDOA method uses distributed sensors to obtain the arrival times t1, t2, and t3 of signals at each location. Based on the wave velocity, the difference in distance from different sensors to the sound source can be calculated.

[0015]

[0016] In the formula: Δd i,j Let Δt be the difference in distance from sensor i and sensor j to the sound source. i,j Let x be the time difference between the arrival of signals from sensors i and j, v be the signal wave velocity, and (x) be the signal wave velocity. i ,y i (x) represents the position of sensor i, and (x) represents the position of sensor i. s ,y s Let (x, y) be the location of the sound source. The set of all points (x, y) that satisfy the distance relationship can form a hyperbola:

[0017]

[0018] By plotting hyperbolas determined by multiple sensors, the intersections between different hyperbolas are calculated, and the average value of all intersections is taken as the final location of the abnormal sound source.

[0019] Furthermore, the specific process for determining other anomalies in S3 using the adaptive threshold method is as follows:

[0020] The energy method is used to locate the sensing grid where the acoustic emission source is located, and anomalies are eliminated from the data of the four sensors in that grid. The sensor signals are then judged using an adaptive threshold method.

[0021] The adaptive thresholding method determines the approximate location of the starting point by calculating the noise level. First, the first 1 / 10 of the signal is taken as the noise segment, and these points are arranged according to size. The average value of the data in the middle 1 / 2 segment after arrangement is taken as the approximate noise level. Assuming the signal length is N, the noise segment length is M = N / 10.

[0022]

[0023] Among them: E noise For noise level, x sorted [k] represents the sorted data, and M represents the length of the noise segment;

[0024] Then, a window of a certain width is established, and the average value of the signal within the window is calculated as the amplitude level of that signal segment. When the amplitude level of a certain window is greater than 30 times the noise level, the starting point of that window is considered to be the approximate starting point of the event.

[0025]

[0026] Where: W is the length of the sliding window, t event If the signals from all four sensors can be used to calculate a reasonable arrival time for the abnormal sound source, the earliest point is selected as the common starting point for all four sensors, and the next 800 points are selected as the signal selection range for subsequent positioning processing. If there is a signal among the four sensors whose starting point cannot be calculated, it indicates that the signal-to-noise ratio of that sensor is too poor, and the sensor is malfunctioning. In this case, nearby sensors need to be selected to supplement it.

[0027] Moreover, S4 specifically refers to:

[0028] (1) If all four sensors are normal, calculate the energy signals of all sensors and obtain the grid X where the signals are located. If the four sensors 11, 12, 19, and 20 in the preferred range X are normal, use the signals of the four sensors and use the TDOA method to locate the abnormal sound source.

[0029] (22) If one sensor is abnormal, calculate the energy signals of all sensors and obtain the grid X where the signal is located. Sensor 20 in the first choice of range X is abnormal. Sensors 11, 12 and 19 are normal. Sensor 20 is discarded. Select the nearest sensor 21 or 28 to form a new positioning area. If the new sensor is normal, perform regular positioning. If the new sensor is still abnormal, continue to select the nearest sensor until the signals of all 4 sensors meet the normal signal-to-noise ratio. Then, use the TDOA method to locate the abnormal sound source.

[0030] (3) If two sensors are abnormal, calculate the energy signals of all sensors to obtain the grid X where the signals are located. If sensors 12 and 20 are abnormal in the preferred range X, select the nearest new sensor according to the direction of the abnormal sensor, discard sensor 20, and select the nearest sensor 21 or 28; discard sensor 12, and select the nearest sensor 4 or 13 to form a new positioning range; if there are still abnormalities, continue to search for new sensors until all four sensors meet the normal signal-to-noise ratio, and use the TDOA method to locate the abnormal sound source.

[0031] Since the preferred region is determined by the energy of all sensor signals, it is unlikely that three or four sensors in the preferred region will be abnormal unless all sensors are abnormal.

[0032] (4) In some extreme cases, only 3 sensors remain in the sensor network and are normal. Sensors that exceed two grid lengths are judged to have insufficient signal-to-noise ratio and cannot be used for positioning. Therefore, only sensors 11, 12, and 19 are normal in grid X and its nearby grids. Since TDOA supports three-sensor positioning, the normal three sensors are used to form the positioning area for positioning.

[0033] However, in some cases, all three normal sensors are on the same straight line. For example, if only sensors 11, 12, and 13 are normal, normal TDOA positioning cannot be performed. Instead, the abnormal sound source location can be determined by analyzing the arrival times of the three signals to see which sensor is closer to it. In the example diagram, the arrival time can be determined by STFT to show that the abnormal signal is closer to sensor 11.

[0034] (5) In some extreme cases, sensors that exceed two grid lengths are judged to have insufficient signal-to-noise ratio and cannot be used for positioning. Only two sensors remain in the sensor network, namely sensors 19 and 21, which cannot perform normal TDOA positioning. The only way to determine which sensor the abnormal sound source is closer to is by analyzing the arrival time of the two signals. In the example figure, the arrival time is determined by STFT, which shows that the abnormal signal is closer to sensor 19.

[0035] (6) In some extreme cases, only one sensor remains in the sensor network. The only way to determine whether there is an abnormal sound source is to analyze whether the signal-to-noise ratio is normal. In the example figure, the presence of an abnormal sound source signal can be determined by the adaptive threshold method.

[0036] The advantages and beneficial effects of this invention are as follows:

[0037] This invention can eliminate abnormal sensors by analyzing sensor signals and can be accurately applied to various abnormal sound sources that can be monitored under sensor networking. It is in line with the material application trends in various future engineering projects and has important significance for structural health monitoring under different working conditions. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the sensor network structure of the present invention;

[0039] Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0040] The present invention will be further described in detail below through specific embodiments. The following embodiments are merely descriptive and not limiting, and should not be used to limit the scope of protection of the present invention.

[0041] A method for compensating and correcting anomalies in sensor network nodes, the innovation of which lies in the following steps:

[0042] 1) Measurement System Setup: The sensor network node anomaly compensation and correction system consists of the measured object, an abnormal sound source excitation device, a preamplifier, acoustic emission sensors, a power supply signal splitter, a signal acquisition card, and a computer. The abnormal sound source excitation device acts on the measured object to simulate the sound signal of a foreign object impacting in real-world working conditions. The acoustic emission sensors are attached to the surface of the measured object using a coupling agent to achieve sensor networking, signal capture, and omnidirectional detection of the measured structure. The acoustic emission sensors amplify the acoustic emission signals through a preamplifier. The signal amplifier is connected to the power supply signal splitter for independent power supply. The power supply signal splitter is connected to the signal acquisition card, which acquires the amplified signals. The signal acquisition card is connected to the computer, which processes and calculates the signals acquired by the signal acquisition card and locates the acoustic emission source. The sensors provide comprehensive and evenly distributed coverage of the measured structure, ensuring accurate detection of any abnormal sound source at any location within the measured structure and guaranteeing a good signal-to-noise ratio for some sensor acquisition signals.

[0043] 2) Sensor grid selection: For ease of description, all sensors are numbered as shown in the appendix. Figure 1 Taking the sensor network structure shown as an example, there are 32 sensors, forming 21 grid systems. Each grid consists of four sensors, and the four sensors in each grid constitute a distributed sensing and detection network. Calculate the energy of all sensor signals for the same length, and let the energy of sensor i be E. i (i = 1, 2, 3…n), the energy set is:

[0044] E = {E1, E2, ..., E} n}

[0045] Because the energy of acoustic emission signals attenuates significantly after propagating over long distances within the structure, the four sensors with the highest energy were selected: max 4 {E i |i=1,2,...,n}. If these four sensors can form a square distributed sensing grid, the sound source can be located and analyzed.

[0046] 3) Troubleshooting Abnormal Sensors: Possible sensor abnormalities include: (a) abnormal sensor connection, no signal; (b) poor contact between the sensor and the measured surface, weak signal; (c) signal bias; (d) sensor malfunction, weak signal. Abnormality (c) can be corrected through filtering, while other abnormalities can be identified using an adaptive threshold method. The specific process is as follows:

[0047] The energy method is used to locate the sensing grid where the acoustic emission source is located, and anomalies are eliminated from the data of the four sensors in that grid. The sensor signals are then judged using an adaptive threshold method.

[0048] The adaptive thresholding method determines the approximate location of the starting point by calculating the noise level. First, the first 1 / 10 of the signal is taken as the noise segment, and these points are arranged in order of size. The average value of the data in the middle half of the sorted segment is taken as the approximate noise level. Assuming the signal length is N, the noise segment length is M = N / 10.

[0049]

[0050] Among them: E noise For noise level, x sorted [k] represents the sorted data, and M represents the length of the noise segment.

[0051] Then, a window of a certain width is established and the average value of the signal within the window is calculated as the amplitude level of the signal segment. When the amplitude level of a certain window is greater than 30 times the noise level, the starting point of the window is considered to be the approximate starting point of the event.

[0052]

[0053] Where: W is the length of the sliding window, t event This is the approximate starting point of the abnormal sound source. If the signals from all four sensors can be calculated to have a reasonable arrival time, then the earliest point is selected as the common starting point for all four sensors, and the next 800 points are selected as the signal selection range for subsequent positioning processing. If there is a signal among the four sensors whose starting point cannot be calculated, it proves that the signal-to-noise ratio of that sensor is too poor, and that the sensor is malfunctioning. It is necessary to select nearby sensors to supplement it.

[0054] 4) Analyze the possible damage to the faulty sensor and handle it according to the specific circumstances:

[0055] (a) All four sensors are normal, as shown in the attached document. Figure 1 As shown in (a), after calculating the energy signals of all sensors, the grid X where the signal is located is obtained. If the four sensors 11, 12, 19, and 20 in the preferred range X are normal, the signals of the four sensors are used to locate the abnormal sound source using the TDOA method.

[0056] (b) One sensor is malfunctioning, as shown in the attached document. Figure 1As shown in (b), after calculating the energy signals of all sensors, the grid X where the signal is located is obtained. Sensor 20 in the preferred range X is abnormal, while sensors 11, 12, and 19 are normal. Sensor 20 is discarded, and the nearest sensor 21 or 28 is selected to form a new positioning area. If the new sensor is normal, regular positioning is performed. If the new sensor is still abnormal, the nearest sensor is selected again until all 4 sensor signals meet the normal signal-to-noise ratio. The abnormal sound source is located using the TDOA method.

[0057] (c) Two sensors are malfunctioning, as shown in the attached document. Figure 1 As shown in (c), after calculating the energy signals of all sensors, the grid X containing the signals is obtained. If sensors 12 and 20 are found to be anomalous within the first-choice range X, the nearest new sensor is selected based on the direction of the anomalous sensor. Sensor 20 is discarded, and the nearest sensor 21 or 28 is selected. Sensor 12 is discarded, and the nearest sensor 4 or 13 is selected to form a new positioning range. If anomalous sensors still exist, the search for new sensors continues until all four sensors meet the normal signal-to-noise ratio. The anomalous sound source is then located using the TDOA method.

[0058] Since the preferred region is determined by the energy of all sensor signals, it is unlikely that three or four sensors within the preferred region will malfunction unless all sensors are abnormal. This situation will not be elaborated further.

[0059] (d) In some extreme cases, only 3 sensors remain functioning in the sensor network, as shown in the attached diagram. Figure 1 As shown in (d), sensors exceeding two grid lengths are judged to have insufficient signal-to-noise ratio and cannot be used for positioning. Therefore, only sensors 11, 12, and 19 are normal in grid X and its surrounding grids. Since TDOA supports three-sensor positioning, the positioning area is directly composed of the normal three sensors for positioning.

[0060] However, in some cases, all three normal sensors are on the same straight line. For example, if only sensors 11, 12, and 13 are normal, normal TDOA positioning cannot be performed. Instead, the abnormal sound source location can be determined by analyzing the arrival times of the three signals to see which sensor is closer to it. In the example diagram, the arrival time can be determined by STFT to show that the abnormal signal is closer to sensor 11.

[0061] (e) In certain extreme cases, such as the appendix Figure 1As shown in (e), sensors that exceed two grid lengths are judged to have insufficient signal-to-noise ratio and cannot be used for positioning. Only two sensors remain in the sensor network, namely sensors 19 and 21, which cannot perform normal TDOA positioning. The only way to determine which sensor the abnormal sound source is closer to is by analyzing the arrival time of the two signals. In the example figure, the arrival time is determined by STFT, which shows that the abnormal signal is closer to sensor 19.

[0062] (f) In certain extreme cases, such as the appendix Figure 1 As shown in (f), there is only one sensor left in the sensor network. The only way to determine whether there is an abnormal sound source is to analyze whether the signal-to-noise ratio is normal. In the example figure, the presence of an abnormal sound source signal can be determined by the adaptive threshold method.

[0063] 5) Arrival Time Calculation: Calculate the specific arrival time of the sensor signal in mode S0. The signal is windowed using STFT, dividing the non-stationary signal into multiple approximately stationary short time intervals. A Fourier transform is then performed on each time interval to obtain the spectrum variation over time. A signal threshold is set by evaluating the noise level, retaining peaks with peak values ​​greater than the threshold, and selecting the peak with the earliest arrival time as the signal arrival time value.

[0064] 6) TDOA Calculation: The TDOA method uses distributed sensors to obtain the arrival times t1, t2, and t3 of signals at each location. Based on the wave velocity, the difference in distance from different sensors to the sound source can be calculated.

[0065]

[0066] In the formula: Δd i,j Let Δt be the difference in distance from sensor i and sensor j to the sound source. i,j Let x be the time difference between the arrival of signals from sensors i and j, v be the signal wave velocity, and (x) be the signal wave velocity. i ,y i (x) represents the position of sensor i, and (x) represents the position of sensor i. s ,y s Let (x, y) be the location of the sound source. The set of all points (x, y) that satisfy the distance relationship can form a hyperbola:

[0067]

[0068] By plotting hyperbolas determined by multiple sensors, the intersections between different hyperbolas are calculated, and the average value of all intersections is taken as the final location of the abnormal sound source.

[0069] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method for compensating and correcting anomalies in sensor network nodes, characterized in that: The steps of the method are as follows: S1. Measurement System Setup: The sensor network node anomaly compensation and correction system consists of a measurement object, an abnormal sound source excitation device, a preamplifier, acoustic emission sensors, a power supply signal splitter, a signal acquisition card, and a computer. The abnormal sound source excitation device acts on the measurement object to simulate the sound signal of a foreign object impacting in real working conditions. The acoustic emission sensors are attached to the surface of the measured object with a coupling agent to achieve sensor network for signal capture and omnidirectional detection of the measured structure. The acoustic emission sensors amplify the acoustic emission signals through the preamplifier. The preamplifier is connected to the power supply signal splitter for independent power supply. The power supply signal splitter is connected to the signal acquisition card, which acquires the amplified signals. The signal acquisition card is connected to the computer, which processes and calculates the signals acquired by the signal acquisition card and locates the acoustic emission source. The acoustic emission sensors provide comprehensive and uniform coverage of the measured structure, ensuring that when an abnormal sound source occurs at any location in the measured structure, the sensors can accurately detect it and ensure that some sensor signals have a good signal-to-noise ratio. S2. Sensor Grid Selection: Number all acoustic emission sensors and group the n sensors into m grid systems. Each grid consists of four sensors, and the four sensors in each grid form a distributed sensing network. Calculate the energy of all sensor signals for the same length. Let the energy of sensor i be E. i (i = 1, 2, 3…n), the energy set is: E={E1,E2,...,E n }; Because the energy of acoustic emission signals attenuates significantly after propagating over long distances within the structure, the four sensors with the highest energy were selected: max 4 {E i If the four sensors can form a square distributed sensing grid, the sound source can be located and analyzed. S3. Eliminate abnormal sensors: The following abnormalities may exist in the sensor: (a) abnormal sensor connection, no signal; (b) poor contact between the sensor and the measured surface, weak signal; (c) signal bias; (d) sensor itself has a problem, weak signal; abnormality (c) can be corrected by filtering, while other abnormalities are judged by adaptive threshold method. S4. Analyze the possible damage to the faulty sensor and handle it according to the specific situation: S5. Arrival Time Calculation: Calculate the specific arrival time of the sensor signal in mode S0. Use STFT to perform short-time windowing on the signal, dividing the non-stationary signal into multiple approximately stationary short time intervals. Then perform Fourier transform on each time interval to obtain the spectrum changes over time. Set a signal judgment threshold by evaluating the noise level, retain peaks with peak values ​​greater than the threshold, and take the peak with the earliest arrival time as the signal arrival time value. S6. TDOA Calculation: The TDOA method uses distributed sensors to obtain the arrival times t1, t2, and t3 of signals at each location. Based on the wave velocity, the difference in distance from different sensors to the sound source can be calculated. In the formula: Δd i,j Let Δt be the difference in distance from sensor i and sensor j to the sound source. i,j Let x be the time difference between the arrival of signals from sensors i and j, v be the signal wave velocity, and (x) be the signal wave velocity. i ,y i (x) represents the position of sensor i, and (x) represents the position of sensor i. s ,y s Let (x, y) be the location of the sound source. The set of all points (x, y) that satisfy the distance relationship can form a hyperbola: By plotting hyperbolas determined by multiple sensors, the intersections between different hyperbolas are calculated, and the average value of all intersections is taken as the final location of the abnormal sound source.

2. The compensation and correction method for sensor network node anomalies according to claim 1, characterized in that: The specific process for judging other S3 anomalies using the adaptive threshold method is as follows: The energy method is used to locate the sensing grid where the acoustic emission source is located, and anomalies are eliminated from the data of the four sensors in that grid. The sensor signals are then judged using an adaptive threshold method. The adaptive thresholding method determines the approximate location of the starting point by calculating the noise level. First, the first 1 / 10 of the signal is taken as the noise segment, and these points are arranged according to size. The average value of the data in the middle 1 / 2 segment after arrangement is taken as the approximate noise level. Assuming the signal length is N, the noise segment length is M = N / 10. Among them: E noise For noise level, x sorted [k] represents the sorted data, and M represents the length of the noise segment; Then, a window of a certain width is established, and the average value of the signal within the window is calculated as the amplitude level of that signal segment. When the amplitude level of a certain window is greater than 30 times the noise level, the starting point of that window is considered to be the approximate starting point of the event. Where: W is the length of the sliding window, t event If the signals from all four sensors can be used to calculate a reasonable arrival time for the abnormal sound source, the earliest point is selected as the common starting point for all four sensors, and the next 800 points are selected as the signal selection range for subsequent positioning processing. If there is a signal among the four sensors whose starting point cannot be calculated, it indicates that the signal-to-noise ratio of that sensor is too poor, and the sensor is malfunctioning. In this case, nearby sensors need to be selected to supplement it.

3. The compensation and correction method for sensor network node anomalies according to claim 1, characterized in that: Specifically, S4 is: (1) If all four sensors are normal, calculate the energy signals of all sensors and obtain the grid X where the signals are located. If the four sensors 11, 12, 19, and 20 in the preferred range X are normal, use the signals of the four sensors and use the TDOA method to locate the abnormal sound source. (22) If one sensor is abnormal, calculate the energy signals of all sensors and obtain the grid X where the signal is located. Sensor 20 in the first choice of range X is abnormal. Sensors 11, 12 and 19 are normal. Sensor 20 is discarded. Select the nearest sensor 21 or 28 to form a new positioning area. If the new sensor is normal, perform regular positioning. If the new sensor is still abnormal, continue to select the nearest sensor until the signals of all 4 sensors meet the normal signal-to-noise ratio. Then, use the TDOA method to locate the abnormal sound source. (3) If two sensors are abnormal, calculate the energy signals of all sensors to obtain the grid X where the signals are located. If sensors 12 and 20 are abnormal in the preferred range X, select the nearest new sensor according to the direction of the abnormal sensor, discard sensor 20, and select the nearest sensor 21 or 28; discard sensor 12, and select the nearest sensor 4 or 13 to form a new positioning range; if there are still abnormalities, continue to search for new sensors until all four sensors meet the normal signal-to-noise ratio, and use the TDOA method to locate the abnormal sound source. Since the preferred region is determined by the energy of all sensor signals, it is unlikely that three or four sensors in the preferred region will be abnormal unless all sensors are abnormal. (4) In some extreme cases, only 3 sensors remain in the sensor network and are normal. Sensors that exceed two grid lengths are judged to have insufficient signal-to-noise ratio and cannot be used for positioning. Therefore, only sensors 11, 12, and 19 are normal in grid X and its nearby grids. Since TDOA supports three-sensor positioning, the normal three sensors are used to form the positioning area for positioning. However, in some cases, all three normal sensors are on the same straight line. For example, if only sensors 11, 12, and 13 are normal, normal TDOA positioning cannot be performed. Instead, the abnormal sound source location can be determined by analyzing the arrival times of the three signals to see which sensor is closer to it. In the example diagram, the arrival time can be determined by STFT to show that the abnormal signal is closer to sensor 11. (5) In some extreme cases, sensors that exceed two grid lengths are judged to have insufficient signal-to-noise ratio and cannot be used for positioning. Only two sensors remain in the sensor network, namely sensors 19 and 21, which cannot perform normal TDOA positioning. The only way to determine which sensor the abnormal sound source is closer to is by analyzing the arrival time of the two signals. In the example figure, the arrival time is determined by STFT, which shows that the abnormal signal is closer to sensor 19. (6) In some extreme cases, only one sensor remains in the sensor network. The only way to determine whether there is an abnormal sound source is to analyze whether the signal-to-noise ratio is normal. In the example figure, the presence of an abnormal sound source signal can be determined by the adaptive threshold method.

Images