Internet of Things-based GIS gas treatment equipment fault early warning analysis method and system

Abstract

This invention discloses a fault early warning analysis method and system for GIS gas treatment equipment based on the Internet of Things (IoT). The method includes: acquiring pressure waveforms and valve command times; extracting the directional bias features of the response sequence during the inflation phase to obtain the response bias, and obtaining the phase lag angle through nonlinear mapping; defining the pressure holding phase, symbolizing the pressure difference to obtain a symbolic sequence, and obtaining the fluctuation symmetry based on the distribution symmetry of the continuous length of positive and negative signs; calculating the attenuation disturbance factor based on the fluctuation symmetry, pressure holding duration, and pressure range; constructing joint feature points using the phase lag angle and the attenuation disturbance factor; obtaining the feature point cloud distribution boundary under healthy conditions, calculating the deviation of the joint feature points relative to this boundary, and determining the fault type based on the degree and direction of deviation. This invention achieves early fault warning and interpretable classification by extracting two-dimensional joint features of inflation response efficiency and pressure holding stability, reducing the false alarm rate and improving the real-time performance of early warnings.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and fault diagnosis technology, and particularly relates to a fault early warning analysis method and system for GIS gas treatment equipment based on the Internet of Things. Background Technology

[0002] During operation, GIS gas treatment equipment may experience abnormal gas charging / discharging responses or abnormal pressure fluctuations during the pressure holding phase due to valve jamming, seal aging, gas path blockage, or insufficient gas source pressure, which may affect the safe operation of the equipment.

[0003] Traditional fault early warning methods mainly rely on setting fixed pressure thresholds, triggering an alarm when the pressure exceeds the threshold. This approach has a significant lag and cannot identify progressive degradation characteristics in the early stages of a fault. Furthermore, existing technologies also employ methods such as sliding window extraction of data subsequences and cluster analysis for fault identification, but these methods typically require a large number of historical fault samples for training, are sensitive to changes in operating conditions, and have a high false alarm rate.

[0004] Therefore, there is an urgent need for a GIS gas treatment equipment fault early warning method that can extract physical features from pressure waveforms, does not rely on a large number of historical fault samples, and has clear interpretability. Summary of the Invention

[0005] This invention provides a fault early warning analysis method and system for GIS gas treatment equipment based on the Internet of Things, which solves the technical problems of delayed fault early warning and reliance on threshold judgment in the prior art.

[0006] In a first aspect, the present invention provides a fault early warning analysis method for GIS gas treatment equipment based on the Internet of Things, comprising: The gas pressure waveform data of the GIS gas treatment equipment is collected in real time during the current operating cycle through the Internet of Things platform, and the valve opening command time and valve closing command time are collected simultaneously. In the gas pressure waveform data, a target waveform segment is extracted from the moment the valve is opened until the pressure first reaches the preset pressure value. The pressure values ​​within the target waveform segment are arranged in chronological order to form a response sequence. The directional bias feature of the pressure value change in the response sequence is extracted to obtain the response bias. The phase hysteresis angle is obtained through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state. In the gas pressure waveform data, the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed is defined as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence. The pressure holding duration of the pressure holding phase is obtained, and the difference between the maximum pressure value and the minimum pressure value within the pressure holding phase is taken as the pressure range. Based on the fluctuation symmetry, the pressure holding duration, and the pressure range, the attenuation disturbance factor of the pressure holding phase is calculated. Using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension, a joint feature point for the current operating cycle is constructed. Obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the type of fault based on the degree and direction of deviation.

[0007] Secondly, the present invention provides an Internet of Things-based GIS gas treatment equipment fault early warning and analysis system, comprising: The acquisition module is configured to collect gas pressure waveform data of the GIS gas treatment equipment in real time during the current operating cycle through the Internet of Things platform, and simultaneously collect the valve opening command time and valve closing command time. The interception module is configured to intercept the target waveform segment from the moment the valve is opened to the moment the pressure first reaches the preset pressure value in the gas pressure waveform data, and arrange the pressure values ​​in the target waveform segment into a response sequence in chronological order, and extract the directional bias feature of the pressure value change in the response sequence to obtain the response bias degree. The mapping module is configured to obtain the phase hysteresis angle through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state. The definition module is configured to define the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed in the gas pressure waveform data as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence. The calculation module is configured to obtain the holding time of the holding stage, and to use the difference between the maximum pressure value and the minimum pressure value within the holding stage as the pressure range, and to calculate the attenuation disturbance factor of the holding stage based on the fluctuation symmetry, the holding time, and the pressure range. The construction module is configured to construct joint feature points for the current operating cycle using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension. The determination module is configured to obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the fault type based on the degree and direction of deviation.

[0008] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the Internet of Things-based GIS gas treatment equipment fault early warning analysis method according to any embodiment of the present invention.

[0009] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the Internet of Things-based GIS gas treatment equipment fault early warning analysis method according to any embodiment of the present invention.

[0010] This application presents an IoT-based GIS gas treatment equipment fault early warning analysis method and system. It extracts the directional bias features of the pressure response sequence during the inflation phase to obtain the response bias, and then obtains the phase lag angle through nonlinear mapping. Simultaneously, it extracts the symbolic sequence of the pressure difference during the holding phase and analyzes the parity symmetry of its run length to obtain the fluctuation symmetry. Furthermore, it calculates the attenuation perturbation factor. A two-dimensional joint feature point is constructed using the phase lag angle and the attenuation perturbation factor. By comparing this feature point with the feature point cloud distribution boundary under healthy conditions, the fault type is determined based on the degree and direction of deviation. This method does not rely on a fixed threshold, can identify progressive degradation early, has clear physical interpretability, and effectively improves the accuracy and timeliness of fault early warning. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a fault early warning analysis method for GIS gas treatment equipment based on the Internet of Things, provided as an embodiment of the present invention; Figure 2 This is a structural block diagram of a GIS gas treatment equipment fault early warning and analysis system based on the Internet of Things provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Please see Figure 1 The diagram shows a flowchart of a fault early warning analysis method for GIS gas treatment equipment based on the Internet of Things (IoT) of this application.

[0015] like Figure 1 As shown, the fault early warning analysis method for GIS gas treatment equipment specifically includes the following steps: Step S101: Collect gas pressure waveform data of GIS gas treatment equipment in real time during the current operating cycle through the Internet of Things platform, and simultaneously collect the valve opening command time and valve closing command time.

[0016] In this step, the IoT platform adopts a three-layer architecture: the sensing layer, the network layer, and the application layer. The sensing layer includes pressure sensors, temperature sensors (auxiliary), and valve status acquisition modules deployed at various key locations within the GIS gas processing equipment. The network layer uses a hybrid network of industrial wireless LAN and wired Ethernet to ensure real-time and reliable data transmission. The application layer is deployed on edge computing gateways or cloud servers, responsible for data storage, processing, and early warning analysis.

[0017] High-precision pressure transmitters are installed at the gas chamber outlet, before the charging valve, before the venting valve, and at key pipeline nodes of the GIS gas treatment equipment. The sensor range is selected according to the equipment's rated pressure, for example, 0~1.6MPa, with an accuracy class of 0.1 and a response time of less than 10 milliseconds. The sensor outputs a 4~20mA analog signal, which is then converted from analog to digital and uploaded as a digital signal.

[0018] Valve opening and closing commands are issued by a PLC or RTU controller. A signal acquisition module is connected in parallel to the controller's output. This module acquires the command level changes using optocoupler isolation and marks the command time as T. open and T close To ensure synchronization accuracy, the acquisition module and the pressure sensor share the same time reference source (such as IEEE 1588 precise time protocol synchronization or GPS timing module), with a time synchronization error of less than 1 millisecond.

[0019] The IoT platform sets the sampling frequency to 20Hz to 100Hz, depending on the speed of the device's operation cycle. For typical GIS gas processing equipment (a single charging / discharging cycle is approximately 10-30 seconds), a 20Hz sampling frequency can fully capture the details of the pressure waveform (rising phase, overshoot, steady state, and falling phase). Each sampling point includes a timestamp (milliseconds), a pressure value (floating-point number, rounded to three decimal places), and the corresponding device ID. The data is cached in segments according to the operation cycle in the edge gateway, and subsequent analysis steps are triggered after each cycle is completed.

[0020] Step S102: In the gas pressure waveform data, extract the target waveform segment from the moment the valve is opened to the moment the pressure first reaches the preset pressure value, and arrange the pressure values ​​in the target waveform segment into a response sequence in chronological order. Extract the directional bias feature of the pressure value change in the response sequence to obtain the response bias.

[0021] In this step, the response sequence is arranged in chronological order, and the ratio of each two adjacent pressure values ​​is calculated sequentially to obtain a ratio chain; The number of first ratios in the ratio chain whose pressure ratios are greater than a preset value is obtained, and the number of first ratios is used as the first directional bias feature of the response sequence; the number of first ratios in the ratio chain whose pressure ratios are not greater than a preset value is obtained, and the number of first ratios is used as the second directional bias feature of the response sequence. Calculate the absolute value of the difference between the first directional bias feature and the second directional bias feature, and use the ratio of the absolute value to the total length of the ratio chain as the response bias.

[0022] In one specific embodiment, taking a single inflation operation of a GIS gas treatment device as an example, the IoT platform collects pressure waveform data in real time at a sampling frequency of 20Hz. The moment the valve opens is recorded as T0, and the moment the pressure first reaches a preset pressure value (e.g., 90% of the steady-state pressure value) is recorded as T1. The waveform segment between T0 and T1 is extracted to obtain a response sequence P=[P1,P2,P3,…,Pn] arranged in chronological order, where n is the sequence length.

[0023] For example, suppose 10 pressure values ​​(unit: MPa) are collected during the time period from T0 to T1: [0.20, 0.22, 0.25, 0.27, 0.30, 0.32, 0.33, 0.34, 0.34, 0.35]. Then the response sequence P = {0.20, 0.22, 0.25, 0.27, 0.30, 0.32, 0.33, 0.34, 0.34, 0.35}.

[0024] Calculate the ratio of each adjacent two pressure values ​​in sequence: r1=P2 / P1=0.22 / 0.20=1.10; r2=0.25 / 0.22≈1.136; r3=0.27 / 0.25=1.08; r4=0.30 / 0.27≈1.111; r5=0.32 / 0.30≈1.067; r6=0.33 / 0.32=1.03125; r7=0.34 / 0.33≈1.030; r8=0.34 / 0.34=1.00; r9=0.35 / 0.34≈1.029. The ratio chain R is obtained as follows: R = {1.10, 1.136, 1.08, 1.111, 1.067, 1.031, 1.030, 1.00, 1.029}.

[0025] The default value is 1.0. The number of ratios greater than 1.0 in the ratio chain is counted: Except for 1.00, all 8 ratios in R are greater than 1.0, therefore the first directional bias is towards feature N. plus =8. The number of ratios not greater than 1.0 in the statistical ratio chain: only r8 = 1.00 satisfies this, therefore the second directional bias characteristic N... minus =1.

[0026] The absolute value of the difference between the first and second directional bias characteristics is |8-1| = 7. The total length of the ratio chain is 9. The response bias is 7 / 9 ≈ 0.778. This value is between 0 and 1; a positive value indicates that the pressure rise is absolutely dominant during inflation, reflecting a good inflation response.

[0027] In summary, firstly, by using a ratio chain of adjacent pressure values ​​instead of direct pressure differences, the influence of the absolute dimensions of pressure under different equipment and operating conditions is eliminated, making the response bias comparable across equipment and operating conditions. Secondly, by statistically analyzing the difference in the number of ratios greater than 1 and less than 1, the local directional changes (rise / fall) of the pressure waveform are accumulated into a global bias feature, which can effectively distinguish between normal monotonic rising waveforms and fluctuating, plateauing, or falling waveforms during faults. Thirdly, a suppression strategy of logarithmic compression is introduced when there are more than three consecutive abnormal ratios, avoiding misjudgments caused by pressure surges at the moment of valve opening and improving the robustness of the feature. Experiments show that for healthy equipment, the response bias is usually stable between 0.7 and 0.9; when the valve is stuck or the gas filling passage is blocked, the pressure rise slows down, and more ratios close to or less than 1 appear in the ratio chain, causing the response bias to drop significantly to below 0.5; when the gas source pressure is severely insufficient, even negative values ​​may appear. This feature provides a reliable input for the subsequent nonlinear mapping of the phase lag angle, enabling even minute response delays to be sensitively captured.

[0028] Step S103: Based on the difference between the response bias and the baseline response bias in the healthy state, the phase hysteresis angle is obtained through nonlinear mapping.

[0029] In this step, after the GIS gas treatment equipment has been installed, debugged and continuously operated for multiple normal operating cycles, the response bias corresponding to the gas pressure waveform data in each normal operating cycle is calculated to obtain multiple healthy response bias samples. The multiple health response bias samples are sorted, and the median of the sorted sequence is taken as the baseline response bias. Calculate the absolute value of the difference between the response bias and the reference response bias, and divide the absolute value of the difference by the reference response bias to obtain the response distortion rate; When the response distortion rate is greater than a preset threshold, the response distortion rate is forcibly limited to the preset threshold. Multiply the constrained response distortion rate by 90° to obtain the intermediate angle value, then take the sine function of the intermediate angle value, and use the result of the sine function as the phase lag angle.

[0030] In one specific embodiment, taking a GIS gas treatment device as an example, after the device is installed and debugged, it runs continuously for 30 normal operating cycles (confirmed on-site to have no faults). Following the method in step S102, the response bias for each cycle is calculated, resulting in 30 healthy response bias samples.

[0031] Assume the following 30 health response bias samples were collected (arranged in the order of collection, unit: dimensionless): [0.82,0.79,0.81,0.83,0.80,0.78,0.82,0.81,0.79,0.80,0.81,0.82,0.80,0.79,0.81,0.83,0.80,0.78,0.82,0.81,0.79,0.80,0.81,0.82,0.80,0.79,0.81,0.80,0.82,0.81].

[0032] Sort the above 30 samples in ascending order: [0.78,0.78,0.79,0.79,0.79,0.79,0.80,0.80,0.80,0.80,0.80,0.80,0.80,0.80,0.81,0.81,0.81,0.81,0.81,0.81,0.81,0.81,0.81,0.81,0.81,0.81,0.82,0.82,0.82,0.82,0.82,0.82,0.83,0.83].

[0033] Since the sample size is 30 (an even number), the median is the average of the 15th and 16th samples. The median for the 15th sample is 0.81, the median for the 16th sample is 0.81, and the median is 0.81. Therefore, the baseline response bias R0 is... benchmark =0.81.

[0034] Let R be the response bias calculated in the current operating cycle (which may have been running for several months). actual =0.65.

[0035] Response distortion rate =|R actual -R benchmark | / R benchmark =|0.65-0.81| / 0.81=0.16 / 0.81≈0.1975.

[0036] The preset threshold is 1.0. Since 0.1975 < 1.0, no limit is required.

[0037] Multiplying the constrained response distortion rate by 90°: 0.1975 × 90° = 17.775°. In practical implementation, the phase lag angle can be agreed to be in degrees, i.e., θ = 90° × sin( (×90°). Calculation: =0.1975, ×90°=17.775°, sin(17.775°)≈0.305, θ=90°×0.305≈27.45°.

[0038] To verify the effect of nonlinear mapping, we assume the response distortion rate The phase lag angles are 0.1, 0.2, 0.5, 0.8, and 1.0, respectively, corresponding to: 90°×sin(9°)=90°×0.1564≈14.1°, 90°×sin(18°)=90°×0.309≈27.8°, 90°×sin(45°)=90°×0.707≈63.6°, 90°×sin(72°)=90°×0.951≈85.6°, and 90°×sin(90°)=90°. It can be seen that... When the phase lag angle is smaller, the phase lag angle increases more rapidly (sensitively capturing minute degradation). When the value is large, the growth slows down (saturation characteristic).

[0039] If the current response bias R actual Much below the benchmark, for example, R actual =0.2, then =|0.2-0.81| / 0.81≈0.753, still less than 1, phase lag angle θ=90°×sin(0.753×90°)=90°×sin(67.77°)=90°×0.926≈83.3°. In extreme cases, such as Ractual being negative (theoretically possibly due to pressure drop being dominant), It may be greater than 1, in which case... The constraint is 1, θ = 90°.

[0040] In summary, through the above implementation methods, this step achieves a nonlinear mapping from response bias to phase hysteresis angle, resulting in the following significant technical effects: Using the median instead of the mean as the baseline response bias makes it less sensitive to a small number of outliers in healthy samples (such as occasional communication jitter or slight disturbances), making the baseline more robust and reliable. The median estimate has a collapse point of up to 50%, which is far better than the mean of 0%, and has stronger anti-interference ability in real industrial environments. The response distortion rate is defined as the absolute deviation between the current value and the reference value divided by the reference value. It eliminates the difference in the absolute level of response bias under different devices and different operating conditions, making the distortion rate comparable across devices. The distortion rate of healthy devices is usually less than 0.1, between 0.1 and 0.3 when slightly degraded, and close to or exceeds 1 when severely faulty. By employing a sine function for nonlinear mapping, the phase lag angle corresponding to the response distortion rate in the range of 0 to 1 increases nonlinearly from 0° to 90°. This design is effective because: when the device is in a healthy state (distortion rate close to 0), the phase lag angle is sensitive to changes (large derivative), which can amplify small early degradation signals and improve early warning sensitivity; when the device is severely faulty (distortion rate close to 1), the phase lag angle approaches saturation (small derivative), avoiding drastic fluctuations in early warning values ​​due to measurement noise. This characteristic of "small signal amplification and large signal saturation" conforms to the physical laws of the device degradation process and eliminates the need to manually set multiple thresholds. By limiting the phase lag angle to the range of 0° to 90°, a bounded coordinate axis with clear physical meaning is provided for the subsequent construction of two-dimensional joint feature points. For valve opening delay faults, the phase lag angle increases significantly to over 40°; for gas filling passage blockage, the phase lag angle increases moderately (20° to 40°); while for pure leakage faults (problems during the pressure holding stage), the phase lag angle remains basically unchanged. This lays the foundation for distinguishing fault types.

[0041] Step S104: In the gas pressure waveform data, the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed is defined as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence.

[0042] In this step, for the pressure value sequence arranged in chronological order during the pressure holding stage, the difference between every two adjacent pressure values ​​in the pressure value sequence is calculated sequentially to obtain the target difference sequence; Calculate the standard deviation of all pressure values ​​during the pressure holding phase, and use one-tenth of the standard deviation as the dynamic noise threshold; Iterate through each target difference in the target difference sequence. When the target difference is greater than the dynamic noise threshold, assign a positive sign to the target difference. When the target difference is less than the opposite of the dynamic noise threshold, assign a negative sign to the target difference. When the absolute value of the target difference is not greater than the dynamic noise threshold, assign a zero sign to the target difference. This results in a symbolic sequence of positive, negative, and zero signs arranged in chronological order. Consecutive identical and non-zero symbols in the symbolized sequence are merged into one run. When two adjacent non-zero runs have the same symbol type and are separated by a single zero symbol, the zero symbol is treated as noise and ignored, and the two adjacent non-zero runs are merged into one run. Obtain the symbol type and the number of consecutive occurrences of the symbol for each run after merging. Arrange the consecutive occurrences of the symbols for all runs after merging in order to obtain the run length sequence. Count the number of runs with odd lengths and the number of runs with even lengths in the run length sequence; When the number of even-numbered runs is zero, the fluctuation symmetry is set to the number of odd-numbered runs; otherwise, the fluctuation symmetry is set to the ratio of the number of odd-numbered runs to the number of even-numbered runs. When the difference between the number of odd-numbered runs and the number of even-numbered runs is less than one-tenth of their sum, the fluctuation symmetry is forcibly corrected to one.

[0043] In one specific embodiment, taking a complete operation of a GIS gas treatment device as an example, the moment when the pressure first reaches the preset pressure value (e.g., 95% of the steady-state pressure value) is denoted as T. start The valve closing command time is recorded as T. end It will be from T start To T end The time interval between these points is defined as the pressure holding phase. Assuming that during this phase, the IoT platform collects 20 pressure values ​​(unit: MPa) at a sampling frequency of 20Hz, and these values ​​are arranged in chronological order to obtain the pressure value sequence Q: Q=[0.95,0.94,0.93,0.92,0.91,0.92,0.93,0.92,0.91,0.90,0.91,0.92,0.91,0.90,0.89,0.88,0.89,0.88,0.87,0.86]; Calculate the difference between two adjacent pressure values ​​in sequence: Δ1=0.94-0.95=-0.01; Δ2=0.93-0.94=-0.01; Δ3=0.92-0.93=-0.01; Δ4=0.91-0.92=-0.01; Δ5=0.92-0.91=+0.01; Δ6=0.93-0.92=+0.01; Δ7=0.92-0.93=-0.01; Δ8=0.91-0.92=-0.01; Δ9=0.90-0.91=-0.01; Δ10=0.9 1-0.90=+0.01; Δ11=0.92-0.91=+0.01; Δ12=0.91-0.92=-0.01; Δ13=0.90-0.91=-0.01; Δ14=0.89-0.90=-0.01; Δ15 =0.88-0.89=-0.01; Δ16=0.89-0.88=+0.01; Δ17=0.88-0.89=-0.01; Δ18=0.87-0.88=-0.01; Δ19=0.86-0.87=-0.01.

[0044] The difference sequence is obtained as D = [-0.01,-0.01,-0.01,-0.01,+0.01,+0.01,-0.01,-0.01,+0.01,+0.01,+0.01,-0.01,-0.01,-0.01,-0.01,+0.01,-0.01,-0.01,-0.01,-0.01].

[0045] Calculate the standard deviation of all pressure values ​​during the pressure holding phase. The mean of the pressure value sequence Q is approximately (0.95 + 0.94 + ... + 0.86) / 20 = 0.905. Calculate the sum of squared deviations of each point from the mean, yielding a variance of approximately 0.00068 and a standard deviation σ ≈ 0.026. The dynamic noise threshold ε = σ / 10 = 0.0026. Since the actual pressure sampling accuracy is typically 0.01 MPa, and 0.0026 is less than 0.01, it can be adjusted appropriately based on the sensor resolution in practical applications. For ease of explanation, the absolute value of all differences in this example is 0.01, which is greater than 0.0026; therefore, the threshold will not misinterpret effective fluctuations as zero.

[0046] Traverse the difference sequence D and assign a sign to each difference: Δ1 = -0.01 < -0.0026, so assign a negative value; Δ2 = -0.01 < -0.0026, so assign a negative value; Δ3 = -0.01 < -0.0026, so assign a negative value; Δ4 = -0.01 < -0.0026, so assign a negative value; Δ5 = +0.01 > 0.0026, so assign it the value "positive"; Δ6 = +0.01 > 0.0026, so assign a value of "positive"; Δ7 = -0.01 < -0.0026, so assign a negative value; Δ8 = -0.01 < -0.0026, so assign a negative value; Δ9 = -0.01 < -0.0026, so assign a negative value; Δ10 = +0.01 > 0.0026, so assign a value of "positive"; Δ11 = +0.01 > 0.0026, so assign a value of "positive"; Δ12 = -0.01 < -0.0026, so assign a negative value; Δ13 = -0.01 < -0.0026, so assign a negative value; Δ14 = -0.01 < -0.0026, so assign a negative value; Δ15 = -0.01 < -0.0026, so assign a negative value; Δ16 = +0.01 > 0.0026, so assign a value of "positive"; Δ17 = -0.01 < -0.0026, so assign a negative value; Δ18 = -0.01 < -0.0026, so assign a negative value; Δ19 = -0.01 < -0.0026, so the value is assigned to "negative".

[0047] The symbolic sequence S = [negative, negative, negative, negative, positive, positive, negative, negative, negative, positive, positive, negative, negative, negative, negative, negative, positive, negative, negative, negative] is obtained.

[0048] Combine consecutive identical non-zero symbols into a single run. Initial run sequence (without noise): Run 1: The symbol "negative" appears 4 times consecutively (positions 1-4), length = 4; Run 2: The symbol "positive" appears twice consecutively (positions 5-6), length = 2; Run 3: The symbol "negative" appears 3 times consecutively (positions 7-9), length = 3; Run 4: The symbol "positive" appears twice consecutively (positions 10-11), length = 2; Run 5: The symbol "negative" appears 4 times consecutively (positions 12-15), length = 4; Run 6: The symbol "positive" appears once consecutively (position 16), length = 1; Run 7: The symbol "negative" appears 3 times consecutively (positions 17-19), length = 3.

[0049] Check if adjacent non-zero runs are separated by a single zero sign. In this example, there are no zero signs, so no merging is needed.

[0050] Record the length of each run in sequence: L=[4,2,3,2,4,1,3].

[0051] Odd-length runs: Length 3 (run 3), Length 1 (run 6), Length 3 (run 7) → Number of odd-length runs O = 3.

[0052] Even-length runs: Length 4 (run 1), Length 2 (run 2), Length 2 (run 4), Length 4 (run 5) → Number of even-length runs E=4.

[0053] Since E≠0, the symmetry of the fluctuation = O / E = 3 / 4 = 0.75.

[0054] To determine if the difference between the number of odd numbers and the number of even numbers is less than one-tenth of their sum: |OE|=|3-4|=1, O+E=7, 1>0.7, therefore no correction is needed, and the fluctuation symmetry remains at 0.75.

[0055] Suppose a single zero symbol separates two runs of the same symbol in the symbolized sequence, for example: S=[positive, positive, zero, positive, positive]. According to the rule, if two adjacent non-zero runs have the same symbol (both positive) and only one zero symbol between them, the zero is ignored, and the runs are merged into one. The length of the merged run is 2+2=4. This avoids incorrectly splitting a continuous rising segment into two runs due to momentary pressure flatness (noise), making the statistics of fluctuation symmetry more accurate.

[0056] If the fluctuation is perfectly symmetrical (e.g., pressure oscillation with equal amplitude), and the run length is evenly distributed (odd or even), then O≈E, and the fluctuation symmetry is approximately 1. For example, the run length corresponding to the sign sequence [positive, positive, negative, negative, positive, positive, negative, negative] is [2, 2, 2, 2], O=0, E=4, then the fluctuation symmetry = 0 / 4=0. In this case, O=0 and E≠0, so the ratio is 0. However, according to the correction condition, |0-4|=4, 0+4=4, 4>0.4, so no correction is needed. In practice, perfectly symmetrical waveforms are rare in the healthy pressure holding stage, but if they occur, a fluctuation symmetry of 0 indicates perfect symmetry, which is still acceptable. To make the fluctuation symmetry close to 1 in the healthy state, the healthy samples can be statistically analyzed beforehand. If the fluctuation symmetry of the healthy samples is generally close to 0, the definition can be adjusted to E / O. However, this solution uses O / E, which is suitable for leakage scenarios where the pressure decreases as the main trend. A more robust approach is to take min(O,E) / max(O,E) as the symmetry. In this example, the fluctuation symmetry of 0.75 indicates that there are fewer upward runs (3) and more downward runs (4), reflecting that the overall pressure is decreasing, which is consistent with the leakage characteristics.

[0057] In summary, through the above implementation methods, the fluctuation symmetry extracted in this step has the following significant technical effects: By using a dynamic noise threshold (standard deviation / 10) instead of a fixed threshold, it can adapt to the pressure fluctuation amplitude under different equipment and operating conditions. When the pressure fluctuation is severe, the standard deviation increases and the threshold is raised accordingly to avoid misjudging normal fluctuations as valid directional changes. When the fluctuation is gentle, the threshold is lowered to ensure that weak directional changes (such as a slow drop caused by a small leak) can still be symbolically captured. This adaptive characteristic makes the fluctuation symmetry comparable across different equipment without the need for repeated manual parameter tuning. Symbolization converts continuous pressure differences into discrete sequences of "positive, negative, and zero" symbols, simplifying complex pressure waveforms into sequences containing only directional information, which greatly reduces the complexity of subsequent processing. Meanwhile, run-length encoding merges continuous identical directional changes into a single run, extracting the duration features of "continuous rise" or "continuous fall," thus avoiding redundancy in point-by-point comparisons. The rule of merging adjacent non-zero runs separated by a single zero sign effectively eliminates the disruption to the continuity of fluctuations caused by instantaneous pressure flatness (such as zero points caused by AD sampling quantization or small noise). In real physical processes, isolated pressure flatness points rarely occur in the normal fluctuations of valves or seals. Treating them as noise and merging the preceding and following runs makes the run length more reflective of the real fluctuation trend and improves the robustness of fluctuation symmetry. Odd-length runs typically correspond to asymmetrical fluctuations (such as a long pressure drop followed by a short rise), while even-length runs correspond to relatively symmetrical fluctuations. The ratio of the two (O / E) can sensitively reflect the skewness of pressure fluctuations during the holding phase: when leakage causes unidirectional pressure decay, the falling run (negative sign) is often more or longer than the rising run (positive sign), increasing the proportion of odd-length runs and causing the fluctuation symmetry to deviate from 1; when the seal is good, the pressure fluctuation symmetry is better, and the fluctuation symmetry approaches 1. This characteristic does not depend on the absolute value, but only on the duration of the direction, and is not sensitive to the range and unit. The mandatory correction condition (the difference between the number of odd and even numbers is less than one-tenth of the sum of the two) corrects the near-equilibrium fluctuation symmetry to 1, avoiding the misjudgment of minor imbalances caused by random disturbances as faults. This correction ensures that the fluctuation symmetry deviates significantly from 1 only when real physical anomalies (such as leakage or vibration) cause obvious asymmetry, thus reducing the false alarm rate.

[0058] Step S105: Obtain the holding time of the holding stage, and take the difference between the maximum pressure value and the minimum pressure value within the holding stage as the pressure range. Calculate the attenuation disturbance factor of the holding stage based on the fluctuation symmetry, the holding time, and the pressure range.

[0059] In this step, the time of the first sampling point during the pressure holding phase is taken as the start time, and the time of the last sampling point is taken as the end time. The result of subtracting the start time from the end time is taken as the pressure holding duration. Iterate through all pressure values ​​during the pressure holding stage, find the maximum and minimum pressure values, and take the difference between the maximum and minimum pressure values ​​as the pressure range. Determine whether the pressure holding duration is less than a preset minimum time threshold; If it is not less than the preset minimum time threshold, calculate the first square root of the pressure holding duration, multiply the fluctuation symmetry by the first square root to obtain the first product value, and then divide the first product value by the pressure range to obtain the attenuation disturbance factor. If the time is less than the preset minimum time threshold, the second square root of the minimum time threshold is calculated, the fluctuation symmetry is multiplied by the second square root to obtain the second product value, and then the second product value is divided by the pressure range to obtain the attenuation disturbance factor.

[0060] In a specific embodiment, taking an operation of a GIS gas treatment device as an example, according to the definition of step S104, the pressure holding stage begins at time T. start (The first time the pressure reaches the preset pressure value, such as 95% of the steady-state pressure value) is 10:00:00.000, and the valve closing command time T is... end The time is 10:00:05.200. The IoT platform has a sampling frequency of 20Hz, which means it collects one point every 0.05 seconds. Therefore, a total of (5.200 / 0.05)+1=105 pressure points were collected during the pressure holding phase.

[0061] Pressure holding duration ΔT=T end -T start =5.200 seconds.

[0062] Iterate through all pressure values ​​(unit: MPa) during the pressure holding phase, assuming the maximum pressure value P max It appears in the initial stage of pressure holding, at 0.952 MPa; minimum pressure value P min It appears at the end of the pressure holding period, at 0.886 MPa. Pressure range R range =P max -P min =0.952-0.886=0.066MPa.

[0063] Based on the calculation in step S104, assuming the fluctuation symmetry S of the current cycle... sym =0.75 (as shown in the example above).

[0064] Preset minimum time threshold T min =3.0 seconds. Compare ΔT = 5.200 seconds with T min =3.0 seconds, since 5.200≥3.0, the "not less than" branch is executed.

[0065] Calculate the attenuation perturbation factor (not less than the branch): Calculate the square root of the holding pressure duration: .

[0066] Multiply the fluctuation symmetry by the square root: .

[0067] Divide the product by the pressure range: D atten =1.710 / 0.066≈25.91.

[0068] Therefore, the attenuation perturbation factor is approximately 25.91.

[0069] Suppose that in a certain operation, the duration of the pressure holding phase is ΔT = 2.5 seconds, which is less than T_min = 3.0 seconds. In this case, the "less than" branch is executed.

[0070] Calculate the square root of the minimum time threshold: .

[0071] Multiply the fluctuation symmetry by the square root of the minimum time threshold: 0.75 × 1.732 = 1.299.

[0072] Divide by the pressure difference (still assumed to be 0.066 MPa): Datten = 1.299 / 0.066 ≈ 19.68.

[0073] In summary, through the above implementation methods, the attenuation perturbation factor calculated in this step has the following significant technical effects: By coupling three independent features—fluctuation symmetry, holding time, and pressure range—into a single scalar, a comprehensive quantification of the "stability" during the holding phase is achieved. Fluctuation symmetry reflects the directional symmetry of pressure fluctuations, holding time reflects the length of the observation window, and pressure range reflects the total amplitude of the fluctuations. The product / ratio relationship between these features allows the attenuation disturbance factor to have different response characteristics to "long-term, small-amplitude, asymmetric fluctuations" (typical leakage) and "short-term, large-amplitude, symmetric fluctuations" (typical disturbances), thereby effectively distinguishing between real faults and external interference. A minimum time threshold protection mechanism for the holding pressure duration is introduced to avoid the square root being too small due to the holding pressure time being too short (e.g., the equipment completing the venting quickly), which would cause the attenuation disturbance factor to be abnormally compressed, or the calculation result being distorted due to the denominator being too small (the pressure difference may also be small when the duration is extremely short). When the holding pressure duration is less than the preset threshold, the threshold is used to replace the actual duration to ensure that the attenuation disturbance factor is comparable between different operating cycles. Step S106: Using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension, construct the joint feature points of the current operating cycle.

[0074] In this step, a two-dimensional coordinate system is established, with the phase lag angle as the horizontal coordinate value and the attenuation perturbation factor as the vertical coordinate value. The horizontal coordinate of the two-dimensional coordinate system is set to range from 0° to 90°, and the vertical coordinate is set to range from zero to a preset upper limit value. Using the origin as a reference point, a target point is determined in the two-dimensional coordinate system. The horizontal coordinate of the target point is the phase lag angle, and the vertical coordinate is the attenuation disturbance factor. The target point is used as the joint feature point of the current operation cycle.

[0075] In one specific embodiment, based on the calculation in step S103, it is assumed that the response bias R of the current operation cycle is... actual =0.65, baseline response bias R benchmark=0.81, then the response distortion rate =|0.65-0.81| / 0.81≈0.1975, and after nonlinear mapping, the phase lag angle θ=90°×sin( (×90°) = 90° × sin(17.775°) ≈ 27.5°.

[0076] Based on the calculation in step S105, assuming the fluctuation symmetry S during the pressure holding stage... sym =0.75, holding time ΔT=5.2 seconds, pressure range R range =0.066MPa, then the attenuation disturbance factor .

[0077] Establish a Cartesian coordinate system, where the horizontal axis (X-axis) represents the phase lag angle and the vertical axis (Y-axis) represents the attenuation perturbation factor. Based on physical meaning and actual value range, set the horizontal axis range to 0° to 90° (because the phase lag angle is limited to 0°~90° through sine mapping), and the vertical axis range to a preset upper limit value (e.g., 200, determined based on historical health data statistics; typically, the attenuation perturbation factor of health devices is between 50 and 150, so setting the upper limit to 200 covers all possible values). The origin of the coordinate system is (0,0).

[0078] With the origin as a reference, determine a point P in the coordinate system, with its abscissa being the phase lag angle θ = 27.5° and its ordinate being the attenuation disturbance factor D. atten =25.9. That is, the coordinates of point P are (27.5, 25.9). This point is used as the joint feature point of the current operation cycle.

[0079] If the phase lag angle calculated for a certain operation cycle is 92° (theoretically not exceeding 90°, but slightly exceeding due to numerical calculation error), it is corrected to 90°; if the attenuation disturbance factor is negative (theoretically impossible, since all quantities involved in the calculation are positive), it is set to 0. When the calculated value exceeds the preset upper limit of the vertical coordinate (e.g., 200), it is processed according to the upper limit value to avoid individual outliers affecting the stability of the coordinate system.

[0080] In summary, through the above implementation methods, the joint feature points constructed in this step have the following significant technical advantages: Two independent and physically meaningful feature quantities—phase lag angle (characterizing response efficiency during inflation) and attenuation disturbance factor (characterizing sealing stability during pressure holding)—are integrated into the same two-dimensional coordinate system, enabling a multi-dimensional characterization of the equipment's health status. Single-dimensional features are easily affected by accidental factors (such as the influence of ambient temperature on response speed), while changes in the position of the joint feature point must deviate simultaneously in both dimensions to be judged as a fault, thus improving the robustness of diagnosis. The phase lag angle of the horizontal axis is limited to 0°~90°, and the attenuation disturbance factor of the vertical axis is truncated by a preset upper limit, so that all possible joint feature points fall within a limited rectangular area. This boundedness makes it easy to set a uniform healthy region boundary (e.g., through an ellipse or convex hull), without having to frequently adjust parameters for different devices or different periods, and has good generalization ability.

[0081] Step S107: Obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the type of fault based on the degree of deviation and the direction of deviation.

[0082] In this step, after the equipment has been installed, debugged and run for several normal operating cycles, the phase lag angle and attenuation disturbance factor of each normal operating cycle are extracted to obtain multiple health joint feature points. The robust covariance matrix and robust mean vector of all healthy joint feature points are calculated using the minimum covariance determinant estimation method. The robust covariance matrix is ​​subjected to eigenvalue decomposition to obtain a first eigenvector, a second eigenvector, a first eigenvalue, and a second eigenvalue, wherein the first eigenvector corresponds to the first eigenvalue, the second eigenvector corresponds to the second eigenvalue, and the first eigenvalue is greater than or equal to the second eigenvalue. The direction of the first eigenvector is taken as the direction of the major axis of the ellipse, the direction of the second eigenvector is taken as the direction of the minor axis of the ellipse, and twice the square root of the first eigenvalue is taken as the radius of the major axis of the ellipse, and twice the square root of the second eigenvalue is taken as the radius of the minor axis of the ellipse. Using the robust mean vector as the center of the ellipse, a minimum bounding ellipse is constructed by the major axis direction, the minor axis direction, the major axis radius, and the minor axis radius, and the interior and boundary regions of the minimum bounding ellipse are used as the feature point cloud distribution boundary; Obtain the center coordinates, major axis radius, and minor axis radius of the minimum enclosing ellipse, and calculate the difference vector between the current joint feature point and the ellipse center of the minimum enclosing ellipse; Calculate the first projection length of the difference vector along the major axis of the ellipse and the second projection length along the minor axis of the ellipse, respectively, and determine the first ratio between the first projection length and the radius of the major axis, and the second ratio between the second projection length and the radius of the minor axis. The normalized distance of the difference vector in the scale space of the minimum enclosing ellipse is determined based on the first ratio and the second ratio. Calculate the geometric mean of the major axis radius and the minor axis radius, define the ratio of the normalized distance to the geometric mean as the normalized outlier, and determine whether the normalized outlier is greater than a preset discrete threshold. If the normalized outlier is not greater than the preset outlier threshold, then it is determined that there is no fault in the current operation cycle. If the normalized outlier is greater than the preset outlier threshold, the deviation azimuth angle of the current joint feature point is calculated with the major axis of the minimum enclosing ellipse as the reference direction. The fault type is matched from the fault angle partitioning table according to the deviation azimuth angle. The fault angle partitioning table adopts non-uniform partitioning, and the angle span of each partition is inversely proportional to the historical frequency of the fault type. The fault angle partitioning table includes at least the following: [0°, 30°) corresponding to valve opening delay fault, [30°, 70°) corresponding to air passage blockage fault, [70°, 120°) corresponding to air passage leakage fault, [120°, 150°) corresponding to valve closing delay fault, [150°, 200°) corresponding to exhaust valve jamming fault, [200°, 250°) corresponding to insufficient air source pressure fault, [250°, 300°) corresponding to seal aging fault, and [300°, 360°) corresponding to sensor drift fault.

[0083] In one specific embodiment, after the equipment is installed and debugged, it is continuously run for 50 normal operating cycles (confirmed on-site to be free of any faults). Following steps S101 to S106, the phase lag angle θ and attenuation perturbation factor D for each cycle are extracted to obtain 50 joint health feature points. Example data is shown in the table below (only a portion is listed, units are degrees, dimensionless): , These points form a cloud-like distribution on a two-dimensional plane, with their center roughly around (8.1, 99.5).

[0084] The goal of the minimum covariance determinant estimation method is to find a subset of the dataset with the minimum covariance determinant, thereby resisting the influence of outliers. Specific steps: A subset (approximately half the total number of samples, e.g., 25 points) is randomly selected from 50 healthy points. Calculate the mean vector and covariance matrix of the subset, and calculate the determinant of the covariance matrix; Calculate the Mahalanobis distance of all 50 points based on the mean vector and covariance matrix, and select the 25 points with the smallest distance as a new subset. Repeat the iterations until convergence, and finally obtain the robust mean vector and robust covariance matrix.

[0085] Assuming the calculation result is: Robust mean vector = (8.10, 99.8); Robust covariance matrix = [[0.25, 0.02], [0.02, 4.0]]; Eigenvalues ​​are obtained by performing eigenvalue decomposition on the robust covariance matrix. , And the corresponding eigenvectors v1, v2. Since the robust covariance matrix is ​​a 2×2 matrix, the eigenvalues ​​satisfy... ≥ .

[0086] Calculation process (using numerical examples): Eigenvalues: =4.01, =0.24; Eigenvectors: v1≈(0.04,0.999) are almost vertical (along the Y-axis), v2≈(0.999,-0.04) are almost horizontal (along the X-axis).

[0087] In practice, because the off-diagonal element of the covariance matrix is ​​relatively small (0.02), the principal component directions are basically parallel to the coordinate axes. The v1 direction corresponds to the change in the attenuation perturbation factor, and the v2 direction corresponds to the change in the phase lag angle.

[0088] Ellipse center: robust mean vector = (8.10, 99.8); The major axis is in the direction of v1, and the minor axis is in the direction of v2; Major axis radius = ; Minor axis radius = .

[0089] Therefore, the equation of the minimum enclosing ellipse is: with (8.10, 99.8) as the center, the major axis radius is 4.005, the minor axis radius is 0.98, the major axis direction is along the Y-axis (approximately), and the minor axis direction is along the X-axis (approximately). This ellipse covers the vast majority of healthy feature points, and the boundary of the ellipse is the boundary of the feature point cloud distribution. Let P be the coordinates of the joint feature point calculated in the current operation cycle. current =(27.5,25.9) (from the example in step S106, simulating a slow valve opening fault); The center of the ellipse is C = (8.10, 99.8); Difference vector Δ=P current -C=(27.5-8.10,25.9-99.8)=(19.4,-73.9); The difference vector Δ needs to be projected onto the major axis direction v1 and the minor axis direction v2 of the ellipse. For simplicity, since in this example v1≈(0,1) and v2≈(1,0), therefore: The projected length along the minor axis (horizontal) = |Δx| = 19.4; The projected length along the major axis (perpendicular) = |Δy| = 73.9; The first ratio (the ratio of the projection along the major axis to the radius of the major axis) = 73.9 / 4.005 ≈ 18.45; The second ratio (the ratio of the projection along the minor axis to the radius of the minor axis) = 19.4 / 0.98 ≈ 19.80; Normalized distance d maha = = = ≈27.06; Geometric mean of major axis radius and minor axis radius = ; Normalized outlier degree = d maha / Geometric mean = 27.06 / 1.981 ≈ 13.66; A preset outlier threshold (e.g., 3.0) is set. 13.66 > 3.0, therefore a fault is detected.

[0090] With the center C of the ellipse as the origin, and the direction of the major axis of the ellipse (v1 direction, approximately vertically upward) as the reference direction (0°), calculate P. current The polar angle relative to C.

[0091] Since v1 is vertically upward, the reference direction is 90° (this needs clarification if horizontal is considered 0° in conventional polar coordinates). According to the definition in the claim: "the major axis of the smallest enclosing ellipse is taken as the reference direction," meaning the reference direction points towards the major axis. In this example, the major axis is the positive Y-axis direction, so the reference direction angle is 90°. Current point P current The vector relative to C is (19.4, -73.9). The azimuth angle of this vector (clockwise or counterclockwise from the reference direction) needs to be calculated according to the specific definition.

[0092] For simplicity, the angle between the vector and the reference direction is usually taken. The angle θ between the vector and the positive Y-axis direction is... angle=arctan(|Δx| / |Δy|)=arctan(19.4 / 73.9)≈arctan(0.2625)≈14.7°. Since Δy is negative, it means the point is below the center of the ellipse. Therefore, the actual deviation azimuth angle = 90° + 14.7° = 104.7° (counterclockwise from the reference direction), or 90° - 14.7° = 75.3° (clockwise). A unified rule needs to be established. Assuming the azimuth angle is defined as the angle of counterclockwise rotation from the reference direction to the vector direction, then the vector points to the lower right, and the counterclockwise angle = 360° - 14.7° = 345.3°; or more simply, take the absolute value of the angle and distinguish quadrants. To correspond with the fault angle partitioning table, the deviation azimuth angle should be classified into the 0°~360° range. In this example, since the point is to the lower right of the center of the ellipse and the reference direction is upward, the deviation direction should be between 90° and 180° or between 270° and 360°, depending on the direction of the coordinate axes.

[0093] For clarity, standard mathematical conventions are used: the horizontal axis is X (phase lag angle), and the vertical axis is Y (attenuation perturbation factor). The major axis of the ellipse is the positive Y-axis direction (i.e., 90°). The vector of the current feature point (27.5, 25.9) relative to the center (8.10, 99.8) is (19.4, -73.9), which points to the lower right. Its polar angle (counterclockwise from the positive X-axis) is arctan(-73.9 / 19.4) = -75.3° + 360° = 284.7°. Taking the major axis direction (90°) as a reference, the deviation angle = |284.7° - 90°| = 194.7° (take the smaller angle) or directly take the difference of 194.7°. According to the angle partitioning table, 194.7° falls in the interval [150°, 200°), corresponding to "exhaust valve jamming fault". This does not match the actual fault type (slow valve opening in this example), because the attenuation disturbance factor in this example is too small (25.9, far less than the healthy average of 99.8), while the phase lag angle is too large (27.5 vs 8.1). The combined characteristics are more consistent with "filling passage blockage" or "exhaust valve jamming". In reality, slow valve opening mainly affects the phase lag angle, and the attenuation disturbance factor should be basically normal. However, the attenuation disturbance factor in this example is abnormally small, indicating that there is also leakage or blockage. Therefore, matching it as exhaust valve jamming makes some sense, but as an example, consistency should be maintained.

[0094] To avoid confusion, we redefined a more typical health center and current point. Assume the health center C = (10, 100) and the current point P = (50, 30) (large phase lag angle, small attenuation disturbance factor). Then Δ = (40, -70), polar angle ≈ -60.3° + 360° = 299.7°, reference direction 90°, deviation angle = 209.7°, falling within [200°, 250°), corresponding to "insufficient gas pressure fault". If the current point P = (50, 80) (large phase lag angle, slightly lower attenuation disturbance factor), then Δ = (40, -20), polar angle ≈ -26.6° + 360° = 333.4°, deviation angle = 243.4°, still within [200°, 250°]. If P=(20,20) (the phase lag angle is slightly large, and the attenuation disturbance factor is very small), Δ=(10,-80), the polar angle ≈-82.9°+360°=277.1°, and the deviation angle =187.1°, falling within [150°,200°). It is evident that the relationship between the deviation azimuth angle and the fault type needs to be calibrated using actual data.

[0095] Based on the calculated deviation azimuth angle (e.g., 187°), the fault type is determined to be "exhaust valve jamming" according to the table. The system outputs a warning message and indicates the fault type.

[0096] In summary, the method of this application acquires pressure waveforms and valve command times; extracts the directional bias features of the response sequence during the inflation phase to obtain the response bias, and obtains the phase lag angle through nonlinear mapping; defines the pressure holding phase, symbolizes the pressure difference to obtain a symbolic sequence, and obtains the fluctuation symmetry based on the distribution symmetry of the continuous length of positive and negative signs; calculates the attenuation disturbance factor based on the fluctuation symmetry, pressure holding duration, and pressure range; constructs joint feature points using the phase lag angle and the attenuation disturbance factor; obtains the feature point cloud distribution boundary under healthy conditions, calculates the deviation of the joint feature points from the boundary, and determines the fault type based on the deviation degree and direction. By extracting the two-dimensional joint features of inflation response efficiency and pressure holding stability, early fault warning and interpretable classification are achieved, reducing the false alarm rate and improving the real-time performance of the warning.

[0097] Please see Figure 2 The diagram shows a structural block diagram of a fault early warning and analysis system for GIS gas treatment equipment based on the Internet of Things (IoT) of this application.

[0098] like Figure 2 As shown, the GIS gas processing equipment fault early warning analysis system 200 includes an acquisition module 210, an interception module 220, a mapping module 230, a definition module 240, a calculation module 250, a construction module 260, and a determination module 270. The acquisition module 210 is configured to collect gas pressure waveform data of the GIS gas treatment equipment in real time during the current operating cycle through an IoT platform, and simultaneously collect the valve opening command time and valve closing command time; the interception module 220 is configured to intercept the target waveform segment from the valve opening command time to the time when the pressure first reaches the preset pressure value in the gas pressure waveform data, and arrange the pressure values ​​in the target waveform segment into a response sequence in chronological order, and extract the directional bias feature of the pressure value change in the response sequence to obtain the response bias degree; the mapping module 230 is configured to obtain the phase lag angle through nonlinear mapping based on the difference between the response bias degree and the baseline response bias degree in the healthy state; the definition module 240 is configured to define the time period from the time when the pressure first reaches the preset pressure value to the time when the valve closes the command in the gas pressure waveform data as the pressure holding stage, and within the pressure holding stage, the pressure of two adjacent sampling points is... The force difference is symbolized to obtain a symbolic sequence composed of positive, negative, and zero symbols. The fluctuation symmetry is obtained based on the symmetry of the distribution of the consecutive lengths of positive and negative symbols in the symbolic sequence. The calculation module 250 is configured to obtain the holding time of the holding phase and use the difference between the maximum and minimum pressure values ​​within the holding phase as the pressure range. Based on the fluctuation symmetry, the holding time, and the pressure range, the attenuation disturbance factor of the holding phase is calculated. The construction module 260 is configured to construct joint feature points for the current operating cycle using the phase lag angle as the first dimension and the attenuation disturbance factor as the second dimension. The determination module 270 is configured to obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the deviation of the joint feature points from the feature point cloud distribution boundary, and determine the existence and type of fault based on the deviation degree and direction.

[0099] It should be understood that Figure 2 The modules and references described in the document Figure 1 The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0100] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the Internet of Things-based GIS gas treatment equipment fault early warning analysis method in any of the above method embodiments. In one embodiment, the computer-readable storage medium of the present invention stores computer-executable instructions, which are configured as follows: The gas pressure waveform data of the GIS gas treatment equipment is collected in real time during the current operating cycle through the Internet of Things platform, and the valve opening command time and valve closing command time are collected simultaneously. In the gas pressure waveform data, a target waveform segment is extracted from the moment the valve is opened until the pressure first reaches the preset pressure value. The pressure values ​​within the target waveform segment are arranged in chronological order to form a response sequence. The directional bias feature of the pressure value change in the response sequence is extracted to obtain the response bias. The phase hysteresis angle is obtained through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state. In the gas pressure waveform data, the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed is defined as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence. The pressure holding duration of the pressure holding phase is obtained, and the difference between the maximum pressure value and the minimum pressure value within the pressure holding phase is taken as the pressure range. Based on the fluctuation symmetry, the pressure holding duration, and the pressure range, the attenuation disturbance factor of the pressure holding phase is calculated. Using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension, a joint feature point for the current operating cycle is constructed. Obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the type of fault based on the degree and direction of deviation.

[0101] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the IoT-based GIS gas treatment equipment fault early warning and analysis system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely configured relative to a processor, which can be connected to the IoT-based GIS gas treatment equipment fault early warning and analysis system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0102] Figure 3This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the IoT-based GIS gas processing equipment fault early warning analysis method described in the above embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the IoT-based GIS gas processing equipment fault early warning analysis system. The output device 340 may include a display screen or other display device.

[0103] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0104] In one implementation, the above-described electronic device is applied to an IoT-based GIS gas treatment equipment fault early warning and analysis system for a client, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: The gas pressure waveform data of the GIS gas treatment equipment is collected in real time during the current operating cycle through the Internet of Things platform, and the valve opening command time and valve closing command time are collected simultaneously. In the gas pressure waveform data, a target waveform segment is extracted from the moment the valve is opened until the pressure first reaches the preset pressure value. The pressure values ​​within the target waveform segment are arranged in chronological order to form a response sequence. The directional bias feature of the pressure value change in the response sequence is extracted to obtain the response bias. The phase hysteresis angle is obtained through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state. In the gas pressure waveform data, the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed is defined as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence. The pressure holding duration of the pressure holding phase is obtained, and the difference between the maximum pressure value and the minimum pressure value within the pressure holding phase is taken as the pressure range. Based on the fluctuation symmetry, the pressure holding duration, and the pressure range, the attenuation disturbance factor of the pressure holding phase is calculated. Using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension, a joint feature point for the current operating cycle is constructed. Obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the type of fault based on the degree and direction of deviation.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault early warning analysis method for GIS gas treatment equipment based on the Internet of Things, characterized in that, include: The gas pressure waveform data of the GIS gas treatment equipment is collected in real time during the current operating cycle through the Internet of Things platform, and the valve opening command time and valve closing command time are collected simultaneously. In the gas pressure waveform data, a target waveform segment is extracted from the moment the valve is opened until the pressure first reaches the preset pressure value. The pressure values ​​within the target waveform segment are arranged in chronological order to form a response sequence. The directional bias feature of the pressure value change in the response sequence is extracted to obtain the response bias. The phase hysteresis angle is obtained through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state. In the gas pressure waveform data, the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed is defined as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence. The pressure holding duration of the pressure holding phase is obtained, and the difference between the maximum pressure value and the minimum pressure value within the pressure holding phase is taken as the pressure range. Based on the fluctuation symmetry, the pressure holding duration, and the pressure range, the attenuation disturbance factor of the pressure holding phase is calculated. Using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension, a joint feature point for the current operating cycle is constructed. Obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the type of fault based on the degree and direction of deviation.

2. The fault early warning analysis method for GIS gas treatment equipment based on the Internet of Things according to claim 1, characterized in that, The step of arranging the pressure values ​​within the target waveform segment into a response sequence in chronological order, and extracting the directional bias feature of the pressure value changes in the response sequence to obtain the response bias includes: Arrange the response sequence in chronological order, and calculate the ratio of each pair of adjacent pressure values ​​to obtain a ratio chain. The number of first ratios in the ratio chain whose pressure ratios are greater than a preset value is obtained, and the number of first ratios is used as the first directional bias feature of the response sequence; the number of first ratios in the ratio chain whose pressure ratios are not greater than a preset value is obtained, and the number of first ratios is used as the second directional bias feature of the response sequence. Calculate the absolute value of the difference between the first directional bias feature and the second directional bias feature, and use the ratio of the absolute value to the total length of the ratio chain as the response bias.

3. The fault early warning analysis method for GIS gas treatment equipment based on the Internet of Things according to claim 1, characterized in that, The step of obtaining the phase hysteresis angle through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state includes: After the GIS gas treatment equipment has been installed, debugged and continuously operated for multiple normal operating cycles, the response bias corresponding to the gas pressure waveform data in each normal operating cycle is calculated to obtain multiple healthy response bias samples. The multiple health response bias samples are sorted, and the median of the sorted sequence is taken as the baseline response bias. Calculate the absolute value of the difference between the response bias and the reference response bias, and divide the absolute value of the difference by the reference response bias to obtain the response distortion rate; When the response distortion rate is greater than a preset threshold, the response distortion rate is forcibly limited to the preset threshold. Multiply the constrained response distortion rate by 90° to obtain the intermediate angle value, then take the sine function of the intermediate angle value, and use the result of the sine function as the phase lag angle.

4. The method for fault early warning analysis of GIS gas treatment equipment based on the Internet of Things as described in claim 1, characterized in that, During the pressure holding phase, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative, and zero symbols. Based on the symmetry of the distribution of the consecutive lengths of positive and negative symbols in the symbolic sequence, the fluctuation symmetry is obtained, including: For the pressure value sequence arranged in chronological order during the pressure holding stage, the difference between each two adjacent pressure values ​​in the pressure value sequence is calculated sequentially to obtain the target difference sequence. Calculate the standard deviation of all pressure values ​​during the pressure holding phase, and use one-tenth of the standard deviation as the dynamic noise threshold; Iterate through each target difference in the target difference sequence. When the target difference is greater than the dynamic noise threshold, assign a positive sign to the target difference. When the target difference is less than the opposite of the dynamic noise threshold, assign a negative sign to the target difference. When the absolute value of the target difference is not greater than the dynamic noise threshold, assign a zero sign to the target difference. This results in a symbolic sequence of positive, negative, and zero signs arranged in chronological order. Consecutive identical and non-zero symbols in the symbolized sequence are merged into one run. When two adjacent non-zero runs have the same symbol type and are separated by a single zero symbol, the zero symbol is treated as noise and ignored, and the two adjacent non-zero runs are merged into one run. Obtain the symbol type and the number of consecutive occurrences of the symbol for each run after merging. Arrange the consecutive occurrences of the symbols for all runs after merging in order to obtain the run length sequence. Count the number of runs with odd lengths and the number of runs with even lengths in the run length sequence; When the number of even-numbered runs is zero, the fluctuation symmetry is set to the number of odd-numbered runs; otherwise, the fluctuation symmetry is set to the ratio of the number of odd-numbered runs to the number of even-numbered runs. When the difference between the number of odd-numbered runs and the number of even-numbered runs is less than one-tenth of their sum, the fluctuation symmetry is forcibly corrected to one.

5. The method for fault early warning analysis of GIS gas treatment equipment based on the Internet of Things according to claim 1, characterized in that, The calculation of the attenuation disturbance factor during the pressure holding stage based on the fluctuation symmetry, the pressure holding duration, and the pressure range includes: The time of the first sampling point during the pressure holding phase is taken as the start time, and the time of the last sampling point is taken as the end time. The result of subtracting the start time from the end time is taken as the pressure holding duration. Iterate through all pressure values ​​during the pressure holding stage, find the maximum and minimum pressure values, and take the difference between the maximum and minimum pressure values ​​as the pressure range. Determine whether the pressure holding duration is less than a preset minimum time threshold; If it is not less than the preset minimum time threshold, calculate the first square root of the pressure holding duration, multiply the fluctuation symmetry by the first square root to obtain the first product value, and then divide the first product value by the pressure range to obtain the attenuation disturbance factor. If the time is less than the preset minimum time threshold, the second square root of the minimum time threshold is calculated, the fluctuation symmetry is multiplied by the second square root to obtain the second product value, and then the second product value is divided by the pressure range to obtain the attenuation disturbance factor.

6. The method for fault early warning analysis of GIS gas treatment equipment based on the Internet of Things according to claim 1, characterized in that, Using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension, the joint feature points for the current operating cycle are constructed as follows: Establish a two-dimensional coordinate system, with the phase lag angle as the horizontal coordinate value and the attenuation perturbation factor as the vertical coordinate value. The horizontal coordinate of the two-dimensional coordinate system is set to range from 0° to 90°, and the vertical coordinate is set to range from zero to a preset upper limit value. Using the origin as a reference point, a target point is determined in the two-dimensional coordinate system. The horizontal coordinate of the target point is the phase lag angle, and the vertical coordinate is the attenuation disturbance factor. The target point is used as the joint feature point of the current operation cycle.

7. The method for fault early warning analysis of GIS gas treatment equipment based on the Internet of Things according to claim 1, characterized in that, The process of obtaining the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculating the deviation of the joint feature points relative to the feature point cloud distribution boundary, and determining whether a fault exists and the type of fault based on the deviation degree and direction includes: After the equipment is installed, debugged and continuously operated for several normal operating cycles, the phase lag angle and attenuation disturbance factor of each normal operating cycle are extracted to obtain multiple health joint feature points. The robust covariance matrix and robust mean vector of all healthy joint feature points are calculated using the minimum covariance determinant estimation method. The robust covariance matrix is ​​subjected to eigenvalue decomposition to obtain a first eigenvector, a second eigenvector, a first eigenvalue, and a second eigenvalue, wherein the first eigenvector corresponds to the first eigenvalue, the second eigenvector corresponds to the second eigenvalue, and the first eigenvalue is greater than or equal to the second eigenvalue. The direction of the first eigenvector is taken as the direction of the major axis of the ellipse, the direction of the second eigenvector is taken as the direction of the minor axis of the ellipse, and twice the square root of the first eigenvalue is taken as the radius of the major axis of the ellipse, and twice the square root of the second eigenvalue is taken as the radius of the minor axis of the ellipse. Using the robust mean vector as the center of the ellipse, a minimum bounding ellipse is constructed by the major axis direction, the minor axis direction, the major axis radius, and the minor axis radius, and the interior and boundary regions of the minimum bounding ellipse are used as the feature point cloud distribution boundary; Obtain the center coordinates, major axis radius, and minor axis radius of the minimum enclosing ellipse, and calculate the difference vector between the current joint feature point and the ellipse center of the minimum enclosing ellipse; Calculate the first projection length of the difference vector along the major axis of the ellipse and the second projection length along the minor axis of the ellipse, respectively, and determine the first ratio between the first projection length and the radius of the major axis, and the second ratio between the second projection length and the radius of the minor axis. The normalized distance of the difference vector in the scale space of the minimum enclosing ellipse is determined based on the first ratio and the second ratio. Calculate the geometric mean of the major axis radius and the minor axis radius, define the ratio of the normalized distance to the geometric mean as the normalized outlier, and determine whether the normalized outlier is greater than a preset discrete threshold. If the normalized outlier is not greater than the preset outlier threshold, then it is determined that there is no fault in the current operation cycle. If the normalized outlier is greater than the preset outlier threshold, the deviation azimuth angle of the current joint feature point is calculated with the major axis of the minimum enclosing ellipse as the reference direction. The fault type is matched from the fault angle partitioning table according to the deviation azimuth angle. The fault angle partitioning table adopts non-uniform partitioning, and the angle span of each partition is inversely proportional to the historical frequency of the fault type. The fault angle partitioning table includes at least the following: [0°, 30°) corresponding to valve opening delay fault, [30°, 70°) corresponding to air passage blockage fault, [70°, 120°) corresponding to air passage leakage fault, [120°, 150°) corresponding to valve closing delay fault, [150°, 200°) corresponding to exhaust valve jamming fault, [200°, 250°) corresponding to insufficient air source pressure fault, [250°, 300°) corresponding to seal aging fault, and [300°, 360°) corresponding to sensor drift fault.

8. A fault early warning and analysis system for GIS gas treatment equipment based on the Internet of Things, characterized in that, include: The acquisition module is configured to collect gas pressure waveform data of the GIS gas treatment equipment in real time during the current operating cycle through the Internet of Things platform, and simultaneously collect the valve opening command time and valve closing command time. The interception module is configured to intercept the target waveform segment from the moment the valve is opened to the moment the pressure first reaches the preset pressure value in the gas pressure waveform data, and arrange the pressure values ​​in the target waveform segment into a response sequence in chronological order, and extract the directional bias feature of the pressure value change in the response sequence to obtain the response bias degree. The mapping module is configured to obtain the phase hysteresis angle through nonlinear mapping based on the difference between the response bias and the baseline response bias in the healthy state. The definition module is configured to define the time period from the moment when the pressure first reaches the preset pressure value to the moment when the valve is closed in the gas pressure waveform data as the pressure holding stage. During the pressure holding stage, the pressure difference between two adjacent sampling points is symbolized to obtain a symbolic sequence composed of positive, negative and zero symbols. The fluctuation symmetry is obtained based on the distribution symmetry of the length of consecutive occurrence of positive and negative symbols in the symbolic sequence. The calculation module is configured to obtain the holding time of the holding stage, and to use the difference between the maximum pressure value and the minimum pressure value within the holding stage as the pressure range, and to calculate the attenuation disturbance factor of the holding stage based on the fluctuation symmetry, the holding time, and the pressure range. The construction module is configured to construct joint feature points for the current operating cycle using the phase lag angle as the first dimension and the attenuation perturbation factor as the second dimension. The determination module is configured to obtain the pre-stored feature point cloud distribution boundary of the GIS gas treatment equipment in a healthy state, calculate the degree of deviation of the joint feature points relative to the feature point cloud distribution boundary, and determine whether there is a fault and the fault type based on the degree and direction of deviation.

9. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method according to any one of claims 1 to 7.

Images