Intelligent fault detection method and system for surface acoustic wave filter

Abstract

The invention relates to the technical field of fault detection, and comprises an intelligent fault detection method and system for a surface acoustic wave filter, and the method comprises the following steps: collecting an input signal and an output signal of the surface acoustic wave filter, extracting an independent frequency band, carrying out the statistics of the input and output power of the frequency band, calculating a power ratio, and obtaining a multi-frequency-band energy ratio sequence. According to the method, the fixed bandwidth of input and output signals is decomposed, the energy characteristic analysis of different frequency bands is refined, the abnormal signal capturing capacity is improved, the judgment of the power attenuation rate is optimized by introducing a dynamic threshold value, the recognition adaptability and accuracy are improved, and meanwhile, the recognition accuracy is improved through comprehensive impedance characteristic analysis and change rate monitoring. According to the method, the identification capability of sudden faults is enhanced, energy attenuation and impedance abrupt change points are jointly analyzed, a multi-dimensional fault classification system is constructed, the accuracy of fault identification and effective judgment of complex faults are ensured, and the stability and applicability of overall fault diagnosis of the surface acoustic wave filter are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and particularly to an intelligent fault detection method and system for surface acoustic wave filters. Background Art

[0002] The technical field of fault detection includes the monitoring of the operating states of various devices, components or systems and the identification of abnormalities, aiming to realize the identification, classification and prediction of potential faults through data acquisition, analysis and processing means, mainly involving core links such as signal acquisition, feature extraction, and abnormality judgment. Among them, signal acquisition usually relies on a variety of sensors to monitor the operating state of the device in real time, feature extraction is to extract fault-related information through methods such as mathematical transformation and pattern recognition, and abnormality judgment usually combines threshold setting, statistical analysis and other methods to identify abnormal device operation. With the development of intelligent technology, fault detection technology has gradually integrated data analysis and modeling means to form a more efficient and accurate intelligent diagnosis system, which is widely used in many fields such as mechanical equipment, power systems, and communication equipment.

[0003] Among them, the intelligent fault detection method for surface acoustic wave filters refers to aiming at the abnormal situations that may occur during the operation of surface acoustic wave filters, based on the collected electrical signal parameters, using mathematical operations and data analysis methods to construct a classification and determination mechanism to realize the intelligent discrimination of fault states. First, the input and output signals of the filter are sampled, and key feature parameters are extracted by means of time series analysis or feature comparison. Subsequently, according to specific classification criteria, the feature parameters are classified and analyzed by means of mapping calculation or statistical modeling to determine whether there are abnormal situations, and the trend of the surface acoustic wave filter operation data is inferred to improve the accuracy of the determination.

[0004] In the existing fault detection process of surface acoustic wave filters, during signal monitoring, usually single or low-dimensional data is used for analysis, resulting in insufficient ability to capture the refinement of energy changes and easily ignoring local abnormal trends. The analysis mode of power attenuation is mainly based on the overall trend, lacking the refinement processing of specific frequency bands, making it difficult to accurately identify some early attenuation signals. Impedance measurement mainly focuses on the individual changes of input or output impedance, and fails to consider the coupling relationship between input and output impedance, resulting in limited ability to locate internal faults. The identification method of mutation points relies on static threshold setting and is difficult to adapt to the influence of different environmental factors, so that some key fault signals may be ignored due to noise interference. The classification method of fault modes lacks in-depth analysis of the distribution of mutation points and is difficult to effectively distinguish periodic interference from random faults, affecting the accuracy of the classification results. Due to the lack of cross-analysis means for multi-dimensional data, the existing methods are difficult to effectively establish an accurate diagnostic model in the case of coexistence of complex faults, resulting in unstable identification effects for some composite faults. Summary of the Invention

[0005] The object of the present invention is to solve the disadvantages existing in the prior art, and to propose an intelligent fault detection method for surface acoustic wave filters.

[0006] To achieve the above object, the present invention adopts the following technical solutions: An intelligent fault detection method for surface acoustic wave filters, comprising the following steps: S1: Collect the input signal and output signal of the surface acoustic wave filter, extract the independent frequency bands, statistically calculate the input and output power of the frequency bands, calculate the power ratio, and obtain a multi-band energy ratio sequence; S2: Based on the multi-band energy ratio sequence, calculate the change in energy ratio between adjacent frequency bands, judge the local mutation mode, extract the power attenuation rate per unit time, analyze whether the attenuation rate exceeds the limit, and obtain the energy attenuation mode recognition result; S3: Collect the input and output impedance of the electrodes of the surface acoustic wave filter, synchronously calculate the coupled equivalent value of the input and output impedance, and obtain the coupled impedance data; S4: Based on the coupled impedance data, calculate the change rate of the input impedance, output impedance, and coupled impedance per unit time respectively, screen the time points where the change rate of the impedance suddenly increases, judge the distribution characteristics of the mutation points, and obtain the impedance mutation point recognition result.

[0007] As a further solution of the present invention, the multi-band energy ratio sequence includes the input power of each frequency band, the output power of each frequency band, and the power ratio of each frequency band. The energy attenuation mode recognition result includes a normal state, an abnormal attenuation state, a filter fault mode, and a natural attenuation of performance. The coupled impedance data includes the input impedance, the output impedance, and the coupled equivalent value of the input and output impedance. The impedance mutation point recognition result includes the input impedance change rate mutation point, the output impedance change rate mutation point, and the coupled impedance change rate mutation point.

[0008] As a further solution of the present invention, the specific steps of S1 are as follows: S111: Collect the input signal and output signal of the surface acoustic wave filter, perform fixed-bandwidth decomposition on the input signal and output signal, call a set of multiple center frequencies, perform frequency division processing on the signal, obtain the input signal and output signal of each independent frequency band, store the extracted signal power data, and obtain the input power and output power data sets of each frequency band; S112: Based on the input power and output power data sets of each frequency band, extract the corresponding frequency band data of the input power and output power, and use the formula: ; Calculate the power ratio of the th frequency band, store and obtain the power ratio sequence of each frequency band, where Represents the output power of the frequency band, represents the average output power of all frequency bands, represents the input power of the frequency band, represents the average input power of all frequency bands; S113: According to the power ratio sequences of the respective frequency bands, perform statistical analysis, integrate the power ratio data of all frequency bands, and obtain a multi - frequency band energy ratio sequence.

[0009] As a further solution of the present invention, the specific steps of S2 are as follows: S211: Based on the multi - frequency band energy ratio sequence, calculate the change in energy ratio between adjacent frequency bands, set an energy ratio change threshold, compare the energy ratio change threshold to screen the frequency bands with abnormal energy ratio changes, and obtain a set of frequency bands with abnormal energy ratio changes; S212: Based on the set of frequency bands with abnormal energy ratio changes, judge the local mutation mode. If the energy ratios of all frequency bands are stable, mark the normal state. Otherwise, judge the distribution of abnormal frequency bands, call the energy ratio change threshold, compare the energy ratio over - limit conditions of each frequency band. If the energy ratios of some frequency bands exceed the threshold, mark the abnormal attenuation state and record the positions of abnormal frequency bands to obtain abnormal frequency band marking data; S213: Based on the abnormal frequency band marking data, extract the power attenuation rate within a unit time, and use the formula: ; Calculate the power attenuation rate of the frequency band , combine it with the upper limit value of the power attenuation rate, compare and screen the over - limit frequency bands. If the attenuation rate is over - limit, mark the filter failure mode. Otherwise, mark the natural attenuation of performance, and establish the recognition result of the energy attenuation mode, where represents the output power of the frequency band at time , represents the output power of the frequency band at the previous time point , represents the time interval.

[0010] As a further solution of the present invention, the specific steps of S3 are as follows: S311: Collect the impedance of the electrode input end and the impedance of the electrode output end of the surface acoustic wave filter, record the impedance data, and store it according to the time series to obtain an input - output impedance sequence; S312: Based on the input - output impedance sequence, use the complex forms of the input impedance and the output impedance, and use the formula: ; Calculate and obtain the coupled equivalent impedance value , where represents the complex value of the impedance at the electrode input end, represents the complex value of the impedance at the electrode output end, represents the phase difference between the input and output impedances; S313: Based on the coupled equivalent impedance value, establish a time series of the coupled equivalent impedance, and combine it with the input-output impedance sequence to organize and obtain the coupled impedance data.

[0011] As a further solution of the present invention, the specific steps of S4 are as follows: S411: Based on the coupled impedance data, perform numerical differential calculation, call the impedance data at adjacent time points, calculate the change rates of the input, output, and coupled impedances in the time dimension, and obtain a sequence of impedance change rates; S412: Based on the sequence of impedance change rates, combine a preset impedance mutation threshold, compare the impedance change rates at each time point, screen the time points with sudden increases in the impedance change rate, and establish a dataset of mutation time points; S413: Based on the dataset of mutation time points, analyze the distribution characteristics of the mutation points, calculate the mutation point intervals, distribution densities, and aggregation trends within the time series, store the calculated distribution characteristic data, and obtain the impedance mutation point recognition result.

[0012] As a further solution of the present invention, the method further includes S5; S5: Based on the impedance mutation point recognition result and the energy attenuation mode recognition result, analyze the frequency band corresponding to the mutation point. If the mutation point occurs in the low frequency and the impedance change rate continues to increase, mark the material aging fault mode. If the mutation point appears in a specific frequency band and shows periodic changes, mark the environmental interference fault mode. If the mutation point is irregularly distributed in multiple frequency bands and is accompanied by power attenuation, mark the electrode structure damage fault mode, and comprehensively obtain the classification result of the surface acoustic wave filter fault mode.

[0013] As a further solution of the present invention, the classification result of the surface acoustic wave filter fault mode includes the material aging fault mode, the environmental interference fault mode, and the electrode structure damage fault mode.

[0014] As a further solution of the present invention, the specific steps of S5 are as follows: S511: Based on the impedance mutation point recognition result and the energy attenuation mode recognition result, extract the mutation point data and the frequency band energy attenuation information, analyze the frequency band position of the mutation point, and extract the frequency range where the mutation point occurs to obtain the data of the frequency band corresponding to the mutation point; S512: Based on the data corresponding to the frequency band of the mutation point, judge the distribution characteristics of the mutation point, combine the impedance change rate sequence, analyze the distribution law of the mutation point in terms of position and time, identify the change trend of the mutation point, screen the concentration of mutation points within the frequency band, and obtain the mutation point distribution pattern; S513: Based on the mutation point distribution pattern, combined with the frequency band characteristics where the mutation point occurs, classify and judge the fault mode. If the mutation point occurs in the low frequency and the impedance change rate continues to increase, mark the material aging fault mode. If the mutation point appears in a specific frequency band and shows periodic changes, mark the environmental interference fault mode. If the mutation point is irregularly distributed in multiple frequency bands and is accompanied by power attenuation, mark the electrode structure damage fault mode. Comprehensively obtain the classification result of the surface acoustic wave filter fault mode.

[0015] An intelligent fault detection system for a surface acoustic wave filter, comprising: The signal energy ratio calculation module collects the input signal and the output signal, performs fixed bandwidth decomposition, extracts independent frequency bands, calculates the input power, output power and power ratio, and generates a multi-frequency band energy ratio sequence; The energy attenuation mode recognition module calculates the change of the energy ratio between adjacent frequency bands based on the multi-frequency band energy ratio sequence, screens abnormal frequency bands, extracts the power attenuation rate, compares with the upper limit value and marks faults and natural attenuation, and generates an energy attenuation mode recognition result; The impedance change rate calculation module collects the input impedance and the output impedance, calculates the coupled equivalent value of the input and output impedances, and generates coupled impedance data; The impedance mutation point screening module calculates the change rates of the input impedance, output impedance and coupled impedance based on the coupled impedance data, compares with a preset threshold, screens mutation points, determines the distribution characteristics, and generates an impedance mutation point recognition result; The fault mode classification module analyzes the mutation point frequency band and distribution characteristics based on the impedance mutation point recognition result and the energy attenuation mode recognition result, marks the fault mode, and generates a classification result of the surface acoustic wave filter fault mode.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by performing fixed-bandwidth decomposition on the input signal and the output signal, independent frequency bands with multiple fixed center frequencies are extracted to ensure accurate analysis of the energy characteristics of different frequency bands. Based on the calculation of the power ratio of each frequency band, a multi-band energy ratio sequence is constructed, and combined with the ratio change of adjacent frequency bands, local energy mutation patterns are identified to enhance the ability to capture abnormal signals. The calculation of the power attenuation rate introduces a dynamic threshold determination method, making the recognition of the energy attenuation pattern more adaptable and effectively distinguishing normal fluctuations from abnormal attenuation. By synchronously collecting the impedance data at the input and output ends of the electrode and calculating the coupling impedance, and comprehensively analyzing the impedance change characteristics, the internal state monitoring of the device is made more comprehensive. The change rates of the input impedance, output impedance, and coupling impedance are used to screen for mutation points, and combined with the distribution characteristics of the mutation points, the ability to identify sudden faults is improved. Based on the joint analysis of the energy attenuation pattern and impedance mutation points, a multi-dimensional fault classification system is constructed to ensure accurate recognition of different fault modes, and combined with the timing characteristics of the mutation points, the ability to judge complex faults is enhanced. Through multi-parameter collaborative analysis, misjudgments caused by a single data source are avoided, and the stability and applicability of fault detection are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is the main process flow chart of the present invention; Figure 2 is the process flow chart of step S1 of the present invention; Figure 3 is the process flow chart of step S2 of the present invention; Figure 4 is the process flow chart of step S3 of the present invention; Figure 5 is the process flow chart of step S4 of the present invention; Figure 6 is the process flow chart of step S5 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0020] Please refer to Figure 1 , an intelligent fault detection method for a surface acoustic wave filter, comprising the following steps: S1: Collect the input signal and output signal of the surface acoustic wave filter, perform fixed-bandwidth decomposition on the input signal and output signal, extract independent frequency bands with multiple fixed center frequencies, statistically calculate and store the input power and output power of each frequency band, calculate the power ratio of each frequency band, and obtain a multi-band energy ratio sequence; S2: Based on the multi-band energy ratio sequence, calculate the change in the energy ratio between adjacent frequency bands, set an energy ratio change threshold, screen the frequency bands with abnormal energy ratio changes, judge the local mutation mode. If the energy ratios of all frequency bands are stable, mark the normal state. If the energy ratios of some frequency bands exceed the energy ratio change threshold, mark the abnormal attenuation state and record the position of the abnormal frequency band, extract the power attenuation rate per unit time, set an upper limit value for the power attenuation rate. If the attenuation rate exceeds the limit, mark the filter fault mode, otherwise mark the natural attenuation of performance, and obtain the energy attenuation mode recognition result; S3: Collect the impedance at the electrode input end and the impedance at the electrode output end of the surface acoustic wave filter, store and obtain the impedance sequence, and synchronously calculate the coupled equivalent value of the input and output impedances to obtain the coupled impedance data; S4: Based on the coupled impedance data, calculate the input impedance change rate, output impedance change rate, and coupled impedance change rate per unit time, compare with the preset impedance mutation threshold, screen the time points with sudden increases in the impedance change rate, and judge the distribution characteristics of the mutation points to obtain the impedance mutation point recognition result; S5: Based on the impedance mutation point recognition result and the energy attenuation mode recognition result, analyze the frequency bands corresponding to the mutation points. If the mutation point occurs at a low frequency and the impedance change rate continues to increase, mark the material aging fault mode. If the mutation point appears in a specific frequency band and shows a periodic change, mark the environmental interference fault mode. If the mutation points are irregularly distributed in multiple frequency bands and are accompanied by power attenuation, mark the electrode structure damage fault mode, and comprehensively obtain the surface acoustic wave filter fault mode classification result.

[0021] The multi - band energy ratio sequence includes the input power of each band, the output power of each band, and the power ratio of each band. The energy attenuation mode recognition result includes the normal state, the abnormal attenuation state, the filter failure mode, and the natural performance attenuation. The coupled impedance data includes the input impedance, the output impedance, and the coupled equivalent value of the input - output impedance. The impedance mutation point recognition result includes the mutation point of the input impedance change rate, the mutation point of the output impedance change rate, and the mutation point of the coupled impedance change rate. The classification result of the surface acoustic wave filter failure mode includes the material aging failure mode, the environmental interference failure mode, and the electrode structure damage failure mode.

[0022] Please refer to Figure 2 , step S1 is as follows: S111: Collect the input signal and output signal of the surface acoustic wave filter, perform fixed - bandwidth decomposition on the input signal and output signal, call a set of multiple center frequencies, perform frequency - division processing on the signal to obtain the input signal and output signal of each independent band, store the extracted signal power data, and obtain the input power and output power data sets of each band; When collecting the input signal and output signal of the surface acoustic wave filter, it is necessary to synchronously measure the input end and output end of the filter through a signal acquisition device (such as an oscilloscope or a data acquisition card). The measured signal data should include the time - domain waveform and power - spectrum information. In practical applications, for example, for a surface acoustic wave filter with a center frequency of 2.4 GHz, the sampling rate of the collected signal usually needs to be much higher than this frequency, such as above 10 GHz, to ensure accurate acquisition of signal characteristics. Then, perform fixed - bandwidth decomposition on the input signal and output signal. Usually, a band - pass filter bank is used. The center - frequency interval of this filter bank can be set according to application requirements, such as 10 MHz, 50 MHz, or 100 MHz. The signal data of each bandwidth is stored separately, and the power - spectrum density of each band is calculated based on FFT (Fast Fourier Transform) to obtain the input signal and output signal of each independent band. For a specific band, such as the 2.45 GHz ± 10 MHz band, a band - pass filter can be used to extract the signal of this band, and the root - mean - square value (RMS) of its time - domain is calculated to characterize the signal power. Store the extracted signal power data to obtain the input power and output power data sets of each band.

[0023] S112: Based on the input power and output power data sets of each band, extract the corresponding - band data of the input power and output power, and use the formula: ; Calculate the power ratio of the th band , and store to obtain the power ratio sequence of each band. Among them, represents the output power of the th band, Represents the average output power of all frequency bands, Represents the Input power of the frequency band, Represents the average input power of all frequency bands; Based on the input power and output power data sets of each frequency band, calculate the power ratio of each frequency band. First, extract the input power of each frequency band from the aforementioned stored data set And the output power , then calculate the average value of the input power of all frequency bands And the average value of the output power , set a specific example. Assume that the system has tested 5 frequency bands, and the input power and output power data are as follows in Table 1: Table 1 Input and output power data collected: ; Calculate the average values of the input power and the output power: ; ; Then, calculate the power ratio of each frequency band , taking the 3rd frequency band as an example: ; Similarly calculate the Values of all frequency bands, and finally store the calculation results to establish a power ratio sequence for each frequency band.

[0024] S113: According to the power ratio sequence of each frequency band, conduct statistical analysis, integrate the power ratio data of all frequency bands, and obtain a multi-band energy ratio sequence; Call the power ratio sequence of each frequency band and conduct statistical analysis on the sequence. First, calculate the mean and standard deviation of the power ratio for all frequency bands to judge the stability of power transmission in each frequency band. Assume that the calculated ratio sequence is as follows: ; Calculate the mean: ; Calculate the standard deviation: ; ; Through this calculation, the stability of the power ratio of different frequency bands can be evaluated. If the ratio of a certain frequency band deviates greatly from the mean value (such as ), there may be an abnormal signal loss situation. Store the finally calculated ratio sequence to obtain a multi-band energy ratio sequence.

[0025] Please refer to Figure 3 , the steps of S2 are: S211: Based on the multi-band energy ratio sequence, calculate the change in the energy ratio between adjacent frequency bands, set the energy ratio change threshold, and compare the energy ratio change threshold to screen for frequency bands with abnormal energy ratio changes, obtaining a set of frequency bands with abnormal energy ratio changes; Based on the multi-band energy ratio sequence, extract the energy ratio data of adjacent frequency bands, for each frequency band and its adjacent frequency band for comparison, calculate the energy ratio change amount, set the energy ratio change threshold, and screen for frequency bands with abnormal energy ratio changes. In practical applications, after a filter of a certain device has been working for a long time, the energy ratios of its different frequency bands may change due to external interference or component aging. For example, in the three frequency bands of 10 MHz, 20 MHz, and 30 MHz, if their energy ratios are 0.85, 0.92, and 1.1 respectively, calculate the ratio change amount between adjacent frequency bands as and . If the preset energy ratio change threshold value is based on the long-term monitoring data of the device, and the standard deviation of the energy ratio change is statistically , according to the setting rule of the energy ratio change threshold = 3 times the standard deviation, then the energy ratio change threshold = 3 x 0.05 = 0.15. If the calculated ratio change amount exceeds 0.15, it is determined that the frequency band has an abnormal change. For example, the change amount of the 30 MHz frequency band, 0.18, exceeds 0.15, and it is marked as an abnormal change frequency band.

[0026] Table 2 Energy ratio and change amount of adjacent frequency bands: ; As shown in Table 2, the energy ratio has a large mutation in the 30 MHz frequency band. Therefore, it is necessary to further analyze its abnormal situation. Call the set of frequency bands with abnormal energy ratio changes, judge the local mutation mode. If the energy ratio change amounts of all frequency bands are lower than the threshold, mark the normal state; otherwise, further analyze the distribution of abnormal frequency bands, and compare and judge the abnormal category based on the change trend.

[0027] S212: Based on the set of frequency bands with abnormal energy ratio changes, judge the local mutation mode. If the energy ratios of all frequency bands are stable, mark the normal state; otherwise, judge the distribution of abnormal frequency bands, call the energy ratio change threshold, compare the energy ratio exceeding limit conditions of each frequency band. If the energy ratios of some frequency bands exceed the threshold, mark the abnormal attenuation state, and record the positions of abnormal frequency bands, obtaining abnormal frequency band marking data; Based on the set of frequency bands with abnormal changes in energy ratio, detect the spatial distribution of the abnormally changing frequency bands, cluster multiple abnormal frequency bands, and determine whether a local mutation pattern is formed. In the example, if the abnormal frequency bands are mainly concentrated in the high-frequency region (for example, the frequency bands of 30 MHz, 32 MHz, and 35 MHz all exceed the threshold), it may indicate that a certain component (such as a capacitor or resonator) of the filter is aging. During this process, use the position data of each abnormal frequency band and its adjacent abnormal frequency bands to calculate the clustering coefficient: ; The setting of this clustering coefficient is based on the device bandwidth characteristics and the circuit resonance frequency distribution. Usually, the value range should be less than the passband width of the filter. For example, for a certain filter, its passband width is 10 MHz. Therefore, the threshold is set to . If the calculation result is lower than this value, it is considered that the frequency band clustering is abnormal, otherwise it is regarded as a random abnormality.

[0028] Calculation example: If the abnormal frequency bands are 30 MHz, 32 MHz, and 35 MHz respectively, then calculate: ; Since 5 MHz is less than the set threshold of 10 MHz, it indicates that the abnormal frequency band clustering phenomenon is obvious and there is a local mutation pattern. If it exceeds 10 MHz, it is judged as a discrete abnormality and finally marked as a local mutation pattern.

[0029] S213: Based on the marked data of abnormal frequency bands, extract the power attenuation rate per unit time, using the formula: ; Calculate the power attenuation rate of the frequency band, combine it with the upper limit value of the power attenuation rate, compare and screen out the over-limit frequency bands. If the attenuation rate exceeds the limit, mark the filter failure mode, otherwise mark the natural attenuation of performance, and establish the energy attenuation mode recognition result, where represents the output power of the frequency band at time , represents the output power of the frequency band at the previous time point , represents the time interval; Based on the marked data of abnormal frequency bands, extract the power attenuation rate per unit time, set the upper limit value of the power attenuation rate, and compare and screen out the over-limit frequency bands. Suppose the power measurement data of a certain device is as follows: Table 3 Power attenuation rate calculation data: ; As shown in Table 3, calculations are performed using formulas. For example, for the 32 MHz frequency band: ; Upper limit value of power attenuation rate Depends on the power stability parameter of the device and the maximum allowable attenuation rate, usually based on the maximum change range of long-term monitoring and time window , calculate . If the calculated attenuation rate exceeds 0.7 dB / s, mark this frequency band as the filter fault mode, and the remaining frequency bands are within the normal range and marked as natural attenuation of performance.

[0030] This result indicates that the power attenuation rate of the 32 MHz frequency band exceeds the set threshold, and there may be internal losses or component damage in the filter. Although there is a certain attenuation in the 30 MHz frequency band, it is still within the normal range, so no further processing is required. The attenuation rate of the 35 MHz frequency band is low, indicating stable signal energy. Finally, an energy attenuation mode recognition result is established for subsequent device status monitoring.

[0031] Please refer to Figure 4 , the steps of S3 are as follows: S311: Collect the input impedance and output impedance of the electrodes of the surface acoustic wave filter, record the impedance data, and store it in time series to obtain the input-output impedance sequence; Collect the input impedance and output impedance of the electrodes of the surface acoustic wave filter. First, connect a high-precision impedance analyzer to the input and output ends of the filter respectively to measure the signal. During the measurement, to ensure the stability and repeatability of the impedance data, a scanning frequency range of 10 kHz to 500 MHz is used, and the impedance values are gradually obtained in steps of 1 kHz. During the measurement, due to the influence of small signal noise, 5 repeated measurements are required and their mean value is calculated to reduce random errors. The measured impedance data includes amplitude and phase information. For example, at a specific measurement point (such as at a frequency of 100 MHz), the impedance measured at the input end is Ω, and the impedance measured at the output end is Ω. To ensure the systematicness of the data, the data of all measurement points are stored in the database and recorded in time series. Each record includes the timestamp, frequency, input impedance, output impedance, and the average value of the number of measurements, as shown in Table 4. Finally, the input-output impedance sequence is obtained.

[0032] Table 4 Input-output impedance data table collected: ; As shown in Table 4, the collected impedance data contains input and output impedance information at multiple frequency points, and the mean value is taken for each data point after multiple measurements to ensure accuracy.

[0033] S312: Based on the input and output impedance sequences, using the complex forms of the input impedance and the output impedance, the formula: ; Calculate to obtain the coupled equivalent impedance value , where represents the complex value of the impedance at the electrode input end, represents the complex value of the impedance at the electrode output end, represents the phase difference between the input and output impedances; Based on the input and output impedance sequences, calculate the coupled equivalent value of the input and output impedances. First, extract the modulus values of the input impedance and the output impedance at the corresponding frequency points, that is and , and the phase angle can be calculated from the phase difference between the input and output impedances. At the 100 MHz frequency point, assuming the input impedance Ω, its modulus value is calculated as follows: ; Similarly, for the output impedance Ω, its modulus value is calculated as follows: ; Calculate the phase angle , that is, the phase difference between the input impedance and the output impedance: ; ; Substitute into the formula to calculate the coupled equivalent impedance: ; ; ; Finally, obtain the coupled equivalent impedance value. The calculation result shows that at the 100 MHz frequency point, the coupled equivalent impedance value of the input and output impedances is 7.31 Ω, and this value is used for subsequent impedance analysis to further study the impedance matching characteristics of the filter.

[0034] S313: Based on the coupled equivalent impedance value, establish a time series of the coupled equivalent impedance, and combine it with the input and output impedance sequences to organize and obtain the coupled impedance data; Based on the coupled equivalent impedance value, establish an impedance time series and store it in combination with the input and output impedance sequences. First, extract the coupled equivalent impedance values calculated at all measured frequency points and store them in chronological order, as shown in Table 5.

[0035] Coupled equivalent impedance data calculated in Table 5: ; As shown in Table 5, the calculated coupled equivalent impedance data varies at different frequency points. Among them, the impedance value increases in the high-frequency case, which is caused by the influence of the amplitude and phase angle difference between the input and output impedances. Finally, the coupled impedance data is obtained for the matching analysis of the filter.

[0036] Please refer to Figure 5 , and the steps of S4 are as follows: S411: Based on the coupled impedance data, perform numerical differentiation calculations, call the impedance data at adjacent time points, calculate the change rates of the input, output, and coupled impedances in the time dimension, and obtain a sequence of impedance change rates; Based on the coupled impedance data, the impedance data is sourced from the impedance data records of the electrode input and output ends of the surface acoustic wave filter by a measuring device. This measurement process usually uses an impedance analyzer (such as Keysight E4990A or Wayne Kerr 6500B), and its measurement frequency range is usually between 1 MHz and 3 GHz. The specific measurement point interval can be set to 10 kHz to obtain sufficiently dense sampling point data. The acquired raw data includes the input impedance and the output impedance in the form of complex values. The real and imaginary parts corresponding to each measurement point are stored, and the storage format is stored in a matrix as shown in the following table: Table 6 Coupled impedance data acquisition table: ; After the data is stored, the numerical differentiation method is used to calculate the change rate per unit time of the measured input and output impedance data, that is, calculate the impedance change amount between adjacent time points and divide it by the measurement time interval. For example, for the adjacent measurement time interval , the calculation of the input impedance change rate is as follows: ; Among them, for the data in the table, the input impedance change rate is calculated at the frequency point of 20 MHz: ; Similarly, the output impedance change rate can be calculated. After the calculation is completed, the data is stored in the impedance change rate sequence.

[0037] S412: Based on the impedance change rate sequence, combined with a preset impedance mutation threshold, compare the impedance change rates at each time point, screen the time points with sudden increases in the impedance change rate, and establish a dataset of mutation time points; Based on the impedance change rate sequence, a preset threshold is used to compare the change rate to screen for mutation time points, and the preset threshold is set by referring to the standard deviation calculation method of the impedance change rate, and the standard deviation of all impedance change rates is calculated , and the threshold is set as the mean plus 2 times the standard deviation: ; Among them, represents the mean value of the impedance change rate, represents the standard deviation of the impedance change rate. Taking the data in Table 1 as an example, assuming its mean value is , substitute the value to calculate the standard deviation: ; Therefore, the mutation threshold: ; Screen out the time points greater than this threshold as the impedance mutation time points. For example, if the change rate of a certain measurement point reaches 0.26 Ω / ms, then mark it as a mutation point and store it in the mutation point data set.

[0038] S413: Based on the mutation time point data set, analyze the distribution characteristics of the mutation points, calculate the mutation point interval, distribution density, and aggregation trend within the time series, store the calculated distribution characteristic data, and obtain the impedance mutation point recognition result; Based on the mutation time point data set, analyze the time distribution characteristics of the mutation points, calculate the average time interval between the mutation points, and count the density of the mutation points within the unit time window. Specifically, the average value of the mutation time interval is calculated as follows: ; Among them, represents the time of the th mutation point, represents the total number of mutation points. Assuming the mutation points occur at 20 ms, 35 ms, and 55 ms respectively, then calculate the mutation time interval: ; For the normal working stable interval of the filter, the setting basis of this value comes from the equivalent circuit characteristics of the filter and its dynamic impedance change range under the working state. Specifically, the rate of change of the input and output impedance of the surface acoustic wave filter with frequency is affected by its resonance characteristics, Since in actual measurement, the minimum resolvable time of the instrument is usually at the microsecond level, for filters in the low-frequency range (such as 10 MHz - 100 MHz), the impedance stability interval is generally determined by the system's response time and measurement resolution. After actual measurement, this value usually falls around 50 ms. This value is affected by factors such as temperature drift, device aging, and electrode coupling characteristics, and usually fluctuates within the range of 40 ms - 60 ms.

[0039] If the calculated mutation interval is much smaller than the impedance stability interval of 50 ms, it can be judged as an abnormal state, and finally the impedance mutation point recognition result is formed.

[0040] Please refer to Figure 6 , and the steps of S5 are as follows: S511: Based on the impedance mutation point recognition result and the energy attenuation mode recognition result, extract the mutation point data and the frequency band energy attenuation information, analyze the frequency band position of the mutation point, extract the frequency range where the mutation point occurs, and obtain the data corresponding to the frequency band of the mutation point; Based on the impedance mutation point recognition result and the energy attenuation mode recognition result, first extract the time stamp and frequency band information corresponding to the impedance mutation point, construct an impedance change time series, record the specific time when the mutation point occurs and the frequency band distribution. To ensure the integrity of the data, the sampling time interval is set to 0.1 second, and the frequency band range where the mutation point appears is classified. Suppose in a certain measurement, the mutation points mainly occur in the frequency band range of 2.5 MHz to 3.2 MHz, and multiple consecutive mutations occur in the time period from 1.2 seconds to 1.5 seconds, then the system records these data and analyzes them in the subsequent steps. For the energy attenuation mode, call the energy ratio change data and compare the power attenuation trends of different frequency bands. For example, if the power of the 2.5 MHz frequency band drops by 3 dB during the period from 1.2 seconds to 1.5 seconds, while the 3.0 MHz frequency band drops by 2.8 dB, then record this information and conduct a joint analysis in combination with the mutation point information. According to the collected data, finally obtain the data corresponding to the frequency band of the mutation point.

[0041] S512: Based on the data corresponding to the frequency band of the mutation point, judge the distribution characteristics of the mutation point, combine the impedance change rate sequence, analyze the distribution law of the mutation point in terms of position and time, identify the change trend of the mutation point, screen the concentration of the mutation points within the frequency band, and obtain the mutation point distribution mode; Based on the data corresponding to the mutation points in the frequency band, first analyze the distribution pattern of the mutation points, including the time interval, the density of the mutation points, and the trend of the mutation points. Suppose in a certain measurement data, multiple mutation points occur between 1.2 seconds and 1.5 seconds, and the interval between each mutation point is about 0.05 seconds, then it can be determined that the mutation points have a certain time concentration. In addition, classify the distribution of the mutation points in each frequency band. For example, the interval of the mutation points in the 2.5 MHz frequency band is shorter, while the interval of the mutation points in the 3.0 MHz frequency band is longer, indicating that the impedance change patterns in different frequency bands are different. On this basis, calculate the trend of the impedance change rate and analyze whether the mutation points increase with time. For example, if the impedance change rate at 1.2 seconds is 0.8 Ω / s and rises to 1.4 Ω / s at 1.5 seconds, it indicates that the impedance change rate shows an increasing trend, which may be related to material aging. If the mutation points occur at periodic intervals (such as once every 0.5 seconds), there may be environmental interference. Through the above analysis, the distribution pattern of the mutation points is finally obtained.

[0042] S513: Based on the distribution pattern of the mutation points, combined with the frequency band characteristics where the mutation points occur, classify and judge the fault mode. If the mutation points occur in the low frequency band and the impedance change rate continues to increase, mark the material aging fault mode. If the mutation points appear in a specific frequency band and show periodic changes, mark the environmental interference fault mode. If the mutation points are irregularly distributed in multiple frequency bands and are accompanied by power attenuation, mark the electrode structure damage fault mode. Synthesize to obtain the classification result of the SAW filter fault mode; Based on the distribution pattern of the mutation points, combined with the frequency band characteristics where the mutation points occur, judge the possible fault mode. First, if the mutation points are concentrated in the low frequency band (for example, below 3 MHz) and the impedance change rate increases with time, it may be caused by material aging. For example, in a certain measurement, the impedance change rate in the 2.5 MHz frequency band rises from 0.8 Ω / s to 1.5 Ω / s, which conforms to the material aging mode. Second, if the mutation points appear in a specific frequency band (such as 3.8 MHz to 4.2 MHz) and show periodic changes, such as an impedance mutation occurring once every 0.5 seconds, it may be the result of environmental interference. In addition, if the mutation points are irregularly distributed in multiple frequency bands and are accompanied by power attenuation, such as mutations occurring in the 2.5 MHz, 3.2 MHz, and 3.8 MHz frequency bands, and the energy ratio drops by more than 2 dB at the same time, it may be caused by electrode structure damage. Through the above analysis, the classification result of the SAW filter fault mode is finally obtained.

[0043] Table 7: Table of the time interval and frequency band distribution of the mutation points: ; As shown in Table 7, the mutation points are relatively concentrated during the period from 1.2 seconds to 1.5 seconds, and the impedance change rate gradually increases with time, indicating that material aging may be the main factor leading to the mutation.

[0044] An intelligent fault detection system for surface acoustic wave filters, comprising: The signal energy ratio calculation module collects the input signal and the output signal, performs fixed bandwidth decomposition, extracts independent frequency bands, calculates the input power, the output power and the power ratio, and generates a multi-band energy ratio sequence; The energy attenuation mode recognition module calculates the change of the energy ratio of adjacent frequency bands based on the multi-band energy ratio sequence, screens abnormal frequency bands, extracts the power attenuation rate, compares with the upper limit value and marks faults and natural attenuation, and generates an energy attenuation mode recognition result; The impedance change rate calculation module collects the input impedance and the output impedance, calculates the coupled equivalent value of the input and output impedances, and generates coupled impedance data; The impedance mutation point screening module calculates the change rates of the input impedance, the output impedance and the coupled impedance based on the coupled impedance data, compares with the preset threshold, screens mutation points, determines the distribution characteristics, and generates an impedance mutation point recognition result; The fault mode classification module analyzes the mutation point frequency band and distribution characteristics based on the impedance mutation point recognition result and the energy attenuation mode recognition result, marks the fault mode, and generates a surface acoustic wave filter fault mode classification result.

[0045] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent fault detection method for a surface acoustic wave filter, characterized in that: The following steps are involved: S1: Collect the input and output signals of the surface acoustic wave filter, extract independent frequency bands, count the input and output power of the frequency bands, calculate the power ratio, and obtain a multi-band energy ratio sequence; S2: Based on the multi-band energy ratio sequence, calculate the energy ratio change between adjacent frequency bands, determine the local mutation mode, extract the power attenuation rate per unit time, analyze whether the attenuation rate exceeds the limit, and obtain the energy attenuation mode recognition result; S3: collecting the impedance of the electrode input and output ends of the surface acoustic wave filter, synchronously calculating the coupling equivalent value of the input and output impedances, and obtaining coupling impedance data; S4: Based on the coupling impedance data, respectively calculate the change rates of the input impedance, output impedance and coupling impedance per unit time, select the time point at which the impedance change rate suddenly increases, determine the distribution characteristics of the mutation point, and obtain the impedance mutation point identification result.

2. The intelligent fault detection method for surface acoustic wave filter according to claim 1, characterized in that: The multi-band energy ratio sequence includes the input power of each frequency band, the output power of each frequency band, and the power ratio of each frequency band. The energy attenuation mode recognition result includes the normal state, the abnormal attenuation state, the filter failure mode, and the natural attenuation of performance. The coupled impedance data includes the input impedance, the output impedance, and the coupled equivalent value of the input and output impedances. The impedance mutation point recognition result includes the input impedance change rate mutation point, the output impedance change rate mutation point, and the coupling impedance change rate mutation point.

3. The intelligent fault detection method for surface acoustic wave filter according to claim 1, characterized in that: The specific steps of S1 are: S111: collecting input signals and output signals of the surface acoustic wave filter, and performing fixed bandwidth decomposition on the input signals and output signals, calling multiple set center frequencies, performing frequency division processing on the signals, obtaining input signals and output signals of each independent frequency band, storing the extracted signal power data, and obtaining input power and output power data sets of each frequency band; S112: Based on the input power and output power data sets of each frequency band, extract the corresponding frequency band data of the input power and the output power, using the formula: ； Calculate the Power ratio of frequency band , store the power ratio sequence of each frequency band, where, Representative The output power of the frequency band, Represents the average output power of all frequency bands, Representative The input power of the frequency band, Represents the average input power in all frequency bands; S113: Performing statistical analysis based on the power ratio sequences of each frequency band, integrating the power ratio data of all frequency bands, and obtaining a multi-frequency band energy ratio sequence.

4. The intelligent fault detection method for surface acoustic wave filter according to claim 1, characterized in that: The specific steps of S2 are: S211: Based on the multi-band energy ratio sequence, calculate the energy ratio change between adjacent frequency bands, set an energy ratio change threshold, and compare the energy ratio change threshold to screen frequency bands with abnormal energy ratio changes to obtain a frequency band set with abnormal energy ratio changes; S212: Based on the frequency band set with abnormal energy ratio changes, determine the local mutation mode. If the energy ratios of all frequency bands are stable, mark them as normal. Otherwise, determine the distribution of abnormal frequency bands, call the energy ratio change threshold, compare the energy ratio exceeding the limit of each frequency band, and if the energy ratio of some frequency bands exceeds the threshold, mark the abnormal attenuation state, and record the abnormal frequency band position to obtain abnormal frequency band marking data. S213: Based on the abnormal frequency band mark data, extract the power attenuation rate per unit time, using the formula: ； Calculate the Power attenuation rate of the frequency band , combined with the upper limit of the power attenuation rate, the frequency band that exceeds the limit is compared and screened. If the attenuation rate exceeds the limit, the filter failure mode is marked, otherwise the performance is marked as natural attenuation, and the energy attenuation mode recognition result is established, where, Representative Frequency band in time The output power, Representative Frequency band at the previous time point The output power, Represents a time interval.

5. The intelligent fault detection method for surface acoustic wave filter according to claim 1, characterized in that: The specific steps of S3 are: S311: collecting the impedance of the electrode input end and the impedance of the electrode output end of the surface acoustic wave filter, recording the impedance data, and storing them in time series to obtain an input and output impedance sequence; S312: Based on the input and output impedance sequence, the complex form of input impedance and output impedance is used, and the formula is used: ； Calculate the coupling equivalent impedance value ,in, represents the complex value of the impedance at the electrode input, represents the complex value of the impedance at the output of the electrode, Represents the phase difference between input and output impedance; S313: Based on the coupling equivalent impedance value, a coupling equivalent impedance time series is established, and the coupling impedance data is obtained by combining the input and output impedance series.

6. The intelligent fault detection method for surface acoustic wave filter according to claim 1, characterized in that: The specific steps of S4 are: S411: Based on the coupling impedance data, perform numerical differential calculation, call the impedance data of adjacent time points, calculate the change rates of input, output and coupling impedance in the time dimension, and obtain an impedance change rate sequence; S412: Based on the impedance change rate sequence and in combination with a preset impedance mutation threshold, the impedance change rate at each time point is compared, the time point at which the impedance change rate suddenly increases is selected, and a mutation time point data set is established; S413: Based on the mutation time point data set, analyze the mutation point distribution characteristics, calculate the mutation point interval, distribution density and aggregation trend in the time series, store the calculated distribution characteristic data, and obtain the impedance mutation point identification result.

7. The intelligent fault detection method for a surface acoustic wave filter according to claim 1, characterized in that: The method further comprises S5; S5: Based on the impedance mutation point identification results and the energy attenuation pattern identification results, the frequency band corresponding to the mutation point is analyzed. If the mutation point occurs at a low frequency and the impedance change rate continues to increase, the material aging failure mode is marked. If the mutation point appears in a specific frequency band and changes periodically, the environmental interference failure mode is marked. If the mutation point is irregularly distributed in multiple frequency bands and is accompanied by power attenuation, the electrode structure damage failure mode is marked, and the surface acoustic wave filter failure mode classification result is comprehensively obtained.

8. The intelligent fault detection method for a surface acoustic wave filter according to claim 7, characterized in that: The surface acoustic wave filter failure mode classification results include material aging failure mode, environmental interference failure mode, and electrode structure damage failure mode.

9. The intelligent fault detection method for a surface acoustic wave filter according to claim 7, characterized in that: The specific steps of S5 are: S511: based on the impedance mutation point identification result and the energy attenuation pattern identification result, extract the mutation point data and frequency band energy attenuation information, analyze the frequency band position of the mutation point, extract the frequency range where the mutation point occurs, and obtain the frequency band data corresponding to the mutation point; S512: Based on the frequency band data corresponding to the mutation point, determine the distribution characteristics of the mutation point, combine the impedance change rate sequence, analyze the distribution law of the mutation point position and time, identify the change trend of the mutation point, screen the concentration of the mutation points in the frequency band, and obtain the mutation point distribution pattern; S513: Based on the mutation point distribution pattern and the frequency band characteristics of the mutation point, the fault mode is classified and judged. If the mutation point occurs at a low frequency and the impedance change rate continues to increase, the material aging fault mode is marked. If the mutation point appears in a specific frequency band and changes periodically, the environmental interference fault mode is marked. If the mutation point is irregularly distributed in multiple frequency bands and is accompanied by power attenuation, the electrode structure damage fault mode is marked, and the surface acoustic wave filter fault mode classification result is obtained comprehensively.

10. An intelligent fault detection system for a surface acoustic wave filter, characterized in that: The system is used to execute the method according to any one of claims 1 to 9, comprising: The signal energy ratio calculation module collects input signals and output signals, performs fixed bandwidth decomposition, extracts independent frequency bands, calculates input power, output power and power ratio, and generates a multi-band energy ratio sequence; The energy attenuation pattern recognition module calculates the energy ratio changes of adjacent frequency bands based on the multi-band energy ratio sequence, screens abnormal frequency bands, extracts the power attenuation rate, compares the upper limit value and marks faults and natural attenuation, and generates an energy attenuation pattern recognition result; The impedance change rate calculation module collects input impedance and output impedance, calculates the coupling equivalent value of input and output impedance, and generates coupling impedance data; The impedance mutation point screening module calculates the change rates of the input impedance, the output impedance and the coupling impedance based on the coupling impedance data, compares the preset thresholds, screens the mutation points, determines the distribution characteristics, and generates the impedance mutation point identification results; The fault mode classification module analyzes the frequency band and distribution characteristics of the mutation point based on the impedance mutation point identification result and the energy attenuation mode identification result, marks the fault mode, and generates a surface acoustic wave filter fault mode classification result.

Images