Fan blade voiceprint monitoring method and system

Abstract

The application discloses a fan blade voiceprint monitoring method and system, and relates to the technical field of wind power equipment state monitoring.The method comprises the following steps: calculating a first synthetic force parameter according to a real-time fan rotating speed and a real-time rotating phase; calculating a first friction sound reference sequence corresponding to a magnetic attraction sensor inside a fan blade based on the first synthetic force parameter; acquiring a first mixed voiceprint sequence collected by the magnetic attraction sensor inside the fan blade; calculating a first difference parameter based on the difference between the first mixed voiceprint sequence and the first friction sound reference sequence; and outputting a fan blade crack early warning information when the amplitude of the first difference parameter is greater than a first preset amplitude.The application solves the problem that dynamic friction sound generated by loosening of the magnetic attraction sensor inside the fan blade is confused with real crack sound, which leads to inaccurate identification of blade cracks and failure to shut down for repair, and can dynamically strip the friction noise generated by loosening of the sensor, accurately extract residual crack acoustic components, and avoid false positives and false negatives.

Description

Technical Field

[0001] This application relates to the field of wind power equipment condition monitoring technology, and in particular to a method and system for monitoring the acoustic signature of wind turbine blades. Background Technology

[0002] Monitoring the health of wind turbine blades is crucial during the long-term operation of wind turbine generators. Current technology typically deploys acoustic sensors inside the blades to identify early cracks by collecting sound signals during blade operation. For ease of installation, these sensors are often magnetically attached to the lightning protection mesh on the inner wall of the blade.

[0003] As the wind turbine operates for an extended period, the magnetic force of the magnetic sensor inevitably weakens. During high-speed rotation of the wind turbine blades, the alternating effects of centrifugal force and gravity cause the magnetic sensor to slip slightly inside the blade, generating a low-frequency friction sound. This friction sound caused by sensor loosening highly overlaps in frequency distribution with the tearing sound caused by early micro-cracks in the fiberglass material of the blades, and the frequency and energy of the friction sound exhibit dynamic time-varying characteristics with changes in wind turbine speed and rotation phase.

[0004] Because wind turbines cannot be shut down to re-secure sensors, existing conventional static spectral subtraction methods are ineffective at separating friction noise from the dynamically changing friction sounds described above, leading to severe cross-contamination of acoustic characteristics. Static filtering not only fails to remove dynamic friction noise but also distorts or even loses the true crack sound signature, causing frequent false alarms or fatal missed alarms in the monitoring system, making it impossible for maintenance personnel to accurately determine the true health status of the blades. Summary of the Invention

[0005] In view of the aforementioned problems, this application is hereby filed.

[0006] Therefore, this application provides a method and system for monitoring the acoustic signature of wind turbine blades, which can solve the problem of the dynamic friction sound generated by the loose magnetic sensor inside the wind turbine blade being confused with the sound of real cracks, resulting in the inability to accurately identify blade cracks and the inability to stop the machine for repair.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: Firstly, this application provides a method for monitoring the acoustic signature of wind turbine blades, including: Obtain the real-time fan speed corresponding to the fan blades and the real-time rotation phase corresponding to the fan blades; The first composite force parameter is calculated based on the real-time fan speed and the real-time rotation phase, and the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the fan blade is calculated based on the first composite force parameter. Acquire the first mixed acoustic signature sequence collected by the magnetic sensor inside the wind turbine blade; The first difference parameter is obtained based on the difference between the first mixed acoustic pattern sequence and the first frictional sound reference sequence; When the amplitude of the first difference parameter is greater than the first preset amplitude, a wind turbine blade crack warning message is output.

[0008] Preferably, the step of calculating the first composite force parameter based on the real-time fan speed and the real-time rotation phase, and calculating the first frictional sound reference sequence corresponding to the magnetic attraction sensor inside the fan blade based on the first composite force parameter, includes: The ratio between the square of the real-time fan speed corresponding to the fan blade and the square of the reference speed corresponding to the fan blade is calculated and used as the first centrifugal force coefficient corresponding to the magnetic attraction sensor inside the fan blade. The product of the phase offset angle of the real-time rotating phase within the preset quadrant interval and the cosine of the reference quadrant angle is used as the first gravity compression coefficient. The sum of the first centrifugal force coefficient and the first gravity compression coefficient is used as the first synthetic force parameter, and the first frictional sound frequency offset is obtained based on the product of the first synthetic force parameter and the preset force-sound conversion ratio. Based on the sum of the first friction sound frequency offset and the reference friction sound frequency, a first friction sound reference sequence corresponding to the magnetic attraction sensor inside the wind turbine blade is constructed.

[0009] Preferably, the step of obtaining the first frictional sound frequency offset based on the product of the first synthetic force parameter and the preset force-sound conversion ratio includes: The first speed change rate is determined based on the difference between the real-time wind turbine speed corresponding to the wind turbine blades currently acquired and the real-time wind turbine speed corresponding to the wind turbine blades acquired at the previous moment. When the absolute value of the first speed change rate is greater than the preset change rate limit, the first dynamic correction coefficient is obtained. The first dynamic correction coefficient is the percentage parameter of the additional inertial impact force generated on the internal magnetic sensor when the wind turbine blades change speed rapidly. The product of the first dynamic correction coefficient and the first synthetic force parameter is used as the first dynamic impact increment. The first dynamic synthetic force parameter is obtained by summing the first dynamic impact increment and the first synthetic force parameter. The first friction sound frequency offset is recalculated based on the product of the first dynamic synthetic force parameter and the preset force-sound conversion ratio. When the absolute value of the first rotational speed change rate is determined to be no greater than the preset change rate limit, the original first frictional sound frequency offset is maintained based on the product between the first synthetic force parameter and the preset force-sound conversion ratio.

[0010] Preferably, the acquisition of the first mixed acoustic signature sequence collected by the magnetic sensor inside the wind turbine blade includes: The raw sound pressure time-series data collected by the magnetic sensor inside the wind turbine blade is obtained. The raw sound pressure time-series data is the continuous sampling data after the magnetic sensor inside the wind turbine blade converts sound fluctuations into voltage fluctuations. The original sound pressure time series data is truncated and segmented according to the real-time rotation phase of the wind turbine blade to obtain the first phase sound texture segment. The first phase sound texture segment is the sound pressure time series data segment corresponding to the wind turbine blade rotating to a single specific angle range. The transmission attenuation coefficient of the wind turbine blade casing for sound waves of different frequency bands is obtained, and the amplitude inverse compensation processing is performed on the first phase acoustic pattern segment based on the transmission attenuation coefficient to obtain the first compensated acoustic pattern segment. Based on the order of the real-time rotation phases corresponding to the wind turbine blades, the first compensated acoustic text segments are spliced ​​and combined to obtain the first mixed acoustic text sequence collected by the magnetic attraction sensor inside the wind turbine blades.

[0011] Preferably, the calculation of the first difference parameter based on the difference between the first mixed acoustic signature sequence and the first frictional sound reference sequence includes: The first frequency band to be processed in the same rotational phase interval of the first mixed acoustic pattern sequence and the first friction sound reference sequence is obtained. The first frequency band to be processed is the frequency interval where the sliding friction sound and crack sound of the magnetic sensor inside the wind turbine blade overlap. A first reference amplitude set is determined based on the amplitude of the first friction sound reference sequence in the first frequency band to be processed, and a first mixed amplitude set is determined based on the amplitude of the first mixed acoustic pattern sequence in the first frequency band to be processed. The difference between each amplitude in the first mixed amplitude set and the amplitude at the corresponding phase and frequency position in the first reference amplitude set is calculated to obtain the first phase frequency difference set. The first phase frequency difference set is the residual amplitude set of the wind turbine blade crack sound at each frequency point after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade. The positive residual amplitude values ​​that are greater than the preset residual amplitude value in the first phase frequency difference set are obtained and accumulated to obtain the first residual energy sum, and the first residual energy sum is used as the first difference parameter.

[0012] Preferably, the step of outputting wind turbine blade crack early warning information when the amplitude of the first difference parameter is greater than the first preset amplitude includes: The first set of historical difference parameters output by the magnetic sensor inside the wind turbine blade during a historical normal operating cycle is obtained. The first set of historical difference parameters is the historical record of residual acoustic energy parameters generated when the wind turbine blade has not cracked and the magnetic sensor has not become seriously loose. The first historical mean is obtained by calculating the mean of each historical first difference parameter in the historical first difference parameter set, and the first historical dispersion is obtained by calculating the dispersion of each historical first difference parameter in the historical first difference parameter set relative to the first historical mean. The dynamic warning threshold is calculated based on the product of the first historical mean and the preset multiple, and the sum of the first historical dispersion. The dynamic warning threshold is a first preset amplitude that is adaptively adjusted to adapt to the gradual aging and loosening of the magnetic sensor inside the wind turbine blade. When the first difference parameter calculated at the moment is greater than the dynamic early warning threshold, the wind turbine blade crack early warning information is output.

[0013] Preferably, the dynamic early warning threshold is calculated based on the product of the first historical mean and a preset multiple, and the sum of the first historical dispersions, including: The first loosening trend parameter is obtained by comparing the amplitude of the first friction sound reference sequence in the first frequency band to be processed in the current calculation cycle with the amplitude of the first friction sound reference sequence in the first frequency band to be processed in the previous calculation cycle. The first loosening trend parameter is a relative change parameter characterizing the degree of decrease in the magnetic attraction force of the magnetic attraction sensor inside the wind turbine blade. Get the current real-time ambient wind speed corresponding to the wind turbine blades, get the preset wind speed correction coefficient reference list, and find the first wind speed correction coefficient based on the current real-time ambient wind speed. The first wind speed correction coefficient is the compensation parameter for the amplification effect of ambient wind speed on the conducted noise of the wind turbine blade shell. When the first loosening trend parameter is greater than the preset loosening limit, the dynamic warning threshold is updated based on the product between the first wind speed correction coefficient and the dynamic warning threshold to obtain the first updated warning threshold. When the first difference parameter calculated at the moment is greater than the first update warning threshold, the wind turbine blade crack warning information is output.

[0014] Preferably, the construction of the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the wind turbine blade based on the sum of the first friction sound frequency offset and the reference friction sound frequency includes: The real-time rotation phase corresponding to the wind turbine blade is divided into multiple equal phase intervals. The peak frequency and peak amplitude of the first friction sound reference sequence in each equal phase interval are obtained to obtain the first standard friction feature set. The first standard friction feature set is the frequency and sound pressure combination of the maximum friction sound generated by the magnetic attraction sensor inside the wind turbine blade at each rotation angle. Based on the comparison of the frequency amplitude distribution of the first hybrid acoustic signature sequence in each equally divided phase interval with the first standard friction feature set, a first separation control sequence is generated. The first separation control sequence is a set of binary control commands that indicate whether to retain or remove data at each time and frequency point. The first separation control sequence and the first mixed acoustic text sequence are multiplied point by point to obtain the first separated acoustic text sequence. The first separated acoustic text sequence is the residual acoustic text data of the wind turbine blade after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade. The residual frequency points of the first separated voiceprint sequence that are greater than the preset separation amplitude in the first frequency band to be processed are accumulated to obtain the second residual energy sum, and the second residual energy sum is used as the calculation reference quantity of the first difference parameter.

[0015] Preferably, the first separation control sequence is generated by comparing the frequency amplitude distribution of the first mixed acoustic signature sequence in each equally divided phase interval with the first standard friction feature set, including: When the difference between the amplitude of a specific time-frequency point in the first mixed acoustic signature sequence and the peak amplitude of the corresponding phase and frequency position in the first standard friction feature set is less than a preset difference limit, the control command corresponding to the specific time-frequency point in the first separation control sequence is determined to be a rejection command. When the difference between the amplitude of a specific time-frequency point in the first mixed acoustic signature sequence and the peak amplitude of the corresponding phase and frequency position in the first standard friction feature set is greater than or equal to a preset difference limit, the control command corresponding to the specific time-frequency point in the first separation control sequence is determined to be a reserved command. The elimination or retention instructions corresponding to each time frequency point in the first separation control sequence are sequentially arranged and combined to form a continuous binary control instruction array; The continuous binary control command array is mapped and multiplied point by point with the first mixed voiceprint sequence, and the time-frequency data of the corresponding removal command is removed, while the time-frequency data of the corresponding retention command is retained.

[0016] Secondly, this application also provides a wind turbine blade acoustic signature monitoring system, including: The first acquisition module is used to acquire the real-time fan speed corresponding to the fan blades and the real-time rotation phase corresponding to the fan blades. The module is used to calculate the first composite force parameter based on the real-time fan speed and the real-time rotation phase, and to calculate the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the fan blade based on the first composite force parameter. The second acquisition module is used to acquire the first mixed acoustic pattern sequence collected by the magnetic attraction sensor inside the wind turbine blades; The calculation module obtains a first difference parameter based on the difference between the first mixed acoustic pattern sequence and the first frictional sound reference sequence; The output module is used to output wind turbine blade crack warning information when the amplitude of the first difference parameter is greater than the first preset amplitude.

[0017] The beneficial effects of implementing this application are as follows: This application provides a method and system for monitoring the acoustic signature of wind turbine blades. By introducing the real-time speed and rotation phase of the wind turbine, a dynamic first frictional sound reference sequence reflecting the physical laws of sensor loosening and friction is constructed, overcoming the shortcomings of conventional static filtering in adapting to dynamic noise. By calculating the difference between the actually collected first mixed acoustic signature sequence and the first frictional sound reference sequence, dynamic frictional noise overlapping with the frequency band of the actual crack can be accurately separated, and the first difference parameter representing the actual crack can be extracted. The above processing logic is completed entirely at the software and data calculation level, automatically eliminating false alarms caused by sensor loosening without downtime, restoring the monitoring system's sensitivity to actual cracks, and avoiding the waste of maintenance resources caused by false alarms and the risk of blade breakage caused by missed alarms. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of the acoustic signature monitoring method for wind turbine blades involved in this application; Figure 2 This is a flowchart of the judgment logic for step S23 of the wind turbine blade acoustic signature monitoring method involved in this application; Figure 3 This is a flowchart of the judgment logic for step S242 of the wind turbine blade acoustic signature monitoring method involved in this application; Figure 4 This is an application environment diagram of the acoustic signature monitoring method for wind turbine blades involved in this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] Figure 1 A flowchart of the wind turbine blade acoustic signature monitoring method provided in this application embodiment is shown, including: Step S1: Obtain the real-time fan speed and the real-time rotation phase of the fan blades.

[0022] It should be understood that the wind turbine main control system usually sends out status data frames at preset time intervals. The acquisition action in step S1 is specifically executed by the first acquisition module in the acoustic fingerprint monitoring system. The first acquisition module listens to the communication bus of the wind turbine main control system and captures data packets containing impeller speed indicators and yaw angle indicators.

[0023] Furthermore, the acquired data packets are verified. If the verification passes, the rotational speed and angle values ​​in the data packets are extracted and recorded as the real-time wind turbine rotational speed and the real-time rotational phase of the wind turbine blades, respectively. If the verification fails, a retransmission request is sent to the wind turbine main control system. If the retransmission request fails after reaching a preset limit, the rotational speed and phase from the previous moment stored locally are used as the current real-time data.

[0024] In this embodiment, the preset time interval is based on the normal communication cycle of the wind turbine controller and the latency tolerance setting for acoustic feature extraction, and the preset number of attempts is based on the stability calibration of the communication link. The first acquisition module generates a corresponding timestamp while acquiring data, and packages the timestamp, rotational speed value, and angle value into a package and sends it to the subsequent processing module to keep the acoustic data and motion attitude data aligned on the time axis.

[0025] Step S2: Calculate the first composite force parameter based on the real-time fan speed and the real-time rotation phase of the fan blades, and calculate the first frictional sound reference sequence corresponding to the magnetic sensor inside the fan blades based on the first composite force parameter.

[0026] Among them, the first friction sound reference sequence is the frequency amplitude distribution data of the acoustic pattern when the magnetic sensor inside the wind turbine blade is subjected to alternating centrifugal force and gravity to generate sliding friction.

[0027] In some embodiments, step S2 includes steps S21 to S24: Step S2: Obtain the ratio between the square of the real-time fan speed corresponding to the fan blade and the square of the reference speed corresponding to the fan blade, and use the ratio result as the first centrifugal force coefficient corresponding to the magnetic attraction sensor inside the fan blade.

[0028] Among them, the first centrifugal force coefficient is the percentage of the outward stretching force generated on the internal magnetic sensor when the fan blades rotate at high speed.

[0029] It should be noted that when the wind turbine blades rotate, the internal magnetic sensors are subjected to centrifugal pull along the blade span, and this pulling force is proportional to the square of the rotational speed. In this application, in order to quantify the centrifugal pulling effect, it is necessary to first set a reference speed, that is, the impeller speed when the wind turbine is in its rated full-load state.

[0030] Furthermore, after obtaining the real-time fan speed sent by the fan main control system, the real-time fan speed and the reference speed are multiplied by themselves to obtain the square of the real-time fan speed and the square of the reference speed corresponding to the fan blade. Then, the two are compared, and the resulting dimensionless value is the first centrifugal force coefficient.

[0031] Specifically, when the real-time fan speed is lower than the reference speed, the first centrifugal force coefficient is less than the preset reference constant, indicating that the current centrifugal pulling force has not reached the limit design load; when the real-time fan speed is equal to the reference speed, the first centrifugal force coefficient is equal to the preset reference constant; when the real-time fan speed momentarily exceeds the reference speed due to gusts, the first centrifugal force coefficient is greater than the preset reference constant, indicating that the slippage trend between the sensor base and the lightning protection metal mesh intensifies.

[0032] Furthermore, the reference speed is obtained by directly reading it from the parameter table on the wind turbine's nameplate and storing it in the storage unit of the acoustic signature monitoring system.

[0033] In practical applications, if the same wind farm contains wind turbines of different models, each wind turbine needs to be configured with a corresponding reference speed to avoid errors in coefficient calculation due to the mixing of different models.

[0034] It should also be noted that the calculated result of the first centrifugal force coefficient will serve as the basic input for constructing the mechanical model. If there is a deviation in the calculation of the first centrifugal force coefficient, it will directly lead to inaccurate prediction of the subsequent frictional sound frequency offset. After each calculation of the first centrifugal force coefficient, it will be compared with the coefficient of the previous calculation cycle. If the absolute value of the difference exceeds the preset jump tolerance, it is determined that the currently acquired real-time fan speed may have sensor reading glitches. The real-time fan speed acquired in this cycle will be automatically discarded, and the speed of the previous cycle will be used to perform self-multiplication and ratio calculations again, thereby ensuring the continuity and smoothness of the first centrifugal force coefficient.

[0035] Furthermore, considering that sensors installed at different spanwise positions on the blade have different radii of rotation, the centrifugal forces at the same rotational speed vary significantly. A mapping table between sensor installation positions and radii of rotation correction factors is pre-configured in the system memory.

[0036] Before performing the self-multiplication operation, the gyration radius correction factor is called first. The real-time fan speed is multiplied by the gyration radius correction factor, and the resulting effective linear velocity equivalent value is used to replace the original speed for subsequent self-multiplication and ratio operations.

[0037] In this embodiment, the preset reference constant is set to constant one, the preset jump tolerance is determined based on the maximum allowable acceleration of the wind turbine under normal gust conditions, and the gyration radius correction factor is calculated based on the geometric dimensions of the sensor's installation drawing inside the blade.

[0038] For example, if a certain type of wind turbine has an impeller speed of 14 revolutions per minute at its rated full capacity, and the current real-time speed is 7 revolutions per minute, then the square of 7 is compared with the square of 14, resulting in a value of 0.25. At this point, the first centrifugal force coefficient is 0.25, and the centrifugal force on the sensor is relatively weak, with the friction noise primarily driven by the gravitational component. However, when encountering strong gusts and the real-time speed spikes to 16 revolutions per minute, the square of 16 is compared with the square of 14, resulting in a value of approximately 1.3. The first centrifugal force coefficient is greater than 1, and the proportion of outward stretching force on the sensor increases significantly, leading to a sharp increase in the slippage trend.

[0039] Step S22: Obtain the phase offset angle of the real-time rotation phase of the wind turbine blade within the preset quadrant interval, and obtain the first gravity compression coefficient based on the cosine product of the phase offset angle and the reference quadrant angle.

[0040] Among them, the first gravity compression coefficient is the percentage of frictional force generated by gravity pressing down on the internal magnetic sensor when the fan blades rotate to different angles.

[0041] It should be noted that the sound in this application is peeling friction, and the change of gravity with the rotation angle of the blade must be taken into account.

[0042] Specifically, since the sensor is attached to the inner wall of the blade, when the blade rotates to the vertically downward position, the sensor's own weight is completely pressed against the lightning protection metal mesh, generating the maximum static friction force; when the blade rotates to the vertically upward position, the sensor's own weight moves away from the metal mesh, and the squeezing force is minimal.

[0043] In this embodiment, based on the real-time rotation phase of the wind turbine blade, its angular position within the circumference is determined, and the angle between this angular position and the vertically downward reference angle is calculated and denoted as the phase offset angle. The cosine value of this phase offset angle is obtained and multiplied by a preset reference constant to obtain the first gravity compression coefficient.

[0044] In one implementation, this can be achieved by directly calling the cosine calculation instruction in the mathematical function library. For example, when the phase offset angle is 0 degrees, the first gravity compression coefficient is 1, which means the compression is the strongest; when the phase offset angle is 90 degrees, the first gravity compression coefficient is 0, which means the compression in the horizontal position is moderate; when the phase offset angle is 180 degrees, the first gravity compression coefficient is negative 1, which means the direction of gravity is away from the contact surface, and at this time the adhesion is maintained only by magnetic force.

[0045] In this application, the preset quadrant interval is a complete circumference range, and the reference quadrant angle is the angle reference corresponding to the vertical downward direction. This setting is based on the physical common sense that the direction of the gravitational field is constant.

[0046] Step S23: The first composite force parameter is obtained by summing the first centrifugal force coefficient and the first gravity compression coefficient, and the first friction sound frequency offset is obtained by multiplying the first composite force parameter and the preset force-sound conversion ratio.

[0047] In some embodiments, the step S23, "obtaining the first frictional sound frequency offset based on the product between the first synthetic force parameter and the preset force-sound conversion ratio," specifically involves the following steps: Figure 2 As shown, steps S231 to S234 are included: Step S231: Calculate the first speed change rate based on the difference between the real-time fan speed corresponding to the currently acquired fan blade and the real-time fan speed corresponding to the fan blade acquired at the previous moment.

[0048] Among them, the first rotational speed change rate is the instantaneous change parameter of the wind turbine blades under the action of gusts of wind.

[0049] It should be noted that the randomness of wind conditions can cause frequent fluctuations in wind turbine speed, which in turn can introduce additional inertial shocks.

[0050] Specifically, the real-time fan speed corresponding to the previous frame of data is retrieved from the local timestamp cache queue, and the difference between the real-time fan speed of the current frame and the real-time fan speed of the previous frame is taken as the absolute difference in fan speed. Then, the ratio of the absolute difference in fan speed to the time interval between the two frames of data is taken as the first rate of change in fan speed.

[0051] In one implementation, angular acceleration can be directly read from the acceleration sensor data of the wind turbine main control system as an alternative method. For example, when the system has a high-precision gyroscope, the angular acceleration value can be directly extracted as the first rate of change of rotational speed to reduce the delay caused by software calculation.

[0052] In this application, the calibration basis for the time interval is the data sampling period set in the step of obtaining the real-time wind turbine speed corresponding to the wind turbine blades and the real-time rotation phase corresponding to the wind turbine blades. The first speed change rate can reflect the dynamic response of the wind turbine under gust impact, providing a quantitative basis for subsequent judgment on whether a dynamic correction coefficient needs to be introduced.

[0053] Step S232: When the absolute value of the first speed change rate is greater than the preset change rate limit, obtain the first dynamic correction coefficient.

[0054] The first dynamic correction coefficient is the percentage of the additional inertial impact force generated on the internal magnetic sensor when the wind turbine blades change speed rapidly.

[0055] It should be noted that inertial impact only dominates the friction noise when the rotational speed fluctuation reaches a certain level. The absolute value of the calculated first rate of change of rotational speed is obtained and compared with a preset rate of change limit.

[0056] In this embodiment, the preset rate of change limit is obtained by statistically analyzing the preset high percentile of the speed fluctuation amplitude under historical normal operating conditions. When the absolute value of the first speed change rate is greater than the preset rate of change limit, the fan is determined to be under gust impact condition, and the corresponding first dynamic correction coefficient is retrieved from a preset mapping table. The mapping table is pre-stored in the monitoring system memory, with the horizontal axis representing the absolute value of the speed change rate and the vertical axis representing the corresponding first dynamic correction coefficient, and the two are positively correlated. When the absolute value of the first speed change rate is not greater than the preset rate of change limit, the fan is determined to be under stable operating condition, and the first dynamic correction coefficient is not retrieved.

[0057] Step S233: The product between the first dynamic correction coefficient and the first synthetic force parameter is used as the first dynamic impact increment. The first dynamic synthetic force parameter is obtained by summing the first dynamic impact increment and the first synthetic force parameter. The first friction sound frequency offset is recalculated based on the product between the first dynamic synthetic force parameter and the preset force-sound conversion ratio.

[0058] In this application, due to the acceleration and deceleration of the wind turbine blades under gust conditions, a tangential inertial shear force is generated on the contact surface between the magnetic sensor base and the lightning protection metal mesh. This shear force, superimposed on the original centrifugal and gravitational normal pressure, will instantly change the stick-slip state of the friction interface, causing a nonlinear shift in the friction howling frequency. Therefore, this increment must be included in the mechanical-to-acoustic conversion model.

[0059] In an alternative implementation, this step is limited by the instantaneous micro-slippage of the contact surface causing a sudden frequency change due to sensor magnetic attenuation and a rapid increase in blade speed caused by typhoon-level gusts. To circumvent these limitations, this step can also be implemented in a different manner than described above, specifically including: The first speed change rate obtained from the step of calculating the absolute value of the first speed change rate is multiplied by the first dynamic correction coefficient obtained from the step of obtaining the first dynamic correction coefficient. The resulting product is the first dynamic impact increment characterizing the tangential inertial shear effect.

[0060] Furthermore, the first dynamic impact increment is added to the first composite force parameter of the initial aggregation to obtain the corrected first dynamic composite force parameter.

[0061] Furthermore, a preset force-sound conversion ratio is retrieved from the memory. This ratio, obtained through laboratory bench calibration, characterizes the shift in the dominant frequency of frictional sound caused by a change in unit normal force. The first dynamic synthetic force parameter is multiplied by the preset force-sound conversion ratio, and the resulting product replaces the original first frictional sound frequency shift that did not consider inertia.

[0062] It should be noted that in scenarios involving extreme gusts of wind and severe degradation of the sensor's magnetic force, the first dynamic synthetic force parameter may exceed the linear response range of the preset force-sound conversion ratio. Direct multiplication in this case would result in a calculated frequency offset value far exceeding the actual physical limit. Therefore, after calculating the product, a limiting logic must be introduced. When the calculated frequency offset exceeds the maximum allowable offset threshold, the frequency offset is forcibly assigned the maximum allowable offset threshold.

[0063] In this embodiment, the maximum allowable offset threshold is calculated based on the size of the acoustic resonance cavity excited by the friction between the magnetic base and the lightning protection metal mesh.

[0064] It should also be noted that if the first rate of change of rotational speed is negative and its absolute value is large, it indicates that the fan is in a state of rapid deceleration. At this time, the sensor tends to maintain its original speed due to inertia, which will generate a shear force on the contact surface in the opposite direction to that during acceleration. When performing multiplication, the positive or negative attribute of the first dynamic correction coefficient will be automatically retained, making the first dynamic impact increment negative.

[0065] When the negative increment is added to the first composite force parameter, the first dynamic composite force parameter may fall below the critical value for the magnetic sensor to detach. To address this anomaly, the system has a built-in adsorption state determination logic: if the first dynamic composite force parameter is lower than the critical adsorption force parameter, it is determined that the sensor has experienced a suspension jump at the current instant. At this time, the friction sound is interrupted, the first friction sound frequency offset at the current moment is forcibly set to zero, and a suspension event is recorded.

[0066] Furthermore, when the first dynamic synthetic force parameter oscillates at a high frequency near the critical adsorption force parameter, the sensor is in a critical stick-slip state, which will excite broadband white noise. In response to this abnormal physical state, after calculating the first frictional sound frequency offset, if it is found that the offset alternates between positive and negative values ​​and the variance exceeds the preset fluctuation variance in multiple consecutive calculation cycles, it is determined that the sensor has entered a broadband excitation state.

[0067] Instead of outputting a single frequency offset, a rectangular window centered on the offset and covering a preset bandwidth is generated as the frequency offset output for that period, ensuring that subsequent acoustic reconstruction can cover the actual excited noise frequency band. In this embodiment, the calibration basis for the preset fluctuation variance is the sensor damping oscillation frequency band range determined in the laboratory.

[0068] For example, in bench testing, a certain type of magnetic sensor was found to have a preset force-to-sound conversion ratio of 2 Hz frequency shift per Newton of force. If the currently calculated first dynamic composite force parameter is 50 Newtons, then the frequency shift obtained by directly multiplying them is 100 Hz. For instance, when encountering a strong typhoon and experiencing rapid deceleration, the first dynamic impact increment is -20 Newtons, and the original first composite force parameter is 30 Newtons. The sum of these two values ​​results in a first dynamic composite force parameter of 10 Newtons. If this value is lower than the preset critical adsorption force parameter of 15 Newtons, the sensor is determined to be suspended, and the frequency shift is forcibly assigned a value of 0 Hz.

[0069] Step S234: When the absolute value of the first rotational speed change rate is determined to be no greater than the preset change rate limit, the original first frictional sound frequency offset is maintained based on the product between the first synthetic force parameter and the preset force-sound conversion ratio.

[0070] Preferably, when the fan is in a stable operating phase, it is determined that no dynamic inertial correction is needed. In this case, the first composite force parameter is directly called, and the first composite force parameter is multiplied by the preset force-sound conversion ratio. The resulting product is the first frictional sound frequency offset in the current cycle.

[0071] Step S234 ensures low consumption of computing resources under normal wind conditions, avoiding system delays caused by unnecessary correction calculations. Simultaneously, the absolute value of the first rotational speed change rate will be continuously monitored; once it exceeds a preset change rate limit, the system will immediately switch back to the calculation branch containing dynamic corrections.

[0072] Step S24: Based on the sum of the first friction sound frequency offset and the reference friction sound frequency, construct the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the wind turbine blade.

[0073] In some embodiments, step S24 includes steps S241 to S24: Step S241: Divide the real-time rotation phase corresponding to the wind turbine blade into multiple equal phase intervals, obtain the peak frequency and peak amplitude of the first friction sound reference sequence in each equal phase interval to obtain the first standard friction feature set. The first standard friction feature set is the combination of frequency and sound pressure of the magnetic sensor inside the wind turbine blade generating the maximum friction sound at each rotation angle.

[0074] It should be noted that continuous acoustic data is difficult to directly perform discrete comparisons and must be processed by gridding.

[0075] Specifically, the complete circumference is divided into multiple equally divided phase intervals according to a preset angular step size.

[0076] In this embodiment, the calibration basis for the preset angle step size is to ensure that there are enough acoustic sampling points in each interval for peak extraction. For each evenly divided phase interval, all frequency and amplitude data points falling within that interval in the first friction sound reference sequence are traversed. The data point with the largest amplitude is found, and the frequency corresponding to this data point is extracted as the peak frequency of that interval, and the amplitude corresponding to this data point is extracted as the peak amplitude of that interval. The peak frequencies and peak amplitudes of all intervals are arranged in phase order to form the first standard friction feature set. For example, in the first evenly divided phase interval, if the largest amplitude in the reference sequence is found to be a preset decibel value and the corresponding frequency is a preset hertz, then this combination is stored in the set. This step S241 realizes the discretization representation of continuous acoustic features, which greatly reduces the computational complexity of subsequent comparison operations.

[0077] Step S242: Based on the frequency amplitude distribution of the first hybrid acoustic signature sequence in each equally divided phase interval, compare it with the first standard friction feature set to generate the first separation control sequence.

[0078] The first separation control sequence is a set of binary control commands that indicate whether to retain or discard data at each time and frequency point.

[0079] In some embodiments, a first separation control sequence is generated by comparing the frequency amplitude distribution of the first mixed acoustic signature sequence in each equally divided phase interval with a first standard friction feature set. Specific operations are as follows: Figure 3 The following are included: Step S2421: When the difference between the amplitude of a specific time-frequency point in the first mixed voiceprint sequence and the peak amplitude of the corresponding phase and frequency position in the first standard friction feature set is less than a preset difference limit, the control command corresponding to the specific time-frequency point in the first separation control sequence is determined to be a rejection command.

[0080] It should be noted that if the amplitude of a certain point in the actual signal is very close to the amplitude of the predicted friction characteristics, then it is highly suspected that the point is friction noise.

[0081] Specifically, the amplitude of any specific time-frequency point is extracted from the first mixed acoustic signature sequence, and feature points with the same phase interval and frequency position are found in the first standard friction feature set. The absolute difference between these two amplitudes is calculated. If the absolute difference is less than a preset difference limit, it is determined that the energy at that time-frequency point is mainly contributed by frictional sound.

[0082] In this embodiment, the preset difference limit is set by measuring the amplitude fluctuation range of the friction sound pattern during the rotation of the wind turbine in a laboratory setting. At this time, a rejection instruction flag is written at the position corresponding to the time-frequency point in the first separation control sequence.

[0083] Step S2422: When the difference between the amplitude of a specific time-frequency point in the first mixed acoustic signature sequence and the peak amplitude of the corresponding phase and frequency position in the first standard friction feature set is greater than or equal to a preset difference limit, the control command corresponding to the specific time-frequency point in the first separation control sequence is determined to be a reserved command.

[0084] Understandably, if the actual signal amplitude is significantly higher than the predicted friction amplitude, it indicates that the energy from other sound sources is superimposed at that point. Using the same search and calculation logic as the rejection criteria, the absolute difference between the actual amplitude at a specific time-frequency point and the corresponding predicted peak amplitude is obtained. When this absolute difference is greater than or equal to a preset difference limit, it indicates that the energy at that time-frequency point, in addition to friction sound, also includes real crack tearing sound or other effective acoustic events.

[0085] At this point, a retention instruction flag, i.e., a retention instruction, is written at the position corresponding to the time-frequency point in the first separation control sequence. This decision logic ensures that the weak crack signal can be captured by penetrating the masking of friction noise.

[0086] Step S2423: The elimination or retention instructions corresponding to each time frequency point in the first separation control sequence are sequentially arranged and combined to form a continuous binary control instruction array.

[0087] Specifically, discrete control commands are concatenated and spliced ​​together according to the order of the time axis and the frequency axis to form a binary matrix with the same dimension as the first mixed voiceprint sequence, which serves as the basis for subsequent mask calculations.

[0088] Step S2424: Perform point-by-point mapping and multiplication on the continuous binary control command array and the first mixed voiceprint sequence, remove the time-frequency data corresponding to the removal command, and retain the time-frequency data corresponding to the retention command.

[0089] Furthermore, a multiplication operation is performed between the binary control command array and the corresponding elements of the first mixed voiceprint sequence. The position data of the discard command is set to zero, while the position data of the retain command is retained as is, thereby achieving signal masking and filtering at the physical level.

[0090] Step S243: The first separation control sequence and the first mixed acoustic pattern sequence are multiplied point by point to obtain the first separated acoustic pattern sequence. The first separated acoustic pattern sequence is the residual acoustic pattern data of the wind turbine blade after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade.

[0091] In this application, the frequency band dominated by the original friction sound in the spectrum of the first separated acoustic sequence output after masking multiplication will show a dip, because only the frequency band containing real crack features will stand out.

[0092] Specifically, the first separated acoustic signature sequence is stored in the system cache, awaiting subsequent energy statistics. Step S243 effectively blocks the interference of frictional noise on the feature extraction operation, avoiding the phase distortion problem caused by traditional filters. At the same time, the system will perform a continuity check on the first separated acoustic signature sequence. If a large area of ​​data is found to be zero, an abnormal alarm will be triggered, indicating that the sensor loosening may have exceeded the model prediction range.

[0093] Step S244: Obtain the residual frequency points of the first separated voiceprint sequence that are greater than the preset separation amplitude in the first frequency band to be processed, and calculate the second residual energy sum by accumulating them. Use the second residual energy sum as the calculation reference quantity of the first difference parameter.

[0094] In this application, to quantify the intensity of the residual acoustic signature, a first frequency band to be processed is defined, which covers the typical excitation frequency band of fiberglass cracks. The amplitude values ​​of all frequency points in this frequency band of the first separated acoustic signature sequence are traversed, and frequency points with amplitudes greater than a preset separation amplitude are selected. The amplitude values ​​of these selected frequency points are summed, and the resulting sum is the second residual energy sum.

[0095] In this embodiment, the first frequency band to be processed is set to cover the frequency band of micro-fractures in fiberglass material, and the preset separation amplitude is set to a preset number of decibels above the background noise reference value. Here, the reference value is obtained statistically based on the background noise collected when the wind turbine is stopped. The second residual energy sum will serve as a direct reference for assessing the health status of the blades.

[0096] Step S3: Obtain the first mixed acoustic pattern sequence collected by the magnetic sensor inside the wind turbine blade. The first mixed acoustic pattern sequence is the acoustic time sequence data containing friction sound and crack sound actually collected by the magnetic sensor inside the wind turbine blade.

[0097] In some embodiments, acquiring the first mixed acoustic signature sequence collected by the magnetic sensor inside the wind turbine blade includes: Step S31: Obtain the raw sound pressure time series data collected by the magnetic sensor inside the wind turbine blade. The raw sound pressure time series data is the continuous sampling data after the magnetic sensor inside the wind turbine blade converts the sound fluctuations into voltage fluctuations.

[0098] It should be noted that the data collected by the sensor usually includes DC bias and high-frequency interference.

[0099] Specifically, the electret condenser microphone inside the magnetic sensor converts sound pressure fluctuations into voltage signals. After passing through a preamplifier circuit and an anti-aliasing filter, the signals are input to the analog-to-digital converter chip of the voiceprint monitoring system. The voltage signals are discretized and sampled according to a preset sampling frequency to obtain a series of digital codes, which are the original sound pressure timing data.

[0100] In one implementation, this can be achieved using an industrial-grade audio acquisition card, such as a high-resolution acquisition device with a sampling frequency covering the ultrasonic band, ensuring the capture of crack sounds across the entire frequency band. In this application, the acquired raw sound pressure timing data is first stored in a first-in-first-out data buffer queue to smooth out the impact of instantaneous data surges on the system bus.

[0101] Step S32: The original sound pressure time sequence data is truncated and segmented according to the real-time rotation phase corresponding to the wind turbine blades to obtain the first phase acoustic text segment.

[0102] The first phase acoustic signature segment is the time sequence data segment of sound pressure corresponding to the rotation of the wind turbine blades to a single specific angle range.

[0103] It should be noted that acoustic characteristics are strongly correlated with rotation angle, and slices must be taken according to angle.

[0104] Specifically, the real-time rotational phase timestamp sequence, obtained synchronously in the steps of acquiring the real-time fan speed corresponding to the fan blades and the real-time rotational phase corresponding to the fan blades, is retrieved. The start and end times of each evenly divided phase interval are marked on the time axis of the original sound pressure time series data. Based on these start and end times, the original sound pressure time series data is truncated, segmenting the time series data into segments corresponding one-to-one with each phase interval, which are the first phase acoustic signature segments. This slicing operation enables subsequent processing to perform directional analysis of acoustic features at specific angles, while eliminating the time axis stretching effect caused by speed fluctuations.

[0105] Step S33: Obtain the transmission attenuation coefficient of the wind turbine blade casing for sound waves of different frequency bands, and perform amplitude reverse compensation processing on the first phase acoustic pattern segment based on the transmission attenuation coefficient to obtain the first compensated acoustic pattern segment.

[0106] It should be noted that the fiberglass outer shell and internal web of the wind turbine blades have drastically different absorption and reflection characteristics for sound at different frequencies. Low-frequency sound waves have strong penetrating power while high-frequency sound waves are severely attenuated. Without frequency response compensation, the collected acoustic spectrum will be severely distorted, which will lead to a systematic deviation in the subsequent comparison with the friction sound reference sequence, making it impossible to accurately extract the real residual energy of the crack.

[0107] In this application, in order to eliminate the amplitude distortion caused by the sound wave propagation path, a frequency band-based inverse compensation operation must be performed on each first phase acoustic waveform segment.

[0108] Specifically, the transmission attenuation coefficient comparison table of the wind turbine blade casing for different frequency bands of sound waves is retrieved from the memory. The transmission attenuation coefficient comparison table is obtained by calibrating through frequency sweep tests by arranging standard sound sources and microphones at different positions on the blade, and records the attenuation decibels at each frequency point on the path from the sound source to the sensor.

[0109] Furthermore, the first phase acoustic signature is segmented and subjected to a Fast Fourier Transform (FFT) to convert the time-domain signal into a frequency-domain amplitude and phase distribution. For each frequency point in the frequency distribution, the corresponding conduction attenuation coefficient is looked up in a lookup table. The current amplitude at that frequency point is then added to the corresponding conduction attenuation coefficient to achieve inverse amplitude amplification compensation.

[0110] Furthermore, during actual compensation, if the conducted attenuation coefficients of certain high-frequency bands are extremely high, direct addition can lead to overshoot in the compensated amplitude, introducing unnecessary high-frequency white noise amplification. Therefore, when the found conducted attenuation coefficient exceeds the preset gain limit, the compensation amplitude at that frequency point is forcibly limited to the level corresponding to the preset gain limit. In this embodiment, the preset gain limit is calibrated based on the dynamic range and noise floor level of the system's analog-to-digital converter chip, preventing overcompensation from drowning out the real signal.

[0111] Furthermore, for abnormal scenarios where the conduction characteristics dynamically change due to shell strain during blade rotation, a fixed attenuation coefficient lookup table may not be fully applicable. After compensation, the energy of the first compensated acoustic signature segment in a known interference-free frequency band is extracted and compared with historical normal levels. If the deviation exceeds the preset tolerance, it is determined that the shell conduction characteristics have changed, triggering the adaptive fine-tuning process of the lookup table.

[0112] In this embodiment, the preset tolerance is derived from the repeatability error statistics of multiple wind tunnel tests. After compensation, the corrected frequency domain data is subjected to inverse Fourier transform back to the time domain to obtain the first compensated acoustic signature segment. This segment eliminates the frequency response distortion caused by the physical transmission path, laying the foundation for the accurate construction of the subsequent hybrid acoustic signature sequence.

[0113] It should also be noted that during amplitude inverse compensation, the stress state of the blade varies at different rotational phases, causing changes in the shell tension and resulting in periodic fluctuations in the conduction attenuation coefficient. To accurately capture these fluctuations, a phase modulation factor is added when retrieving the conduction attenuation coefficient lookup table. This phase modulation factor is obtained by interpolation based on the quadrant of the real-time rotational phase. The original queried conduction attenuation coefficient is multiplied by this phase modulation factor, and the product is used as the final effective attenuation coefficient in subsequent amplitude addition calculations. By introducing the phase modulation factor, the compensation process can dynamically conform to the stiffness change pattern of the blade shell during one rotation, significantly improving the reproduction accuracy of high-frequency crack sounds.

[0114] Step S34: Based on the order of the real-time rotation phases corresponding to the wind turbine blades, the first compensated acoustic text segments are spliced ​​and combined to obtain the first mixed acoustic text sequence collected by the magnetic attraction sensor inside the wind turbine blades.

[0115] Specifically, following the natural increasing order of rotational phase starting from the reference angle, the compensated first acoustic signature segments are spliced ​​together end-to-end. An overlapping addition method is used at the splicing points to eliminate glitches caused by phase jumps, ultimately forming a continuous and complete first hybrid acoustic signature sequence. This sequence not only retains the original acoustic characteristics but also carries precise phase attributes, facilitating alignment with the frictional sound reference sequence.

[0116] Step S4: Calculate the first difference parameter based on the difference between the first mixed acoustic pattern sequence and the first frictional sound reference sequence.

[0117] Among them, the first difference parameter is the amplitude parameter of the sound component of the wind turbine blade crack after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade.

[0118] In some embodiments, the calculation of the first difference parameter based on the difference between the first mixed acoustic signature sequence and the first frictional sound reference sequence includes: Step S41: Obtain the first frequency band to be processed in the same rotating phase interval between the first mixed acoustic pattern sequence and the first frictional sound reference sequence.

[0119] The first frequency band to be processed is the frequency range where the sliding friction sound and crack sound of the magnetic sensor inside the wind turbine blade overlap.

[0120] Specifically, the first frequency band to be processed is determined by comparing the overlapping region of the pure friction sound spectrum and the pure fiberglass tearing sound spectrum collected in the laboratory. Data slices of the first mixed acoustic waveform sequence and the first friction sound reference sequence within this frequency band and in the same evenly divided phase interval are extracted and used as the input source for subsequent difference calculations. This frequency band setting is based on the spectral envelope characteristics of fiberglass material under microscopic fracture, eliminating interference from irrelevant frequency bands and improving the targeting of the difference calculation.

[0121] Step S42: Determine a first reference amplitude set based on the amplitude of the first friction sound reference sequence in the first frequency band to be processed, and determine a first mixed amplitude set based on the amplitude of the first mixed acoustic pattern sequence in the first frequency band to be processed.

[0122] Specifically, for each frequency point within the first frequency band to be processed, the amplitude corresponding to the phase and frequency is extracted from the first frictional sound reference sequence and arranged sequentially to form a first reference amplitude set. Using the same logic, amplitudes are extracted from the first mixed acoustic pattern sequence to form a first mixed amplitude set. The dimensions and indices of the two sets are completely identical to ensure that subsequent subtraction operations do not result in misalignment.

[0123] Step S43: Calculate the difference between each amplitude in the first mixed amplitude set and the amplitude at the corresponding phase and frequency position in the first reference amplitude set to obtain the first phase-frequency difference set.

[0124] Among them, the first phase frequency difference set is the set of residual amplitude values ​​of the wind turbine blade crack sound at each frequency point after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade.

[0125] It should be noted that due to the complex aerodynamic noise generated by the rotating wind turbine blades, and the non-uniform attenuation of the magnetic force of the magnetic sensor, the actual first mixed acoustic waveform sequence collected includes not only real crack sounds and friction sounds, but also low-frequency modulation noise caused by wind shear. If the first mixed amplitude set is directly subtracted from the first reference amplitude set, a negative difference will occur when the mixed amplitude is lower than the reference amplitude. These negative differences do not physically represent the absence of crack sounds, but rather are due to the instantaneous increase in contact impedance caused by gravity unloading at a specific phase of the sensor, resulting in friction sound energy absorption exceeding sound radiation, or refraction attenuation of the acoustic propagation path due to sudden changes in wind speed.

[0126] In this application, in order to accurately separate friction sound and extract residual crack sound, the difference process must be subject to strict physical boundary constraints and logical judgments.

[0127] Specifically, each amplitude in the first mixed amplitude set is subtracted from the corresponding amplitude in the first reference amplitude set to obtain an initial difference. These initial differences are then iterated over; differences greater than zero are identified as potential crack excitation energy and retained; differences less than or equal to zero are identified as overestimation of frictional sound or signal attenuation and are forcibly set to zero. All processed values ​​are then arranged in their original order to form the first phase frequency difference set.

[0128] Furthermore, in actual calculations, if the amplitude at a specific frequency point in the first mixed amplitude set far exceeds the corresponding value in the first reference amplitude set, for example, if the difference exceeds the preset abnormal increment limit, this is usually not caused by a single micro crack, but is very likely caused by a sudden detachment and heavy impact between the sensor base and the metal mesh, or by the impact of foreign objects falling from inside the blade.

[0129] In this application, when the difference at a certain frequency point exceeds a preset abnormal increment limit, it is not included in the first phase frequency difference set, but is isolated and stored in the abnormal event log. In this embodiment, the preset abnormal increment limit is calibrated based on the maximum transient sound amplitude value during normal loosening and slippage of the sensor.

[0130] Furthermore, considering the damped oscillation characteristics of the acoustic excitation of fiberglass, a crack excitation event will produce amplitude fluctuations at multiple consecutive frequency points. After generating the first set of phase frequency differences, the continuity of the differences between adjacent frequency points is also checked. If a frequency point is found to have a difference greater than zero, but the differences between its two adjacent frequency points are both zero, then this isolated difference is determined to be random electrical noise interference, and it is removed from the set and set to zero. Through multi-layered logic checks, it is ensured that each residual amplitude in the first set of phase frequency differences truly reflects the acoustic energy distribution generated by the fiberglass tear, providing a clean data source for subsequent energy aggregation.

[0131] For example, if the amplitude of a specific frequency point extracted from the first mixed amplitude set is 15 dB, while the amplitude at the corresponding position in the first reference amplitude set is 12 dB, the difference between the two is 3 dB, which is greater than zero and less than the preset anomaly increment limit. This 3 dB residual amplitude is determined to be a possible crack excitation energy and is retained in the first phase frequency difference set. If the amplitude of this specific frequency point suddenly reaches 35 dB, the difference between this and 12 dB is 23 dB. If this value exceeds the preset anomaly increment limit of 20 dB, it is determined that this point is not a crack sound but a mechanical impact, and it is isolated to the anomaly event log, not participating in subsequent energy accumulation.

[0132] Step S44: Obtain the positive residual amplitude values ​​in the first phase frequency difference set that are greater than the preset residual amplitude value, accumulate them to obtain the first residual energy sum, and use the first residual energy sum as the first difference parameter.

[0133] Preferably, in order to filter out background noise fluctuations, a preset residual amplitude is set for filtering.

[0134] Specifically, the first phase frequency difference set is traversed, and all values ​​greater than a preset residual amplitude are extracted. In this embodiment, the preset residual amplitude is calibrated based on the background noise level setting of the analog-to-digital converter chip in the monitoring system. The extracted values ​​are summed to obtain the first residual energy sum, which is output as the first difference parameter. This parameter directly quantifies the intensity of the actual crack sound in the current cycle.

[0135] Step S5: When the amplitude of the first difference parameter is greater than the first preset amplitude, output wind turbine blade crack warning information.

[0136] In some embodiments, the step of outputting wind turbine blade crack early warning information when the amplitude of the first difference parameter is greater than a first preset amplitude includes: Step S51: Obtain the set of historical first difference parameters output by the magnetic sensor inside the wind turbine blade during a historical normal operating cycle.

[0137] Among them, the first set of historical difference parameters is the historical record of residual acoustic energy parameters generated when the wind turbine blades have not cracked and the magnetic sensor has not become seriously loose.

[0138] Specifically, the system retrieves the first difference parameter recorded within a preset time period from the system database.

[0139] In this embodiment, the calibration basis for the preset duration is a complete cycle setting of the wind turbine experiencing various wind conditions. Data from periods when the wind turbine is operating normally and without alarm records is selected to form the first set of historical difference parameters. This set reflects the background residual energy level of the blades and sensors at the current aging stage. If the amount of historical data stored in the database is insufficient for the preset duration, the retrieval cycle will be automatically extended or a data insufficiency warning will be issued, temporarily suspending the generation of dynamic thresholds.

[0140] Step S52: Calculate the mean of each historical first difference parameter in the historical first difference parameter set to obtain the first historical mean, and calculate the dispersion of each historical first difference parameter in the historical first difference parameter set relative to the first historical mean to obtain the first historical dispersion.

[0141] Specifically, the first historical difference parameters within the set are summed, and the sum is divided by the total number of parameters in the set to obtain the first historical mean, representing the average background energy. Further, the squared difference between each parameter and the first historical mean is calculated, and the sum of all squared differences is divided by the total number of differences to obtain the variance, i.e., the first historical dispersion. The first historical dispersion reflects the fluctuation range of background residual energy affected by factors such as wind conditions. The combination of the first historical mean and the first historical dispersion provides solid statistical support for the subsequent construction of adaptive thresholds.

[0142] Step S53: Calculate the dynamic warning threshold based on the product of the first historical mean and the preset multiple, and the sum of the first historical dispersion.

[0143] Among them, the dynamic early warning threshold is a first preset amplitude value that is adaptively adjusted to adapt to the gradual aging and loosening of the magnetic sensor inside the wind turbine blade.

[0144] In some embodiments, the calculation of the dynamic early warning threshold based on the product of the first historical mean and a preset multiple, and the summation of the first historical dispersion, includes: Step S531: Obtain the first loosening trend parameter by comparing the amplitude of the first friction sound reference sequence in the first frequency band to be processed within the current calculation period with the amplitude of the first friction sound reference sequence in the first frequency band to be processed within the previous calculation period.

[0145] The first loosening trend parameter is a relative change parameter characterizing the degree of decrease in the magnetic attraction force of the magnetic sensor inside the wind turbine blade; It should be noted that a decrease in magnetic attraction will lead to a change in the frictional contact area, which in turn will affect the amplitude of the frictional sound.

[0146] In this application, the average amplitude of the first frictional sound reference sequence constructed in the current cycle within the first frequency band to be processed, and the corresponding average amplitude stored in the previous cycle are extracted. The average amplitude of the current cycle is divided by the average amplitude of the previous cycle, and the quotient is the first loosening trend parameter. When this parameter is greater than a preset reference constant, it indicates that the frictional sound amplitude is increasing, and the decline in magnetic attraction leads to increased slippage. Step S531 provides a quantitative basis for evaluating the health status of the sensor.

[0147] Step S532: Obtain the current real-time ambient wind speed corresponding to the wind turbine blades, obtain the preset wind speed correction coefficient comparison list, and find the first wind speed correction coefficient based on the current real-time ambient wind speed.

[0148] The first wind speed correction factor is a compensation parameter for the amplification effect of ambient wind speed on the conducted noise of the wind turbine blade casing; It should be noted that high wind speeds can lead to increased vibration of the blade casing, amplifying acoustic transmission noise.

[0149] Specifically, this application obtains the current real-time ambient wind speed from the wind turbine meteorological station. A coefficient value matching the wind speed is then searched in a preset wind speed correction coefficient lookup list.

[0150] In this embodiment, the list is set as a piecewise function, with different correction coefficients corresponding to different wind speed intervals. The list data is obtained through acoustic transmission path testing and calibration in wind tunnel experiments. The first wind speed correction coefficient is used to adjust the warning threshold to prevent false alarms caused by amplified background noise at high wind speeds.

[0151] Step S533: When the first loosening trend parameter is greater than the preset loosening limit, the dynamic warning threshold is updated based on the product between the first wind speed correction coefficient and the dynamic warning threshold to obtain the first updated warning threshold. In this application, to prevent the sensor from becoming severely loose and causing a surge in friction noise that drowns out the crack signal, the alarm sensitivity needs to be adjusted.

[0152] Specifically, the first loosening trend parameter is compared with a preset loosening limit. When it exceeds this limit, the calculated dynamic warning threshold is multiplied by a first wind speed correction coefficient. Since the first wind speed correction coefficient is usually greater than a preset baseline constant, the resulting first updated warning threshold will be higher than the original dynamic warning threshold. This effectively relaxes the alarm conditions, avoiding frequent false alarms caused by sensor malfunctions and providing maintenance personnel with a buffer period to replace the sensor. The calibration basis for the preset loosening limit here is the inflection point of the nonlinear relationship curve between the magnetic attenuation model of the magnet material and the friction-induced sound pressure.

[0153] Step S5.34: When the first difference parameter calculated at the moment is greater than the first update warning threshold, output the wind turbine blade crack warning information; Specifically, the first difference parameter calculated and output in the current cycle is compared with the updated first warning threshold. If the value of the first difference parameter is greater than the first warning threshold, it is determined that there is an abnormal acoustic event inside the blade that exceeds the background noise tolerance. A wind turbine blade crack warning message containing the unit number, phase interval, and difference parameter value is immediately pushed to the remote monitoring center. If the first difference parameter is not greater than the first warning threshold, the system returns to a silent state.

[0154] Step S54: When the first difference parameter calculated at the moment is greater than the dynamic early warning threshold, output the wind turbine blade crack early warning information.

[0155] It should be noted that the initiation and propagation of cracks in wind turbine blades is a dynamic evolution process accompanied by complex acoustic emission characteristics. In the early stages, micro-cracks may only produce transient tearing sounds within a specific high-stress rotational phase range. Furthermore, this acoustic signal is modulated by the resonance of the internal cavity of the blade when it is transmitted to the sensor, causing the first difference parameter to fluctuate intermittently over multiple consecutive calculation cycles. If only a single-cycle instantaneous comparison is used for judgment, it is very easy to miss the detection during signal troughs or to generate false alarms during signal peaks due to occasional mechanical collisions.

[0156] In this application, in order to suppress the instability of judgment caused by intermittent fluctuations, it is necessary to introduce a dual judgment logic based on time windows, which combines persistence and accumulation.

[0157] Specifically, after calculating the first difference parameter each time, it is not only instantaneously compared with the dynamic warning threshold of the current cycle, but also stored in a sliding time window cache queue along with its corresponding calculation timestamp. The length of this queue is set according to the wind turbine's rotation cycle and can accommodate historical difference parameters for multiple consecutive cycles.

[0158] Furthermore, when actually executing the trigger judgment, it is first determined whether the first difference parameter calculated in the current single calculation is greater than the dynamic warning threshold.

[0159] If the value is greater than the threshold, record a suspected triggering event.

[0160] Furthermore, the sliding time window cache queue is traversed to count the number of suspected triggering events within the past preset observation period. Only when the number of occurrences reaches the preset continuous triggering limit is the action of outputting wind turbine blade crack warning information actually executed. In this embodiment, the preset observation period is calibrated based on the acoustic emission frequency characteristics during the statistical analysis of historical real crack propagation processes, effectively filtering out transient acoustic interference caused by single blade tip lightning strikes or bird impacts.

[0161] Furthermore, considering that under certain extreme operating conditions, the real-time rotation phase of the wind turbine may remain stagnant in the low-stress range for an extended period, causing actual cracks to remain silent, the number of suspected triggers within the sliding time window may not reach the preset continuous trigger limit. In this application, if the number of suspected triggers within the sliding time window does not reach the preset continuous trigger limit, but the sum of all first difference parameters exceeds the preset cumulative energy limit, it is also determined that the blade has progressive damage, and wind turbine blade crack warning information is directly output.

[0162] In this embodiment, the calibration basis for the preset cumulative energy upper limit is the product of the dynamic early warning threshold and the preset accumulation coefficient. Through dual cross-validation of instantaneous frequency and long-term energy accumulation, it is ensured that the monitoring system can accurately detect early cracks in the blades under various complex operating conditions, while keeping the false alarm rate at an extremely low level.

[0163] It should also be noted that after issuing a wind turbine blade crack warning, the monitoring and processing of subsequent acoustic fingerprint data will not immediately cease. After the warning is issued, the current time node will be marked in the sliding time window cache queue, and the warning review mode will be entered.

[0164] In the early warning review mode, the system will continuously monitor the changing trend of the first difference parameter. If the first difference parameter remains below the dynamic early warning threshold within the preset review period, and the number of suspected triggers within the sliding time window is cleared, the system will automatically lift the early warning status and send an early warning cancellation notification to the monitoring center.

[0165] If the first difference parameter triggers the preset continuous trigger limit or the preset cumulative energy limit again within the preset review period, the system will upgrade the warning level and attach a spectrum characteristic change trend chart to the warning information to guide maintenance personnel to prioritize the shutdown and inspection of the unit. In this embodiment, the preset review period is set based on the maximum time required for the wind turbine to complete one complete pitch and yaw wind-following process.

[0166] This embodiment also provides a wind turbine blade acoustic signature monitoring system, including: a first acquisition module, used to acquire the real-time wind turbine speed corresponding to the wind turbine blade and the real-time rotation phase corresponding to the wind turbine blade; The module is used to calculate the first composite force parameter based on the real-time wind turbine speed and real-time rotation phase, and to calculate the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the wind turbine blade based on the first composite force parameter. The second acquisition module is used to acquire the first mixed acoustic pattern sequence collected by the magnetic attraction sensor inside the wind turbine blades; The calculation module obtains a first difference parameter based on the difference between the first mixed acoustic pattern sequence and the first frictional sound reference sequence; The output module is used to output wind turbine blade crack warning information when the amplitude of the first difference parameter is greater than the first preset amplitude.

[0167] like Figure 4 The diagram illustrates the application environment of this application. The method and system can be implemented in this environment, which mainly includes a wind turbine generator set, an acoustic signature monitoring system, and a remote monitoring center. The wind turbine blades are equipped with a lightning protection metal mesh, and a magnetic sensor is magnetically fixed to this mesh to collect raw sound pressure timing data. The main control system of the wind turbine outputs real-time speed and rotation phase data frames; a weather station provides real-time environmental wind speed. The acoustic signature monitoring system comprises five modules: wind turbine operating parameter acquisition, friction sound reference sequence construction, mixed acoustic signature sequence acquisition, difference parameter calculation, and crack early warning information output. The system receives data from various hardware components, dynamically constructs a friction sound reference sequence, calculates difference parameters, and outputs wind turbine blade crack early warning information to the remote monitoring center when early warning conditions are triggered, guiding maintenance personnel to conduct precise shutdown inspections.

[0168] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring the acoustic signature of wind turbine blades, characterized in that, include: Obtain the real-time fan speed corresponding to the fan blades and the real-time rotation phase corresponding to the fan blades; The first composite force parameter is calculated based on the real-time fan speed and the real-time rotation phase, and the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the fan blade is calculated based on the first composite force parameter. Acquire the first mixed acoustic signature sequence collected by the magnetic sensor inside the wind turbine blade; The first difference parameter is obtained based on the difference between the first mixed acoustic pattern sequence and the first frictional sound reference sequence; When the amplitude of the first difference parameter is greater than the first preset amplitude, a wind turbine blade crack warning message is output.

2. The method as described in claim 1, characterized in that: The step of calculating the first composite force parameter based on the real-time fan speed and the real-time rotation phase, and calculating the first frictional sound reference sequence corresponding to the magnetic attraction sensor inside the fan blade based on the first composite force parameter, includes: The ratio between the square of the real-time fan speed corresponding to the fan blade and the square of the reference speed corresponding to the fan blade is calculated and used as the first centrifugal force coefficient corresponding to the magnetic attraction sensor inside the fan blade. The product of the phase offset angle of the real-time rotating phase within the preset quadrant interval and the cosine of the reference quadrant angle is used as the first gravity compression coefficient. The sum of the first centrifugal force coefficient and the first gravity compression coefficient is used as the first synthetic force parameter, and the first frictional sound frequency offset is obtained based on the product of the first synthetic force parameter and the preset force-sound conversion ratio. Based on the sum of the first friction sound frequency offset and the reference friction sound frequency, a first friction sound reference sequence corresponding to the magnetic attraction sensor inside the wind turbine blade is constructed.

3. The method as described in claim 2, characterized in that: The method of obtaining the first frictional sound frequency offset based on the product of the first synthetic force parameter and the preset force-sound conversion ratio includes: The first speed change rate is determined based on the difference between the real-time wind turbine speed corresponding to the wind turbine blades currently acquired and the real-time wind turbine speed corresponding to the wind turbine blades acquired at the previous moment. When the absolute value of the first speed change rate is greater than the preset change rate limit, the first dynamic correction coefficient is obtained. The first dynamic correction coefficient is the percentage parameter of the additional inertial impact force generated on the internal magnetic sensor when the wind turbine blades change speed rapidly. The product of the first dynamic correction coefficient and the first synthetic force parameter is used as the first dynamic impact increment. The first dynamic synthetic force parameter is obtained by summing the first dynamic impact increment and the first synthetic force parameter. The first friction sound frequency offset is recalculated based on the product of the first dynamic synthetic force parameter and the preset force-sound conversion ratio. When the absolute value of the first rotational speed change rate is determined to be no greater than the preset change rate limit, the original first frictional sound frequency offset is maintained based on the product between the first synthetic force parameter and the preset force-sound conversion ratio.

4. The method for monitoring the acoustic signature of wind turbine blades as described in claim 1, characterized in that: The acquisition of the first mixed acoustic signature sequence collected by the magnetic sensor inside the wind turbine blade includes: The raw sound pressure time-series data collected by the magnetic sensor inside the wind turbine blade is obtained. The raw sound pressure time-series data is the continuous sampling data after the magnetic sensor inside the wind turbine blade converts sound fluctuations into voltage fluctuations. The original sound pressure time series data is truncated and segmented according to the real-time rotation phase of the wind turbine blade to obtain the first phase sound texture segment. The first phase sound texture segment is the sound pressure time series data segment corresponding to the wind turbine blade rotating to a single specific angle range. The transmission attenuation coefficient of the wind turbine blade casing for sound waves of different frequency bands is obtained, and the amplitude inverse compensation processing is performed on the first phase acoustic pattern segment based on the transmission attenuation coefficient to obtain the first compensated acoustic pattern segment. Based on the order of the real-time rotation phases corresponding to the wind turbine blades, the first compensated acoustic text segments are spliced ​​and combined to obtain the first mixed acoustic text sequence collected by the magnetic attraction sensor inside the wind turbine blades.

5. The method for monitoring the acoustic signature of wind turbine blades as described in claim 1, characterized in that: The first difference parameter is obtained based on the difference between the first mixed acoustic signature sequence and the first frictional sound reference sequence, including: The first frequency band to be processed in the same rotational phase interval of the first mixed acoustic pattern sequence and the first friction sound reference sequence is obtained. The first frequency band to be processed is the frequency interval where the sliding friction sound and crack sound of the magnetic sensor inside the wind turbine blade overlap. A first reference amplitude set is determined based on the amplitude of the first friction sound reference sequence in the first frequency band to be processed, and a first mixed amplitude set is determined based on the amplitude of the first mixed acoustic pattern sequence in the first frequency band to be processed. The difference between each amplitude in the first mixed amplitude set and the amplitude at the corresponding phase and frequency position in the first reference amplitude set is calculated to obtain the first phase frequency difference set. The first phase frequency difference set is the residual amplitude set of the wind turbine blade crack sound at each frequency point after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade. The positive residual amplitude values ​​that are greater than the preset residual amplitude value in the first phase frequency difference set are obtained and accumulated to obtain the first residual energy sum, and the first residual energy sum is used as the first difference parameter.

6. The method for monitoring the acoustic signature of wind turbine blades as described in claim 5, characterized in that: When the amplitude of the first difference parameter is greater than the first preset amplitude, the wind turbine blade crack early warning information is output, including: The first set of historical difference parameters output by the magnetic sensor inside the wind turbine blade during a historical normal operating cycle is obtained. The first set of historical difference parameters is the historical record of residual acoustic energy parameters generated when the wind turbine blade has not cracked and the magnetic sensor has not become seriously loose. The first historical mean is obtained by calculating the mean of each historical first difference parameter in the historical first difference parameter set, and the first historical dispersion is obtained by calculating the dispersion of each historical first difference parameter in the historical first difference parameter set relative to the first historical mean. The dynamic warning threshold is calculated based on the product of the first historical mean and the preset multiple, and the sum of the first historical dispersion. The dynamic warning threshold is a first preset amplitude that is adaptively adjusted to adapt to the gradual aging and loosening of the magnetic sensor inside the wind turbine blade. When the first difference parameter calculated at the moment is greater than the dynamic early warning threshold, the wind turbine blade crack early warning information is output.

7. The method for monitoring the acoustic signature of wind turbine blades as described in claim 6, characterized in that: The dynamic early warning threshold is calculated based on the product of the first historical mean and a preset multiple, and the sum of the first historical dispersions, including: The first loosening trend parameter is obtained by comparing the amplitude of the first friction sound reference sequence in the first frequency band to be processed in the current calculation cycle with the amplitude of the first friction sound reference sequence in the first frequency band to be processed in the previous calculation cycle. The first loosening trend parameter is a relative change parameter characterizing the degree of decrease in the magnetic attraction force of the magnetic attraction sensor inside the wind turbine blade. Get the current real-time ambient wind speed corresponding to the wind turbine blades, get the preset wind speed correction coefficient reference list, and find the first wind speed correction coefficient based on the current real-time ambient wind speed. The first wind speed correction coefficient is the compensation parameter for the amplification effect of ambient wind speed on the conducted noise of the wind turbine blade shell. When the first loosening trend parameter is greater than the preset loosening limit, the dynamic warning threshold is updated based on the product between the first wind speed correction coefficient and the dynamic warning threshold to obtain the first updated warning threshold. When the first difference parameter calculated at the moment is greater than the first update warning threshold, the wind turbine blade crack warning information is output.

8. The method for monitoring the acoustic signature of wind turbine blades as described in claim 2, characterized in that: The first friction sound reference sequence corresponding to the magnetic attraction sensor inside the wind turbine blade is constructed based on the sum of the first friction sound frequency offset and the reference friction sound frequency, including: The real-time rotation phase corresponding to the wind turbine blade is divided into multiple equal phase intervals. The peak frequency and peak amplitude of the first friction sound reference sequence in each equal phase interval are obtained to obtain the first standard friction feature set. The first standard friction feature set is the frequency and sound pressure combination of the maximum friction sound generated by the magnetic attraction sensor inside the wind turbine blade at each rotation angle. Based on the comparison of the frequency amplitude distribution of the first hybrid acoustic signature sequence in each equally divided phase interval with the first standard friction feature set, a first separation control sequence is generated. The first separation control sequence is a set of binary control commands that indicate whether to retain or remove data at each time and frequency point. The first separation control sequence and the first mixed acoustic text sequence are multiplied point by point to obtain the first separated acoustic text sequence. The first separated acoustic text sequence is the residual acoustic text data of the wind turbine blade after removing the sliding friction sound of the magnetic sensor inside the wind turbine blade. The residual frequency points of the first separated voiceprint sequence that are greater than the preset separation amplitude in the first frequency band to be processed are accumulated to obtain the second residual energy sum, and the second residual energy sum is used as the calculation reference quantity of the first difference parameter.

9. The method for monitoring the acoustic signature of wind turbine blades as described in claim 8, characterized in that: The first separation control sequence is generated by comparing the frequency amplitude distribution of the first hybrid acoustic signature sequence in each equally divided phase interval with the first standard friction feature set, including: When the difference between the amplitude of a specific time-frequency point in the first mixed acoustic signature sequence and the peak amplitude of the corresponding phase and frequency position in the first standard friction feature set is less than a preset difference limit, the control command corresponding to the specific time-frequency point in the first separation control sequence is determined to be a rejection command. When the difference between the amplitude of a specific time-frequency point in the first mixed acoustic signature sequence and the peak amplitude of the corresponding phase and frequency position in the first standard friction feature set is greater than or equal to a preset difference limit, the control command corresponding to the specific time-frequency point in the first separation control sequence is determined to be a reserved command. The elimination or retention instructions corresponding to each time frequency point in the first separation control sequence are sequentially arranged and combined to form a continuous binary control instruction array; The continuous binary control command array is mapped and multiplied point by point with the first mixed voiceprint sequence, and the time-frequency data of the corresponding removal command is removed, while the time-frequency data of the corresponding retention command is retained.

10. A wind turbine blade acoustic signature monitoring system, employing the wind turbine blade acoustic signature monitoring method as described in any one of claims 1 to 9, characterized in that, include: The first acquisition module is used to acquire the real-time fan speed corresponding to the fan blades and the real-time rotation phase corresponding to the fan blades. The module is used to calculate the first composite force parameter based on the real-time fan speed and the real-time rotation phase, and to calculate the first friction sound reference sequence corresponding to the magnetic attraction sensor inside the fan blade based on the first composite force parameter. The second acquisition module is used to acquire the first mixed acoustic pattern sequence collected by the magnetic attraction sensor inside the wind turbine blades; The calculation module obtains a first difference parameter based on the difference between the first mixed acoustic pattern sequence and the first frictional sound reference sequence; The output module is used to output wind turbine blade crack warning information when the amplitude of the first difference parameter is greater than the first preset amplitude.

Images