Fan dynamic balance correction method and system based on multi-source sensor fusion

Abstract

The application provides a fan dynamic balance correction method and system based on multi-source sensing fusion, relates to the technical field, and performs rapid verification measurement analysis on the equipment, corrects the equipment calibration process through a reference deviation value, performs initial unbalance measurement analysis on the fan blade to be measured through a reliable reference and synchronously records main shaft operation parameters, performs trial weight measurement on each correction plane through the initial unbalance and generates a correction scheme, simultaneously performs self-adaptive matching on the sensing response coefficient of the verification link through a dynamic characteristic fingerprint, completes qualification determination through comparison of the accurate measurement data and the acceptance value, archives complete data, periodically verifies the reference standard part through the accumulated correction trend of the same type in the archived data, and the application improves the precision consistency, efficiency and quality traceability of the fan dynamic balance correction, greatly reduces the batch defective product risk caused by reference drift and the time and labor cost caused by repeated invalid correction.

Description

Technical Field

[0001] This invention proposes a fan dynamic balance correction method and system based on multi-source sensor fusion, which relates to the field of balance correction technology, specifically to the field of fan dynamic balance correction technology based on multi-source sensor fusion. Background Technology

[0002] Fan dynamic balancing is a critical process for ensuring stable, low-vibration, and low-noise operation of rotating products at high speeds. It is widely used in the manufacturing of small rotating equipment such as vacuum cleaners, hair dryers, and cooling fans. Existing dynamic balancing methods typically use vibration sensors to acquire imbalance data, and operators add or remove material on various correction planes based on the measurement results. However, existing systems lack proactive verification of the equipment's own condition before calibration to maintain the reliability of the measurement reference, potentially leading to batch deviations due to measurements taken with a deviated reference. The measurement process neglects the synchronous recording and reuse of operating parameters such as spindle speed and drive current, resulting in a lack of consistency between the measurement conditions in subsequent correction and verification stages and the initial measurement stages. Furthermore, a large amount of historical measurement data is used only for single-case judgments, failing to form a closed-loop feedback and trend warning system for the equipment's reference state, causing reference drift to often only be discovered after a large number of defective products have been produced. These problems collectively limit the accuracy, efficiency, and quality stability of the balancing process. Summary of the Invention

[0003] This invention provides a fan dynamic balance correction method and system based on multi-source sensor fusion to solve the above-mentioned problems: The present invention proposes a fan dynamic balance correction method and system based on multi-source sensor fusion, wherein the method includes: S1. Use independently stored reference standard parts to quickly verify and measure the equipment to obtain the reference deviation value before calibration; S2. Correct the equipment calibration process by using the benchmark deviation value to obtain a verified reliable benchmark. Use the reliable benchmark to perform initial imbalance measurement and analysis on the fan blade under test and record the main shaft operating parameters simultaneously to obtain the initial imbalance amount and dynamic feature fingerprint. S3. Perform trial weight measurements on each correction plane using the initial imbalance and generate a correction scheme. At the same time, use dynamic feature fingerprints to adaptively match the sensing response coefficients of the verification process to obtain accurate measurement data. S4. The pass / fail determination is completed by comparing the accurate measurement data with the acceptance value, and the complete data is archived. The reference standard parts are periodically verified by the accumulated correction trend of the same model in the archived data, and a systematic deviation warning signal is obtained and the reference is updated.

[0004] Furthermore, the system includes: The measurement and analysis module is used to quickly verify and analyze the equipment using independently stored reference standard parts to obtain the reference deviation value before calibration. The imbalance feature analysis module is used to correct the equipment calibration process by using the benchmark deviation value to obtain a verified reliable benchmark. The module then uses the reliable benchmark to perform initial imbalance measurement and analysis on the fan blade under test and simultaneously records the spindle operating parameters to obtain the initial imbalance amount and dynamic feature fingerprint. The adaptive matching module is used to perform trial weight measurements on each correction plane based on the initial imbalance and generate a correction scheme. At the same time, it adaptively matches the sensing response coefficients of the verification stage through dynamic feature fingerprints to obtain accurate measurement data. The periodic verification module is used to complete the qualification judgment by comparing the accurate measurement data with the acceptance value, and archive the complete data. The module performs periodic verification on the reference standard parts by the accumulated correction trend of the same model in the archived data, obtains the systematic deviation warning signal and triggers the reference update.

[0005] The beneficial effects of this invention are as follows: This invention solves the technical problem in the prior art where equipment lacks the ability to proactively recognize its current state before each calibration process is initiated, and cannot predict whether the calibration starting point has shifted. It enables rapid pre-checking of the equipment's health status and measurement accuracy before calibration, allowing operators to grasp the current degree and trend of equipment deviation before formal calibration begins. This improves the effectiveness and relevance of the calibration process initiation, avoiding distorted calibration results caused by blindly executing standard calibration procedures when significant equipment deviations already exist. It reduces the time and labor costs caused by initiating invalid calibrations due to unknown equipment status, and also reduces the potential quality risk of mass production defects due to incorrect calibration benchmarks. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of a fan dynamic balance correction method based on multi-source sensor fusion. Detailed Implementation

[0007] Example 1 The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0008] In one embodiment of the present invention, a fan dynamic balance correction method and system based on multi-source sensor fusion is proposed, the method comprising: S1. Use independently stored reference standard parts to quickly verify and measure the equipment to obtain the reference deviation value before calibration; S2. Correct the equipment calibration process by using the benchmark deviation value to obtain a verified reliable benchmark. Use the reliable benchmark to perform initial imbalance measurement and analysis on the fan blade under test and record the main shaft operating parameters simultaneously to obtain the initial imbalance amount and dynamic feature fingerprint. S3. Perform trial weight measurements on each correction plane using the initial imbalance and generate a correction scheme. At the same time, use dynamic feature fingerprints to adaptively match the sensing response coefficients of the verification process to obtain accurate measurement data after correction. S4. Complete the acceptance judgment by comparing precise measurement data with acceptable values, and archive the complete data. Periodically verify the benchmark standard parts based on the accumulated correction trends of the same model in the archived data, obtain systematic deviation warning signals, and trigger benchmark updates. Figure 1 As shown.

[0009] Multi-source sensor fusion refers to the comprehensive acquisition and collaborative processing of output information from various types of sensors, such as vibration sensors, laser optical positioning, motor drive current monitoring, and spindle encoder feedback. This enables the system to simultaneously acquire state information of the measured object and the equipment itself from multiple physical dimensions, providing a richer and more reliable data foundation than a single sensing method for subsequent deviation assessment, dynamic feature extraction, and adaptive adjustment.

[0010] Fan dynamic balancing refers to the process of adjusting the mass distribution of fan blades by measuring the magnitude and direction of the centrifugal force caused by uneven mass distribution when the fan blades are rotating. This adjustment is achieved by adding counterweights or removing materials to bring the fan blades within an acceptable range of imbalance during rotation, thereby reducing vibration and noise and extending bearing life.

[0011] A reference standard part refers to a standardized fan blade that has been precision machined and pre-balanced, has known and stable unbalance characteristics, and is used specifically for equipment calibration and verification. This standard part needs to be balanced with hardened balancing soil and stored independently to ensure that its mass distribution characteristics do not change significantly over a long period of time, serving as a reference benchmark for the accuracy of equipment measurements.

[0012] Rapid verification measurement analysis refers to performing a simplified measurement process using a reference standard as the test object before formal calibration. Vibration response signals of the standard in its current state are collected by vibration sensors and quickly compared with historical reference records of the standard to assess whether there are any deviations in the equipment. This avoids inaccurate calibration results caused by directly starting calibration when the equipment has changed its state.

[0013] The reference deviation value before calibration refers to the actual vibration amplitude and phase obtained by measuring the reference standard part under the current state of the equipment, and the degree of deviation between them and the corresponding standard amplitude and standard phase in the historical reference record of the reference standard part. This deviation value is expressed in numerical form and is a quantitative basis for judging whether the equipment needs to adjust the calibration parameters and how to adjust them.

[0014] Laser alignment reference point refers to using a laser light module installed on the balancing machine to emit a visible laser line. The operator aligns a specific mark on the fan blade or standard part with this laser line to ensure that the angle and orientation of the fan blade relative to the spindle remain consistent each time it is placed on the spindle. This operation eliminates the phase measurement error introduced by different fan blade installation angles and is a key step in ensuring measurement repeatability accuracy.

[0015] Spindle positioning and alignment confirmation refer to controlling the clamping and release of the fan blades by pressing the spindle positioning button to tighten or loosen the spindle chuck, and visually confirming whether the laser line accurately passes through the fan blade reference point. Together, these two steps constitute a standardized procedure for the fan blade installation status before each measurement, ensuring the consistency of the measurement starting conditions.

[0016] Vibration sensors are high-sensitivity accelerometers or velocity sensors installed near the main shaft of a balancing machine. They are used to sense the vibration signals generated by the unbalanced mass when the fan blades rotate. After amplification and analog-to-digital conversion, the signals are sent to the host computer for amplitude and phase extraction. Vibration sensors are the core sensing elements in dynamic balancing measurements.

[0017] The measured vibration characteristics refer to the vibration amplitude and phase data collected by the vibration sensor in the current measurement and extracted through signal processing. The amplitude reflects the magnitude of the vibration energy, and the phase reflects the position of the vibration peak relative to the reference point. Together, they constitute a complete vibration vector.

[0018] Historical benchmark records refer to the standard vibration amplitude and standard phase data of the benchmark component obtained through precise measurement when it was calibrated at the factory or when the equipment was last confirmed to be in good condition. This data is stored in the system database as a comparison reference for all subsequent verification measurements.

[0019] Quantitative assessment refers to transforming the difference between measured vibration characteristics and historical benchmark records into measurable numerical indicators through mathematical calculations, including the magnitude of amplitude deviation and the angular magnitude of phase deviation, thereby transforming equipment deviation from abstract observational perception into concrete and processable data.

[0020] Amplitude deviation refers to the algebraic difference between the measured vibration amplitude and the standard amplitude in the historical benchmark record. This difference quantifies the degree of change in the equipment's vibration energy response sensitivity. A positive deviation indicates that the current sensitivity is higher than the standard state, while a negative deviation indicates that the sensitivity has decreased.

[0021] Phase deviation refers to the angular difference between the measured vibration phase and the standard phase in the historical benchmark record. This difference quantifies the degree of change in the vibration response time delay or lead of the equipment, reflecting the phase drift that may exist in the mechanical transmission chain or electrical drive link.

[0022] A two-dimensional deviation vector is a vector expression consisting of two independent components: amplitude deviation and phase deviation. The amplitude deviation, as the magnitude component of the vector, determines the magnitude of the deviation intensity, while the phase deviation, as the direction angular component of the vector, determines the dominant direction of the deviation. The combination of the two gives a complete two-dimensional representation of the equipment deviation state.

[0023] The complex domain deviation vector refers to the form in which a two-dimensional deviation vector is expressed in a complex coordinate system, that is, the real part represents the magnitude deviation component and the imaginary part represents the phase deviation component. The rules of complex number operations facilitate subsequent vector addition, subtraction, rotation and magnitude-argument calculations.

[0024] The modulus is the straight-line distance from the origin to the endpoint of the vector in the complex plane of the deviation vector in the complex field. It is calculated by taking the square root of the sum of the squares of the vector components. The magnitude of this modulus reflects the overall strength level of the equipment's current deviation from the standard state.

[0025] Argument refers to the angle between the complex domain deviation vector and the positive real axis in the complex plane. It is obtained by calculating the arctangent function of the ratio of the imaginary part to the real part of the vector. This angle reflects the dominant direction of the equipment deviation in the two-dimensional space of amplitude and phase.

[0026] Polar coordinates refers to the way of expressing the same vector by replacing the horizontal and vertical components in rectangular coordinates with two parameters: magnitude and argument. This form of expression separates the intensity and direction of the deviation into two independent scalars, which makes it easier to determine whether to trigger calibration correction based on the magnitude of the intensity and to determine the source type of the deviation based on the direction.

[0027] A reliable reference is a measurement reference origin that can truly reflect the current actual response characteristics of the equipment. This reference is obtained by embedding the reference deviation value as a compensation coefficient into the calibration algorithm and correcting the original calibration parameter matrix item by item. This reference has been compensated for the current system deviation of the equipment, and all subsequent measurements are based on this corrected reference.

[0028] Initial imbalance measurement analysis refers to the process of measuring the original imbalance state of the fan blade under test before any correction operation is applied, after a reliable benchmark is established. This measurement eliminates the influence of installation eccentricity by averaging two flips and obtains the initial vibration amplitude and phase data of the fan blade.

[0029] The spindle operating parameters refer to three types of data recorded synchronously by the servo motor encoder, driver current monitoring module, and positioning sensor during the fan blade rotation and positioning process: rotation speed, drive current waveform, and positioning position coordinates. These parameters reflect the dynamic working status of the fan blade and spindle system during the measurement process.

[0030] The 180-degree rotation measurement refers to the operation of rotating the fan blade half a revolution around its central axis on the spindle chuck, realigning it with the reference point with laser light, and performing the measurement again. This operation changes the relative phase relationship between the fan blade's own imbalance vector and the installation eccentricity vector. By superimposing the two measurement results, the systematic error introduced by the installation eccentricity can be separated from the measurement results.

[0031] The comparative calculation refers to performing a vector difference operation in the complex domain between the vibration vector measured for the first time and the vibration vector measured after flipping 180 degrees. This eliminates the false vibration components caused by the fan blade installation position deviating from the main shaft rotation center, thereby obtaining the true unbalance vector caused only by the uneven mass distribution of the fan blade itself.

[0032] The initial imbalance refers to the magnitude of the equivalent imbalance mass corresponding to the uneven mass distribution of the fan blade itself, obtained after comparison and calculation, and its angular position on the circumference of the fan blade. This vector is the original basis for the subsequent trial weight measurement and correction scheme generation.

[0033] Dynamic fingerprint refers to a set of multi-dimensional feature parameters that integrate and process the spindle speed data, drive current data and position data recorded synchronously during two measurement processes. This set of parameters can comprehensively reflect the dynamic response characteristics of a specific fan blade under the combined action of the current spindle, drive and sensing system. This fingerprint has a unique identification function for the current measurement condition.

[0034] The drive current curve is a waveform record of the output current of the servo motor driver changing over time while maintaining the constant speed rotation of the spindle. The fluctuation pattern of the curve reflects the instantaneous change of the motor load torque when the fan blades rotate, and is an important physical quantity characterizing the load characteristics of the fan blades.

[0035] Current fluctuation characteristics refer to two parameters extracted from the driving current curve: fluctuation amplitude and fluctuation frequency. The fluctuation amplitude reflects the maximum range of current fluctuation around the stable average value and is related to the degree of imbalance of the fan blades and aerodynamic non-uniformity. The fluctuation frequency reflects the number of repetition cycles of current fluctuation per unit time and is related to the rotation frequency of the fan blades and the number of blades.

[0036] Speed ​​stability characteristics refer to two parameters extracted from spindle speed data: mean offset and standard deviation. The mean offset reflects the degree of deviation of the actual average speed from the set speed, while the standard deviation reflects the degree of dispersion of each speed data point relative to the average value. Together, they characterize the spindle's speed maintenance capability under load conditions.

[0037] The positioning repeatability feature refers to two parameters extracted from the spindle stop position data: the maximum deviation and the deviation distribution range. The maximum deviation reflects the maximum deviation from the target position among all stop positions, and the deviation distribution range reflects the statistical scattering width of the deviation values ​​of each stop position. Together, they characterize the repeatability accuracy of the spindle positioning system.

[0038] A multidimensional feature vector is a structured numerical combination formed by arranging and splicing feature parameters of different physical dimensions in a preset order. Mathematically, this vector is a coordinate point in a multidimensional space. Different fan blades or systems in different states will correspond to different coordinate positions, thereby achieving unique identification.

[0039] A six-dimensional feature vector is a specific multi-dimensional vector formed by sequentially splicing two-dimensional current sub-vectors, two-dimensional speed sub-vectors, and two-dimensional positioning sub-vectors. It has six dimensions, which correspond to the current fluctuation amplitude, current fluctuation frequency, speed mean offset, speed standard deviation, maximum positioning deviation, and positioning deviation distribution range, forming a unique six-parameter code for the dynamic characteristics of the system.

[0040] The current subvector is a two-dimensional vector composed of two parameters in the current fluctuation characteristics: fluctuation amplitude and fluctuation frequency. This vector expresses the intensity and rhythm of the dynamic characteristics of the motor load on a two-dimensional plane, and is the first two dimensions of the six-dimensional feature vector.

[0041] The rotational speed subvector is a two-dimensional vector composed of two parameters, the mean offset and the standard deviation, in the rotational speed stability characteristics. This vector expresses the systematic deviation and random fluctuation of the spindle speed holding capability on a two-dimensional plane, and is the middle two dimensions of the six-dimensional feature vector.

[0042] The positioning sub-vector is a two-dimensional vector composed of two parameters: the maximum deviation and the deviation distribution range in the positioning repeatability features. This vector expresses the extreme deviation and statistical dispersion of the principal axis positioning accuracy on a two-dimensional plane, and is the last two dimensions of the six-dimensional feature vector.

[0043] The correction planes refer to the two axial height positions for the dual-plane dynamic balancing correction method. The first correction plane is close to the top of the fan blade, and the second correction plane is close to the bottom of the fan blade. The operator applies counterweights or removes materials at the respective circumferential angle positions of these two planes to simultaneously correct static and couple imbalances.

[0044] Trial weight measurement refers to the process of adding a trial weight of known weight to the zero-degree angle position of each correction plane, restarting the measurement, and recording the vibration response. This operation changes the total vibration vector by introducing a known additional imbalance. The system calculates the influence coefficient by observing the change in the vibration vector before and after the trial weight, and then determines the actual required correction weight and angle.

[0045] The correction scheme refers to the complete set of operation instructions calculated by the system using the dual-plane influence coefficient method based on the changes in vibration vectors before and after each trial weight on the two planes. This set includes the corrected weight, correction angle, and direction of adding or subtracting mass on the first correction plane, as well as the corrected weight, correction angle, and direction of adding or subtracting mass on the second correction plane.

[0046] The sensing response coefficient is the scaling factor when a vibration sensor converts a mechanical vibration signal into an electrical signal output. This coefficient determines how much voltage or digital quantity is output for a unit vibration acceleration or velocity, and its value directly affects the amplitude accuracy of the measurement data.

[0047] Adaptive matching refers to the process by which the system compares the dynamic feature fingerprint stored during the initial measurement with the spindle operating parameters collected in real time during the current verification measurement. Based on the comparison results, it determines whether the current system state is consistent with the initial measurement. If there is a difference, it automatically adjusts the sensing response coefficient to ensure that the system response conditions during the verification measurement are consistent with those during the initial measurement.

[0048] Load matching degree is a quantitative index calculated by comparing the difference between the real-time drive current parameters during the current verification measurement and the current fluctuation characteristics during the initial measurement stored in the dynamic feature fingerprint. This index reflects the degree of deviation of the current motor load state from the load state during the initial measurement. The higher the value, the closer the load state is to the initial measurement.

[0049] Speed ​​matching degree refers to a quantitative index calculated by comparing the deviation of the real-time speed fluctuation parameters during the current verification measurement with the speed stability characteristics during the initial measurement stored in the dynamic feature fingerprint. This index reflects the degree of deviation of the current spindle speed holding state from the speed holding state during the initial measurement. The higher the value, the closer the speed stability is.

[0050] The operating condition consistency coefficient is a single quantitative index obtained by weighting the load matching degree and the speed matching degree according to a preset weight ratio. This index takes a value in the range of 0 to 1, and comprehensively reflects the overall similarity between the current measurement operating condition and the initial measurement operating condition. It is the core criterion for determining whether to adjust the sensor response gain.

[0051] Reverse correction refers to adjusting the sensor response sensitivity in the opposite direction to the deviation from the operating conditions when the operating condition consistency coefficient is lower than the preset threshold. That is, the more serious the deviation from the operating conditions and the more obvious the signal attenuation, the more the gain needs to be increased to compensate for the signal loss. This correction method ensures that the signal strength output by the sensor is comparable under different operating conditions.

[0052] Response gain refers to the multiple amplification factor of the sensor response sensitivity. The reference gain is 1. When the consistency of operating conditions decreases, the system increases the gain to more than 1 according to the preset compensation curve to compensate for signal attenuation. When the operating conditions are highly consistent, the gain is maintained at 1 without adjustment.

[0053] The preset compensation curve is a mapping rule stored internally by the system that represents the functional relationship between the operating condition consistency coefficient and the response gain amplification factor. This curve has a linear relationship in the intermediate matching region and tends to saturate in the severe mismatch region. It is used to smoothly map the numerical change of the consistency coefficient into a continuous adjustment of the gain.

[0054] The high threshold and the low threshold are two preset dividing points for determining the consistency coefficient of the working conditions. The high threshold is used to distinguish between highly consistent and intermediate matching states, while the low threshold is used to distinguish between intermediate matching and severe mismatch states. Together, the two thresholds divide the degree of similarity of the working conditions into three levels.

[0055] A linear compensation curve refers to a linear proportional relationship between the adjustment of the response gain and the change in the operating condition consistency coefficient within the intermediate matching region. That is, for every certain decrease in the consistency coefficient, the gain increases proportionally with a fixed slope. This linear relationship ensures the smoothness and continuity of the gain adjustment.

[0056] The pass / fail determination refers to the logical decision-making process of comparing the residual unbalance values ​​of each plane obtained after correction with the corresponding plane's preset acceptable value. Only when the residual values ​​of all planes are not greater than their respective acceptable values ​​is the fan blade judged as a qualified product, and any plane exceeding the standard is judged as a non-qualified product.

[0057] The acceptable value refers to the maximum residual imbalance that is allowed to exist on each correction plane according to the International Organization for Standardization's dynamic balance grade standard or user-defined requirements. This value is the qualification limit for determining whether the fan blade meets the balance quality requirements.

[0058] Data archiving refers to the process of writing all information related to the fan blade calibration process, including the fan blade model number, initial unbalance vector data, correction scheme parameters, post-correction measurement data, and spindle operating parameters during the measurement process, into the system database for permanent storage according to a preset database table structure, so that each calibration process can be completely traced.

[0059] The correction trend statistical result refers to the set of statistical characteristics, including the central tendency and dispersion of the correction amount, obtained by the system after filtering all historical qualified records of the same type of fan blade from the database and performing time series statistical processing on the correction weight data of each record.

[0060] The correction weight time sequence refers to a sequence of data formed by arranging the correction weight data of all qualified records of the same model in chronological order of measurement time. This sequence reflects the change trajectory of the correction amount of the fan blade of this model over the course of production time.

[0061] Sliding window mean calculation refers to setting a data window of fixed length, which slides gradually along the corrected weight time series. At each window position, the arithmetic mean of all data within the window is calculated and assigned to the center time of the window. The calculation results of all window positions form a smooth curve reflecting the central tendency of the correction amount as it changes over time.

[0062] The corrected mean series is a collection of all sliding window mean calculations. This series eliminates random fluctuations in individual data points and reflects the overall trend of the correction amount.

[0063] Sliding window standard deviation calculation refers to calculating the standard deviation of each corrected weight data within the window relative to the mean of the window at each position of the same sliding window. This standard deviation reflects the degree of dispersion and fluctuation of each correction amount around the central tendency within the time period. The calculation results of all window positions form a corrected dispersion sequence.

[0064] The corrected dispersion sequence is a collection of all sliding window standard deviation calculation results. This sequence reflects the change in the difference in correction amount between individual fan blades of this model over time.

[0065] Linear regression fitting refers to fitting an optimal straight line with the measurement time sequence as the independent variable and the mean in the corrected mean sequence as the dependent variable, so as to minimize the sum of squares of the vertical distances from each data point to the straight line along the time axis. The slope of this straight line is called the drift slope, which characterizes the systematic rate of change of the corrected mean per unit time.

[0066] The drift slope is a numerical expression of the slope of the fitted line in a linear regression. A positive slope indicates that the correction amount shows an increasing drift trend over time, while a negative slope indicates a decreasing drift trend. The larger the absolute value, the faster the drift speed. This slope is the core indicator for determining whether the equipment reference has undergone a systematic shift.

[0067] The residual refers to the vertical deviation of the actual value of each data point in the corrected mean sequence from the predicted value on the fitted regression line at the corresponding time. The magnitude of the residual for each data point reflects the intensity of the random fluctuations of the mean sequence around the trend line.

[0068] Mean fluctuation amplitude is a quantitative indicator of the overall distribution range or concentration obtained after statistical processing of all residual values. The larger the amplitude, the more significant the interference of random factors on the correction amount.

[0069] The monotonicity trend test refers to the process of determining whether the dispersion shows a continuous increasing or decreasing trend with the time of use by statistically analyzing the direction and magnitude of the difference between each two adjacent data points in the corrected dispersion sequence.

[0070] The direction of dispersion change is the output of the monotonicity trend test. This direction indicates whether the dispersion tends to increase, decrease, or remain basically unchanged over time. An increasing dispersion trend indicates that the consistency among individual fan blades is deteriorating.

[0071] The ternary statistical result is a combined statistic composed of three independent parameters: drift slope, mean fluctuation amplitude, and dispersion change direction. This combination jointly characterizes the complete picture of the correction trend from three different perspectives: system drift speed, random fluctuation intensity, and individual consistency change.

[0072] Benchmark update refers to the process where, when the statistical results of the correction trend indicate that the equipment benchmark has undergone a systematic shift that exceeds the allowable range, the system triggers the operator to remake a qualified benchmark standard part, remeasure and save the standard vibration characteristics of the new standard part as a new historical benchmark record, and discard the original standard part.

[0073] The working principle and technical effect of the above technical solution are as follows: Before starting the measurement, the operator takes out the independently sealed and stored reference standard part from the special storage container, installs it on the spindle chuck of the balancing machine, and precisely aligns the laser line emitted by the laser light module with the reference mark point on the standard part to complete the dual confirmation of spindle clamping and positioning and optical alignment. After the installation and alignment status are confirmed, the operator presses the balancing machine start button, and the equipment starts to run at the balancing speed preset by the system, and the spindle drives the standard part to rotate. During the rotation, the high-sensitivity vibration sensor installed on the machine base collects the vibration signal of the standard part in real time during the rotation. After signal conditioning and analog-to-digital conversion, the signal is input into the host computer software. The software extracts the measured vibration amplitude and phase data under the current state as the measured vibration characteristics. The system compares the measured vibration characteristics with the historical benchmark vibration characteristics of the standard part stored in the database when it was calibrated at the factory or when it was in good condition. Through difference calculation and deviation analysis, it quantitatively evaluates the comprehensive deviation of the current mechanical state, sensor system response characteristics and electrical drive characteristics of the equipment from the historical standard state. Finally, it outputs the benchmark deviation value before calibration in numerical form.

[0074] This invention addresses the technical problem in existing technologies where equipment lacks the ability to proactively assess its current state before each calibration cycle begins, and cannot predict whether the calibration starting point has shifted. It enables rapid pre-checking of equipment health and measurement accuracy before calibration, allowing operators to grasp the current degree and trend of equipment deviation before formal calibration begins. This improves the effectiveness and relevance of the calibration process, avoiding distorted calibration results caused by blindly executing standard calibration procedures when equipment already exhibits significant deviations. It reduces time and labor costs associated with ineffective calibrations due to unknown equipment status, and also mitigates the potential quality risk of mass production defects due to incorrect calibration benchmarks.

[0075] In one embodiment of the present invention, S1 includes: By taking out the sealed standard parts and installing them in the spindle chuck, and aligning the laser light with the reference point, the spindle positioning and light alignment status are confirmed, and a signal indicating that the standard parts installation is complete is obtained. By starting the equipment and running it at a preset speed, and with the vibration sensor collecting data synchronously, the standard parts are quickly verified and measured to obtain the measured vibration characteristics under the current state. By comparing and analyzing the measured vibration characteristics with historical benchmark records, the current degree of deviation of the equipment is quantitatively evaluated, and the benchmark deviation value before calibration is obtained.

[0076] The working principle and technical effect of the above technical solution are as follows: The vibration amplitude component is separated from the measured characteristic data, and its difference is calculated with the standard amplitude parameter at the corresponding equilibrium speed in the historical benchmark record to quantify the degree of deviation of the current equipment in the vibration energy response level. The output is the amplitude deviation in milligrams or millimeters per second. The system also separates the vibration phase component from the measured characteristic data, and its angle difference is calculated with the standard phase parameter at the corresponding speed in the historical benchmark record to quantify the phase drift of the current equipment in the vibration response time delay or lead level. The output is the phase deviation in angle. After completing the deviation quantification of the above two independent dimensions, the system combines the amplitude deviation and phase deviation according to the inherent complex domain expression of the vibration signal: using the amplitude deviation as the magnitude component of the complex vector and the phase deviation as the direction angle component of the complex vector, a two-dimensional complex domain deviation vector is constructed. This vector simultaneously represents the comprehensive intensity and dominant direction of the equipment's deviation from the standard state on the complex plane. The system calculates the magnitude of the complex-domain deviation vector, using it as a scalar index reflecting the overall deviation intensity of the equipment; it also calculates the argument of the vector, using it as a scalar index reflecting the dominant direction of the equipment deviation. Finally, the deviation intensity scalar and the deviation direction scalar are combined into a reference deviation vector in polar coordinates, so that the deviation state of the equipment can be fully characterized simultaneously in both numerical magnitude and spatial direction.

[0077] This invention solves the technical problem in existing technologies where standard part calibration results are output only as qualitative conclusions such as pass or fail, failing to express the specific degree and direction of equipment deviation from the standard state in a calculable numerical form. It achieves simultaneous quantification of the amplitude and phase of equipment deviation state in two dimensions, transforming the abstract pass / fail judgment into concrete two-dimensional vector parameters that can participate in subsequent mathematical calculations. This improves the quantitative level and numerical accuracy of deviation assessment before calibration, enabling the system to select different calibration correction strategies based on the specific magnitude of the deviation vector and determine the source type of the deviation based on its direction. It provides structured input data for the accurate calculation and embedding of compensation coefficients in subsequent calibration processes, significantly enhancing the pertinence and effectiveness of the entire calibration correction process.

[0078] In one embodiment of the present invention, the step of comparing and analyzing measured vibration characteristics with historical benchmark records to quantitatively assess the current degree of equipment deviation and obtain the benchmark deviation value before calibration includes: By performing difference calculation between the amplitude data in the measured vibration characteristics and the standard amplitude at the corresponding rotational speed in the historical benchmark record, the vibration energy deviation is quantified to obtain the amplitude deviation amount; By calculating the angle difference between the phase data in the measured vibration characteristics and the standard phase at the corresponding rotational speed in the historical benchmark record, the degree of phase drift is quantified, and the phase deviation is obtained. By combining the amplitude deviation and phase deviation into a two-dimensional deviation vector, the degree and direction of the equipment's current deviation from the standard state are comprehensively characterized, and a reference deviation vector is obtained.

[0079] Specifically, by combining the amplitude deviation and phase deviation into a two-dimensional deviation vector, the degree and direction of the equipment's current deviation from the standard state are comprehensively characterized, resulting in a reference deviation vector, which includes: By using the amplitude deviation as the magnitude component and the phase deviation as the direction angle component, a two-dimensional complex domain deviation vector is constructed to express the overall degree of deviation of the equipment from the standard state, thus obtaining the complex domain deviation vector. By calculating the magnitude of the deviation vector in the complex field, the overall strength of the equipment's current deviation from the standard state is quantified, and a deviation strength scalar is obtained. By calculating the argument of the deviation vector in the complex field, the main direction of the equipment's current deviation from the standard state is identified, and the deviation direction scalar is obtained. By combining the deviation intensity scalar and the deviation direction scalar into a deviation vector in polar coordinates, a two-dimensional comprehensive characterization of the degree and direction of the equipment's deviation from the standard state is performed, thus obtaining the reference deviation vector.

[0080] The working principle and technical effect of the above technical solution are as follows: The reference deviation value is embedded as a compensation coefficient into the equipment calibration algorithm, and the original calibration parameter matrix is ​​corrected and updated item by item to obtain a reliable measurement reference that has been compensated for the deviation and can truly reflect the actual response characteristics of the current equipment. The establishment of this reliable reference means that all subsequent measurement data are referenced to this compensated reference. The operator places the fan blade to be tested in the spindle chuck and aligns the laser beam with the reference point. The system starts the first measurement according to this reliable reference: the spindle drives the fan blade to accelerate to the preset equilibrium speed, the vibration sensor collects the vibration signal generated by the fan blade rotation, the servo motor encoder provides real-time feedback of the spindle speed data, the motor driver output current monitoring module synchronously records the drive current change curve, and the spindle positioning sensor records the position coordinates at each stop. The above vibration signals and three types of operating parameters are synchronously collected and stored in the cache. After the fan blade stops rotating, the system prompts the operator to manually rotate the fan blade 180 degrees and realign the laser beam with the reference point. Then, the second measurement is started, and the vibration signal and operating parameters in the reverse rotation state are collected again. After completing two measurements, the system performs a vector comparison and calculation with the vibration data from the two opposing measurements. Based on the basic principle of the influence coefficient method, it eliminates systematic errors such as eccentricity in the reference installation and outputs the weight magnitude and phase angle of the initial imbalance of the fan blade. The system integrates and processes the spindle speed data, drive current data, and position data recorded synchronously in the two measurements to form a set of operating parameters that comprehensively reflect the dynamic response characteristics of the fan blade under the combined action of the current spindle, drive, and sensing system. This set is defined as the dynamic characteristic fingerprint of the fan blade and spindle system.

[0081] This invention addresses the technical problem in existing technologies where the measurement process focuses solely on the imbalance result while neglecting the system state at the time of measurement, leading to a lack of comparability of measurement results for the same fan blade at different times. It achieves simultaneous acquisition of imbalance numerical measurements and system dynamic state records, ensuring that each imbalance data point is accompanied by complete system state information at the time of its generation. This improves the traceability and state correlation of measurement data, enabling subsequent data analysis to distinguish whether changes in measurement results stem from changes in the fan blade's own imbalance or from fluctuations in the system state. A quantifiable comparison benchmark is established for determining whether measurement conditions are consistent with the initial measurement in the verification process, significantly enhancing the consistency and comparability between multiple batches of measurements.

[0082] In one embodiment of the present invention, S2 includes: By embedding the benchmark deviation value as a compensation coefficient into the calibration algorithm, the original calibration parameters are corrected to obtain a verified reliable benchmark. The first measurement is initiated by a reliable benchmark. The vibration signal of the fan blade under test is collected and the spindle speed, drive current and position are recorded synchronously to obtain the first vibration data and operating parameters. By rotating the fan blades 180 degrees and re-aligning them with the light, a second measurement was initiated to collect the reverse vibration signal and obtain the opposing vibration data. By comparing and calculating the initial vibration data with the opposing vibration data, the weight and orientation of the unbalance are analyzed to obtain the initial unbalance. By integrating the spindle speed, drive current, and position recorded synchronously in two measurements into a combination of operating parameters, the dynamic characteristics of the fan blade spindle system are characterized, and a dynamic feature fingerprint is obtained.

[0083] The working principle and technical effect of the above technical solution are as follows: The drive current curves recorded in the initial measurement and the reversal measurement are superimposed and compared to analyze the fluctuation pattern of the current waveform within the fan blade rotation cycle. The fluctuation amplitude reflecting the fan blade load characteristics (i.e., the maximum amplitude of current fluctuation around the stable average value) and the fluctuation frequency (i.e., the number of repetition cycles of current fluctuation per unit time) are extracted. The fluctuation amplitude and fluctuation frequency are combined into a two-dimensional current sub-vector to numerically express the dynamic drive characteristics of the motor under the current fan blade load conditions. The system statistically processes all spindle speed data recorded in the two measurements, calculating the deviation between the average speed data and the theoretically set speed (i.e., the mean deviation) and the dispersion of each speed data point relative to the average (i.e., the standard deviation). The mean deviation and standard deviation are combined into a two-dimensional speed sub-vector to numerically express the spindle's speed holding capability and stability under the current load and drive conditions. The system performs a repeatability comparison analysis on the position data recorded when the spindle stops during two measurements. It extracts the maximum deviation (maximum deviation) relative to the target position from all stopping positions, along with the statistical distribution range of the deviation values ​​for each stopping position. The maximum deviation and the deviation distribution range are combined into a two-dimensional positioning sub-vector to numerically represent the repeatability and reliability of the spindle positioning system. After extracting the above three two-dimensional sub-vectors, the system concatenates the six independent parameters into a complete six-dimensional feature vector according to a preset vector dimension order: current sub-vector first, speed sub-vector in the middle, and positioning sub-vector last. This six-dimensional vector constitutes a unique identifier in numerical space for the dynamic response characteristics of a specific fan blade under the current specific spindle state. That is, different fan blade and spindle combinations or different equipment states will produce different six-dimensional vector numerical distributions, while repeated measurements of the same fan blade under the same state will produce highly consistent six-dimensional vectors.

[0084] This invention solves the technical problem in existing technologies where operating parameters are stored in a scattered manner as raw waveforms or scattered numerical values, making it impossible to form a unified structured identifier for state comparison. It achieves multi-dimensional, structured, and unique numerical encoding and expression of the dynamic characteristics of the fan blade and spindle system, unifying the previously scattered drive current, speed fluctuations, and positioning deviations into a single vector space for overall representation. This improves the comparability of system states between different measurement batches, enabling the system to quickly determine whether the current state is consistent with a historical state through simple vector distance calculation. It provides a structured, multi-dimensional fusion comparison benchmark that can directly participate in numerical calculations for determining the consistency of operating conditions in the adaptive matching process, significantly improving the efficiency and accuracy of operating condition similarity judgment.

[0085] In one embodiment of the present invention, the dynamic characteristics of the fan blade spindle system are characterized by integrating the spindle speed, drive current, and position recorded synchronously in two measurements into a combination of operating parameters to obtain a dynamic feature fingerprint, including: By superimposing and comparing the drive current curves recorded in the initial measurement and the flip measurement, the load fluctuation pattern of the motor when the fan blades rotate is extracted to obtain the current fluctuation characteristics. By calculating the mean and standard deviation of the spindle speed data recorded in the initial measurement and the overturning measurement, the speed stability is quantified, and the speed stability characteristics are obtained. By comparing the repeatability of the position data recorded in the initial measurement and the flip measurement, the deviation range of the spindle's stop position each time is quantified to obtain the positioning repeatability characteristics. By combining and encoding current fluctuation characteristics, rotational speed stability characteristics, and positioning repeatability characteristics into a multi-dimensional feature vector, the dynamic response characteristics of the fan blade under the current spindle state are uniquely identified, thus obtaining a dynamic feature fingerprint.

[0086] Specifically, by combining and encoding current fluctuation characteristics, rotational speed stability characteristics, and positioning repeatability characteristics into a multi-dimensional feature vector, the dynamic response characteristics of the fan blade under the current spindle state are uniquely identified, resulting in a dynamic feature fingerprint, including: By extracting the fluctuation amplitude and fluctuation frequency from the current fluctuation characteristics into a two-dimensional current sub-vector, the dynamic characteristics of the motor load are numerically expressed to obtain the current sub-vector; By extracting the mean offset and standard deviation from the speed stability characteristics into a two-dimensional speed sub-vector, the spindle speed holding capability is numerically expressed, and the speed sub-vector is obtained. By extracting the maximum deviation and deviation distribution range from the positioning repeatability features into a two-dimensional positioning sub-vector, the main axis stopping accuracy is numerically expressed to obtain the positioning sub-vector; By concatenating the current sub-vector, rotation speed sub-vector, and positioning sub-vector into a six-dimensional feature vector in a preset dimensional order, a unique numerical identifier is obtained for the dynamic response characteristics of the fan blade under the current spindle state, thus obtaining a dynamic feature fingerprint.

[0087] The working principle and technical effect of the above technical solution are as follows: The specific angle position for adding trial weights on the first correction plane is determined. A trial weight of predetermined weight is added at the zero-degree position of this plane. Measurements are then initiated to complete two data acquisitions (one forward and one reverse), recording the vibration response changes after the addition of the trial weight. After completing the trial weight measurement on the first plane, the operator removes the trial weight from the first plane. The system determines the trial weight addition position on the second correction plane based on the initial unbalance. A trial weight of the same weight is added at the zero-degree position of the second plane, and the same two forward and reverse measurements are completed, recording the vibration response changes. After completing all trial weight measurements on both planes, the system comprehensively calculates the vibration vector data before and after each trial weight. Based on the mathematical principle of the dual-plane influence coefficient method, the system calculates the required correction weight, correction angle position, and the direction of operation (adding or removing mass) for each of the first and second planes, and integrates these three parameters to output a complete correction scheme. Simultaneously with the generation of the correction scheme, the system extracts the dynamic feature fingerprint stored in Example 3 and compares it item by item with the current spindle operating parameters to determine whether the system state of the current verification stage is consistent with the system state during the initial measurement. If there is a difference, the vibration sensing response coefficient of the verification stage is adaptively adjusted according to the degree of difference to ensure that the sensing sensitivity during the verification measurement maintains the same system response conditions as the initial measurement. After completing the adaptive matching, the operator performs the actual correction operation on the fan blades according to the correction scheme, then places the corrected fan blades on the spindle, starts the final measurement under the matched measurement conditions, collects the corrected vibration data, and outputs the corrected accurate measurement data after system calculation.

[0088] This invention addresses the technical problem in existing technologies where the generation of correction schemes and the execution of verification measurements are independent, and the measurement conditions used in the verification phase differ from those used in the initial measurement, making it difficult to accurately assess the correction effect. It achieves coordinated and adaptive adjustment of correction scheme generation and measurement conditions, enabling the corrected verification measurement to be completed under the same system response conditions as the initial measurement. This improves the accuracy and reliability of the corrected measurement data, allowing operators to be confident that the verification results reflect the true effect of the correction operation itself rather than spurious deviations caused by changes in system state. It reduces the probability of misjudging the correction effect due to inconsistencies in measurement conditions, thereby reducing the number of repeated, ineffective trial corrections due to misjudgments, and significantly improving the success rate of single corrections and overall calibration efficiency.

[0089] In one embodiment of the present invention, S3 includes: The test weight position of the first correction plane is determined by the initial unbalance. A predetermined test weight is added at zero degrees and two measurements are completed in both directions. The vibration response change of the first plane is recorded to obtain the data before and after the test weight of the first plane. After removing the test weight of the first plane, the position of the test weight of the second corrected plane is determined based on the initial unbalance. The same test weight is added at zero degrees and two measurements are completed in both directions. The changes in the vibration response of the second plane are recorded to obtain the data before and after the second plane test weight. By comprehensively analyzing the differences in data before and after each trial weight on the two planes, the corrected weight, angle, and direction are analyzed to obtain a correction scheme. By comparing and matching the dynamic feature fingerprint with the current spindle operating parameters, the sensing response coefficient of the verification process is adjusted to obtain the measurement conditions after adaptive matching. After performing the operation according to the correction scheme, the measurement is started under matching conditions to collect the corrected vibration data and obtain the corrected accurate measurement data.

[0090] The working principle and technical effect of the above technical solution are as follows: The system compares the difference between the drive current parameters collected in real time during the current verification measurement and the current fluctuation characteristics stored in the dynamic feature fingerprint during the initial measurement. Based on the magnitude of the difference, it determines the degree of deviation between the current motor load state and the load state during the initial measurement, and outputs the load matching degree expressed as a percentage or dimensionless coefficient. The system compares the deviation between the currently collected spindle speed fluctuation parameters and the speed stability characteristics stored in the fingerprint during the initial measurement. Based on the magnitude of the deviation, it determines the degree of difference between the current spindle speed holding state and the speed holding state during the initial measurement, and outputs the speed matching degree. After completing the independent calculation of the above two matching degrees, the system assigns corresponding weight ratios to the load matching degree and the speed matching degree according to preset weighting coefficients, and performs a weighted comprehensive calculation to obtain a comprehensive coefficient that can reflect the degree of similarity between the current measurement condition and the initial measurement condition, namely the condition consistency coefficient. The closer the coefficient is to 1, the higher the condition consistency; the closer it is to 0, the more serious the condition deviation. After obtaining the operating condition consistency coefficient, the system compares it step by step with preset high and low thresholds: if the consistency coefficient is greater than or equal to the high threshold, the current operating condition is determined to be highly consistent with the initial operating condition, and no adjustment of the sensor response sensitivity is required; the original sensor gain is maintained for subsequent measurements. If the consistency coefficient is in the middle range between the high and low thresholds, it is determined to be a medium matching state. The system gradually increases the sensor response gain proportionally as the consistency coefficient decreases, according to the preset linear compensation curve rules, so that the sensor sensitivity is adequately compensated to make up for the signal loss caused by the difference in operating conditions. If the consistency coefficient is less than or equal to the low threshold, it is determined to be a severe mismatch state. The system increases the sensor response gain to the highest level according to the preset maximum compensation ratio, and at the same time sends an abnormal operating condition prompt message to the operator interface, reminding them to check the equipment status or recalibrate the reference.

[0091] This invention solves the technical problems of existing technologies where the sensor response sensitivity is fixed and the same gain coefficient is used regardless of changes in measurement conditions, leading to signal distortion or signal-to-noise ratio degradation when the operating conditions deviate. It achieves graded, segmented, and reverse adaptive adjustment of the sensor response sensitivity based on the consistency between the current and initial operating conditions, ensuring that the sensor outputs comparable signal strengths under different operating conditions. This improves the adaptability and signal quality stability of the sensing system across the entire operating range, avoiding signal saturation distortion or weak signal acquisition problems caused by fixed gain when operating conditions change. It reduces the probability of unusable measurement data due to operating condition mismatch and the resulting time and cost of repeated measurements. Furthermore, an anomaly alert function proactively notifies the operator of severe operating condition mismatches, transforming equipment maintenance from reactive post-event handling to proactive pre-event warning.

[0092] In one embodiment of the present invention, the step of comparing and matching dynamic feature fingerprints with current spindle operating parameters to adjust the sensing response coefficients in the verification process and obtain adaptively matched measurement conditions includes: By comparing the difference between the real-time drive current in the current spindle operating parameters and the current fluctuation characteristics in the dynamic feature fingerprint, the consistency between the current load state and the initial measurement is determined, and the load matching degree is obtained. By comparing the real-time speed fluctuations in the current spindle operating parameters with the speed stability features in the dynamic feature fingerprint, the degree of difference between the current stable speed state and the initial measurement is quantified, and the speed matching degree is obtained; By weighting and combining load matching degree and speed matching degree, the overall similarity between the current measurement condition and the initial measurement condition is quantitatively evaluated to obtain the condition consistency coefficient. The sensor response sensitivity in the verification process is reversed by using the consistency coefficient of the operating conditions. When the consistency coefficient is lower than the threshold, the response gain is increased according to the preset compensation curve. When the consistency coefficient is higher than the threshold, the original gain is maintained, thus obtaining the measurement conditions after adaptive matching.

[0093] Specifically, the sensor response sensitivity in the verification process is inversely corrected using the consistency coefficient of operating conditions. When the consistency coefficient is below a threshold, the response gain is increased according to a preset compensation curve; when the consistency coefficient is above the threshold, the original gain is maintained, thus obtaining the measurement conditions after adaptive matching, including: By comparing the working condition consistency coefficient with a preset high threshold, a highly consistent state is determined when the consistency coefficient is greater than or equal to the high threshold, maintaining the original sensing response sensitivity and obtaining measurement conditions under high matching conditions.

[0094] By comparing the operating condition consistency coefficient with a preset low threshold, a severe mismatch is determined when the consistency coefficient is less than or equal to the low threshold. The response gain is increased by the preset maximum compensation ratio and an abnormal operating condition prompt is issued to obtain the measurement conditions under low matching conditions.

[0095] By placing the operating condition consistency coefficient in an intermediate state between high and low thresholds, and gradually increasing the response gain as the consistency coefficient decreases according to a preset linear compensation curve, the measurement conditions under medium matching conditions are obtained.

[0096] By implementing corresponding gain adjustment strategies based on three judgment results—high consistency, severe mismatch, and intermediate matching—the sensor response sensitivity in the verification stage is corrected by reverse compensation to match the current operating conditions, thereby obtaining the measurement conditions after adaptive matching.

[0097] The working principle and technical effect of the above technical solution are as follows: The system retrieves the corrected precise measurement data from the cache and compares the residual unbalance of the first plane, the residual unbalance of the second plane, and the residual unbalance of the composite plane with the corresponding acceptable values ​​preset in the system setting page or automatically generated by an international standard calculator. Only when the residual unbalance of all planes is less than or equal to its corresponding acceptable value, the system determines that the fan blade is a qualified product and outputs the qualification judgment result; if the residual unbalance of any plane exceeds the corresponding acceptable value, it is determined to be a non-qualified product, and the system prompts the operator to return to the correction stage to re-execute the correction and verification process. After the qualification judgment is completed, the system writes all measurement and correction data related to the fan blade, including the fan blade model and number information, the initial unbalance data obtained in Example 3, the correction scheme data generated in Example 5, the corrected precise measurement data obtained in Example 5, and the spindle running parameters recorded in Example 3, into the corresponding fields of the system database one by one according to the preset database table structure, completing the archiving and storage of the complete measurement process data. After data archiving is completed, the system filters all historical qualified records of the same model as the fan blade currently under test from the database, extracts the corrected weight data from each record, arranges them in chronological order of measurement to form a corrected weight time series, and then calculates the mean and standard deviation of this time series. The mean reflects the overall central tendency of the correction amount for this model of fan blade, and the standard deviation reflects the dispersion and fluctuation of each correction amount around the mean, forming a preliminary correction trend statistical result. The system compares the correction trend statistical result with the preset systematic offset threshold. If the mean or standard deviation in the statistical result exceeds the allowable offset range, it is determined that the equipment reference may have drifted. The system immediately issues an early warning signal and displays a prompt message on the operation interface, while triggering the reference standard part update process, requiring the operator to remake and seal the new reference standard part to replace the original standard part.

[0098] This technology solves the technical problems of existing calibration processes ending at the pass / fail determination stage, measurement data being idle after the current determination, and failing to provide reverse decision-making basis for subsequent equipment maintenance. It achieves a comprehensive upgrade from a unidirectional "measurement, correction, determination" process to a closed-loop "measurement, correction, determination, feedback" process, ensuring that each measurement not only serves the quality determination of the current fan blade but also supports continuous monitoring and maintenance decisions regarding the equipment's own status. It enhances the secondary utilization value of data resources, transforming accumulated historical measurement data from static archives into a dynamic source of equipment status monitoring information. It improves the initiative and foresight of equipment calibration and maintenance, making benchmark updates no longer dependent on passive execution at fixed intervals but based on actual data-driven triggers. It reduces the quality risks and economic losses from the continuous mass production of defective products due to undetected benchmark drift.

[0099] In one embodiment of the present invention, S4 includes: By comparing the corrected precise measurement data with the system's preset acceptable values ​​item by item, the residual unbalance of each plane is judged to determine whether it is within the allowable range, and the pass / fail judgment result is obtained. By writing all the information, including blade information, initial imbalance, correction scheme, post-correction measurement data and operating parameters, into the database, the complete measurement process data is archived and data records are obtained. By extracting the cumulative correction amount of the same type of fan blade from the data records and statistically analyzing its mean and dispersion, the correction trend is analyzed and the statistical results of the correction trend are obtained. By comparing the corrected trend statistics with the preset systematic offset threshold, it is determined whether the benchmark has drifted, an early warning signal is obtained, and the benchmark is updated.

[0100] The working principle and technical effect of the above technical solution are as follows: The system performs a structured query from the archived database records, filters out all qualified records belonging to the same product model as the fan blade currently under test, sorts them in ascending order of time, and reads the corrected weight data field of each record one by one. These data are then arranged sequentially according to the measurement date and time, forming a corrected weight time series sequence for that product model on the continuous production timeline. The length of this sequence is the cumulative number of qualified products of that model up to the current time. After obtaining the corrected weight time series sequence, the system performs a sliding window mean calculation: a fixed-length sliding window is set, starting from the first data point in the time series, and sliding one data point forward each time. At each window position, the arithmetic mean of all corrected weight data within the window is calculated, and this value is used as the central tendency representation value of the corrected amount for that window. The mean calculation results of all windows are arranged according to the center time point of the window to form a corrected mean sequence. The system calculates the standard deviation of the same corrected weight time series using a sliding window: at each position within the same sliding window, the standard deviation is calculated to reflect the dispersion of each corrected weight data point relative to the mean of that window. This standard deviation reflects the random fluctuation amplitude of the correction amount within that window period. The standard deviation calculation results of all windows are arranged to form a corrected dispersion sequence. After extracting the mean sequence and the dispersion sequence, the system performs a joint comparative analysis of the two sequences: by observing the overall trend of the corrected mean sequence over time, it identifies whether there is a directional drift trend that is continuously rising or falling; by observing the width change of the corrected dispersion sequence over time, it identifies whether the random fluctuation amplitude is gradually increasing. Thus, the change of the correction amount over time is decomposed into two different types of trend components: systematic drift with a clear direction and random fluctuation without a clear direction.

[0101] This invention addresses the technical problems of existing technologies, such as storing archived data only in raw record form, lacking the ability to mine and extract trends from cumulative data over time, and burying a large amount of useful information. It achieves the extraction of time-series statistical features and preliminary separation of trend components from cumulative correction data, transforming the chaotic raw correction data into a mean and dispersion sequence with clear statistical significance. This enhances the secondary utilization value and information density of historical data resources, enabling the system to grasp the overall trend of correction values ​​for the same product model from a macro perspective. It provides statistically processed intermediate data products for the accurate separation criteria of subsequent systematic drift and random fluctuations, significantly reducing the probability of misjudgment due to excessive data noise when directly using raw data for trend determination.

[0102] In one embodiment of the present invention, the step of extracting the cumulative correction amount of the same type of fan blade from data records and statistically analyzing its mean and dispersion to obtain statistical results of the correction trend includes: By filtering all qualified records of the same type of fan blade from the data records, the corrected weight data in each record is arranged in chronological order to obtain the corrected weight time sequence. By calculating the sliding window mean of the corrected weight time series, the central tendency of the correction amount within each window is extracted to obtain the corrected mean series. By calculating the standard deviation of the corrected weight time series using a sliding window, the degree of fluctuation of the correction amount within each window is quantified, and the corrected dispersion series is obtained. By jointly analyzing the corrected mean sequence and the corrected dispersion sequence, the directional drift trend and non-directional fluctuation trend of the correction amount over time are separated, and the statistical results of the correction trend including the drift slope and fluctuation amplitude are obtained.

[0103] Specifically, by jointly analyzing the corrected mean sequence and the corrected dispersion sequence, the directional drift trend and non-directional fluctuation trend of the correction over time are separated, obtaining statistical results of the correction trend including the drift slope and fluctuation amplitude, including: By performing linear regression fitting on the modified mean sequence in chronological order, the slope of the fitted line is extracted to obtain the drift slope that characterizes the rate of systematic directional drift.

[0104] By extracting the residuals of each data point in the modified mean sequence relative to the fitted regression line, the dispersion of the mean sequence around the trend line is quantified, and the mean fluctuation amplitude is obtained.

[0105] By performing a monotonicity trend test on the corrected dispersion sequence in chronological order, it is determined whether the dispersion shows an increasing trend with the usage time, thus obtaining the direction of dispersion change.

[0106] By combining the drift slope, mean fluctuation amplitude, and dispersion change direction into a ternary statistical result, the directional drift and non-directional fluctuation of the correction amount over time are separated and characterized, and the statistical result of the correction trend including the drift slope and fluctuation amplitude is obtained.

[0107] The working principle and technical effect of the above technical solution are as follows: The system uses the corrected mean sequence as input data, the measurement time sequence as the independent variable, and the corrected mean as the dependent variable. It performs a univariate linear regression fitting calculation on the sequence, that is, it finds the best-fitting straight line that minimizes the sum of the squares of the vertical distances from each data point to the line along the time axis. After fitting, the system extracts the slope parameter from the analytical expression of the fitted straight line. This slope value is the drift slope, which characterizes the rate at which the corrected mean undergoes systematic change along the time axis per unit time: a positive slope indicates that the correction amount has an increasing drift trend, a negative slope indicates that the correction amount has a decreasing drift trend, and the larger the absolute value of the slope, the faster the drift speed. The system calculates the deviation of each data point in the corrected mean sequence from the fitted regression line. This is achieved by subtracting the predicted value for that point at the corresponding time from the actual value at each point, resulting in a residual value. Then, all residual values ​​are statistically processed to calculate their overall distribution range and concentration, outputting a mean fluctuation amplitude parameter. This parameter reflects the intensity of the corrected mean sequence's fluctuations around the trend line; a larger fluctuation amplitude indicates a more significant influence of random factors. The system performs a monotonicity trend test on the corrected dispersion sequence in chronological order: it calculates the difference between every two adjacent data points in the dispersion sequence and statistically analyzes the positive and negative sign distribution of all adjacent differences. If the number and average amplitude of positive differences are significantly greater than those of negative differences, the dispersion is determined to show an increasing trend over time; otherwise, it shows a decreasing trend. This determination is output as a dispersion change direction parameter. Ultimately, the system combines the three parameters—drift slope, mean fluctuation amplitude, and dispersion change direction—into a set of ternary statistical results. The drift slope reflects the systematic, directional, long-term drift trend; the mean fluctuation amplitude reflects the degree of short-term, undirected random fluctuation; and the dispersion change direction reflects whether the difference in correction amount among different individual fan blades is increasing over time. These three parameters work together from different perspectives to quantitatively separate and characterize the deterministic and random components in the trend of correction amount change.

[0108] This invention solves the technical problem in existing technologies where it is impossible to distinguish whether changes in correction values ​​are caused by systematic drift of the equipment baseline or by random factors such as raw material batch fluctuations and environmental changes, thus making it impossible to accurately determine whether a baseline update should be triggered. It achieves quantitative separation of systematic drift and random fluctuation components in the trend of correction value changes, enabling the system to quantify deterministic drift velocity and random fluctuation amplitude separately. This improves the distinguishability and recognition accuracy between baseline drift determination and random fluctuation misjudgment, ensuring that the system only triggers a baseline update when the drift slope truly exceeds a threshold, avoiding false alarms caused by random fluctuation peaks. Simultaneously, the introduction of a dispersion change direction parameter allows the system to identify whether equipment performance is gradually deteriorating, providing additional quantitative evidence for predictive maintenance, significantly reducing the probability of false alarms and missed alarms, and improving the reliability and practicality of early warning signals.

[0109] According to one embodiment of the present invention, the system includes: The measurement and analysis module is used to quickly verify and analyze the equipment using independently stored reference standard parts to obtain the reference deviation value before calibration. The imbalance feature analysis module is used to correct the equipment calibration process by using the benchmark deviation value to obtain a verified reliable benchmark. The module then uses the reliable benchmark to perform initial imbalance measurement and analysis on the fan blade under test and simultaneously records the spindle operating parameters to obtain the initial imbalance amount and dynamic feature fingerprint. The adaptive matching module is used to perform trial weight measurements on each correction plane based on the initial imbalance and generate a correction scheme. At the same time, it adaptively matches the sensing response coefficients of the verification stage through dynamic feature fingerprints to obtain accurate measurement data after correction. The periodic verification module is used to complete the qualification judgment by comparing the accurate measurement data with the acceptance value, and archive the complete data. The module performs periodic verification on the reference standard parts by the accumulated correction trend of the same model in the archived data, obtains the systematic deviation warning signal and triggers the reference update.

[0110] The working principle and technical effect of the above technical solution are as follows: Before the system starts measurement, the operator takes out the independently sealed and stored reference standard part from the special storage container, installs it on the spindle chuck of the balancing machine, and precisely aligns the laser line emitted by the laser light module with the reference mark point on the standard part to complete the dual confirmation of spindle clamping and positioning and optical alignment. After the installation and alignment status are confirmed, the operator presses the balancing machine start button, and the equipment starts to run at the preset balancing speed, with the spindle driving the standard part to rotate. During the rotation, the high-sensitivity vibration sensor installed on the machine base collects the vibration signal of the standard part in real time during rotation. After signal conditioning and analog-to-digital conversion, the signal is input into the host computer software, which extracts the measured vibration amplitude and phase data of the current state as the measured vibration characteristics. The system compares the measured vibration characteristics with the historical benchmark vibration characteristics of the standard part stored in the database when it was calibrated at the factory or when it was in good condition. Through difference calculation and deviation analysis, it quantitatively evaluates the comprehensive deviation of the current mechanical state, sensor system response characteristics and electrical drive characteristics of the equipment from the historical standard state. Finally, it outputs the benchmark deviation value before calibration in numerical form.

[0111] This system addresses the technical problem in existing technologies where equipment lacks the ability to proactively assess its current state before each calibration cycle begins, and cannot predict whether the calibration starting point has shifted. It enables rapid pre-checking of equipment health and measurement accuracy before calibration, allowing operators to grasp the current degree and trend of equipment deviation before formal calibration begins. This improves the effectiveness and relevance of the calibration process, avoiding distorted calibration results caused by blindly executing standard calibration procedures when significant equipment deviations already exist. It reduces time and labor costs associated with ineffective calibrations due to unknown equipment status, and also mitigates the potential quality risk of mass production defects due to incorrect calibration benchmarks.

[0112] Example 2 This embodiment further combines the verification results of the benchmark standard parts, dynamic feature fingerprints, common variations of different fan blade models, production batch boundaries, the correlation between the initial imbalance and the correction amount, and the dual-plane correction direction to distinguish the sources of abnormal correction trends.

[0113] The fan dynamic balance correction method based on multi-source sensor fusion provided in this embodiment includes steps S1 to S5.

[0114] S1. Perform rapid verification measurements on the equipment using reference standard parts to obtain the reference deviation value, and correct the equipment calibration parameters based on the reference deviation value to establish a reliable reference.

[0115] S2. Perform the first measurement on the fan blade under test and the second measurement after rotating it 180 degrees to obtain the initial imbalance. Simultaneously record the spindle speed, drive current and position to form a dynamic feature fingerprint.

[0116] S3. Perform trial weight measurements on the first and second correction planes respectively, generate correction schemes, and obtain accurate measurement data after correction under conditions that match the initial measurement conditions.

[0117] S4. Complete the qualification determination and data archiving, and obtain the corrected weight time series, corrected mean series, corrected dispersion series and corrected trend statistical results from the continuous qualified records of the same type of fan blade.

[0118] S5. Extract the persistence of the correction trend and the consistency of the dual-plane direction based on the statistical results of the correction trend. Perform correlation analysis between the persistence of the correction trend and the consistency of the dual-plane direction and the equipment-side state parameters and the product-side state parameters, respectively, to obtain the confidence level of equipment baseline drift and the confidence level of product batch offset. Determine the anomaly attribution based on the relative relationship between the two confidence levels, and perform baseline verification, equipment recalibration, production batch warning or continued observation, respectively.

[0119] Specifically, step S5 includes steps S51 to S53.

[0120] S51. Determine the reliability of the equipment reference drift based on the degree of persistence of the correction trend, the degree of consistency of the dual-plane direction, the degree of deviation of the reference standard, the degree of deviation of the spindle dynamic feature fingerprint, and the degree of cross-model same-direction change.

[0121] First, the direction and magnitude of change of multiple consecutive sliding windows are extracted from the corrected mean sequence. In this embodiment, each sliding window includes measurement records of thirty consecutive qualified fan blades of the same model, and adjacent sliding windows are shifted forward by five measurement records.

[0122] When the corrected mean changes in the same direction in three or more consecutive sliding windows, and the magnitude of the change continuously exceeds the normal fluctuation range under the historical stable state of the fan blade model, the degree of correction trend persistence is increased; when the corrected mean changes only in a single sliding window, or the direction of change in adjacent sliding windows changes repeatedly, the degree of correction trend persistence is decreased.

[0123] The corrected weight, corrected angle, and direction of adding or removing mass are read from the first and second corrected planes, respectively, and the changes in the corrected vectors of the two corrected planes are compared. When the corrected amounts of the first and second corrected planes increase or decrease simultaneously, and the direction of change of the corrected angle remains stable, the consistency of the two plane directions is increased; when the directions of change of the corrected amounts of the two corrected planes are opposite, or only one corrected plane changes significantly, the consistency of the two plane directions is decreased.

[0124] Retrieve the verification records of the benchmark standard within the time period corresponding to the abnormal trend, and compare the current measured vibration amplitude and phase of the benchmark standard with the historical benchmark records. When both amplitude and phase deviations increase simultaneously and persist, increase the deviation level of the benchmark standard; when the vibration response of the benchmark standard is still within the historical allowable range, decrease the deviation level of the benchmark standard.

[0125] The dynamic characteristic fingerprints before and after the occurrence of abnormal trends are retrieved, and the fluctuation amplitude and frequency of the drive current, the mean deviation and standard deviation of the spindle speed, and the maximum deviation and distribution range of the positioning position are compared. When multiple parameters among the drive current, spindle speed, and positioning position deviate from their historical stable states simultaneously, the deviation degree of the spindle dynamic characteristic fingerprint is increased; when the operating parameters do not change significantly, the deviation degree of the spindle dynamic characteristic fingerprint is decreased.

[0126] Further, records of other fan blade models that completed calibration within the same device and time period were extracted from the database. Since the correction weight ranges differ for different fan blade models, the correction amount was first normalized according to the historical average and historical fluctuation range of each model, and then the correction trends of different fan blade models were compared. When multiple models showed offsets in the same direction within a similar timeframe, the degree of cross-model unidirectional change was increased; when the anomaly only existed in the current model, the degree of cross-model unidirectional change was decreased.

[0127] The reliability of equipment reference drift is determined as follows: the degree of deviation of the reference standard part is used as the main judgment basis, the degree of deviation of the spindle dynamic feature fingerprint and the degree of cross-model same direction change are used as the equipment side supplementary basis, and the degree of persistence of correction trend and the degree of consistency of dual plane direction are used as the constraint conditions for whether the anomaly has systematicity and continuity.

[0128] When the degree of persistence of the correction trend and the degree of consistency of the dual-plane direction are high, and two or more of the following factors increase simultaneously: the degree of deviation of the reference standard, the degree of deviation of the spindle dynamic characteristic fingerprint, and the degree of change in the same direction across models, the reliability of the equipment reference drift is determined to be at a higher level; when only the correction trend changes, but the reference standard, the spindle operating status, and other models do not change accordingly, the reliability of the equipment reference drift is determined to be at a lower level.

[0129] The above processing can avoid directly assuming that the equipment reference has drifted based solely on changes in the correction amount of a certain type of fan blade.

[0130] S52. Based on the same degree of persistence of correction trend and degree of consistency of dual-plane direction as in step S51, and combined with the degree of concentration of batch boundary, degree of correlation between initial imbalance and correction amount, degree of model limitation and degree of concentration of correction angle, determine the reliability of product batch offset.

[0131] The production batch identifier is read from the data records of each fan blade, and the abnormal start position of the corrected mean sequence is compared with the production batch switching position. When the abrupt change position of the corrected mean or corrected angle distribution is close to the production batch switching position, and subsequent fan blades in the batch continue to maintain a similar direction of change, the batch boundary concentration is increased; when the abnormality forms slowly within the same production batch, or when there is no obvious correspondence between the abnormality start position and the batch switching position, the batch boundary concentration is decreased.

[0132] The initial imbalance of each blade is correlated with the correction weight and correction angle in the corresponding correction scheme. When the production batch changes, if the initial imbalance and correction weight increase or decrease synchronously, and the direction of change of the two is consistent in multiple consecutive blades, the correlation between the initial imbalance and the correction is increased; if the correction weight changes but the initial imbalance does not change accordingly, the correlation is decreased.

[0133] The correction trend of the current model of fan blades is compared with the correction trends of other models of fan blades within the same time period. When only the current model shows a continuous shift, while the correction mean, correction dispersion, and residual imbalance of other models are still within their respective historical stable ranges, the model limitation is increased; when multiple models show changes in the same direction at the same time, the model limitation is decreased.

[0134] Based on the fan blade reference point, the correction angles of the first and second correction planes are mapped to the corresponding circumferential positions of the fan blade. When the correction positions of most fan blades in the same production batch are concentrated in similar circumferential areas or in the same structural parts of the fan blade, the concentration of correction angles is increased; when the correction angles are randomly dispersed in the circumferential direction of the fan blade, the concentration of correction angles is decreased.

[0135] The reliability of product batch offset is determined as follows: the degree of concentration of batch boundaries and the correlation between the initial imbalance and the correction amount are used as the main judgment criteria, the degree of model limitation and the degree of concentration of correction angle are used as supplementary criteria on the product side, and the degree of persistence of correction trend and the degree of consistency of dual-plane direction are used as constraints on trend stability.

[0136] When the correction trend persists, the changes in the two correction planes are consistent, the anomaly start position corresponds to the production batch switching position, the initial imbalance amount changes synchronously with the correction amount, and the anomaly is mainly limited to the current model, the product batch offset confidence level is determined to be high. When the correction angle is scattered, the anomaly is unrelated to the batch switching, or multiple models show the same change at the same time, the product batch offset confidence level is determined to be low.

[0137] Therefore, the determination of product batch offset does not rely solely on the production batch identifier, but is jointly determined by multiple physically meaningful parameters such as production batch, initial imbalance state, correction position, and model range.

[0138] S53. Based on the relative relationship between the reliability of equipment baseline drift and the reliability of product batch offset, determine the attribution type of the correction trend anomaly.

[0139] When both the equipment baseline drift confidence level and the product batch offset confidence level are at a low level, it indicates that both equipment-side and product-side evidence are insufficient. The current anomaly is judged as random fluctuation, and no baseline update or production batch handling is triggered. Subsequent sliding window data collection continues. If the subsequent correction trend returns to the historical stable range, the anomaly label is removed; if the anomaly persists, steps S51 to S53 are re-executed.

[0140] When the reliability level of the equipment reference drift reaches a high level and is significantly higher than the reliability level of the product batch offset, the current anomaly is determined to originate primarily from the equipment reference drift. The system suspends continued batch calibration using the current reliable reference and calls upon the reference standard to perform rapid verification measurements again. If the amplitude and phase deviations of the reference standard still exceed the preset range, the equipment calibration parameters are updated, and a new dynamic feature fingerprint is generated after recalibration. For fan blades that have already been calibrated during the anomaly period, a review list is generated based on the data records in the database.

[0141] When the product batch offset confidence level reaches a high level and is significantly higher than the equipment reference drift confidence level, it is determined that the current anomaly mainly originates from the product batch offset. The system maintains the current reliable reference and sensor response conditions, does not perform reference standard component updates, and outputs production batch warning information to the production end, including the fan blade model, production batch identifier, anomaly start time, main correction position of the first correction plane, and main correction position of the second correction plane. It also increases the sampling retesting ratio of the current batch of fan blades.

[0142] When both the equipment baseline drift confidence level and the product batch offset confidence level reach a high level and are close in magnitude, it is determined to be an anomaly involving both equipment and product. The system first uses independently stored reference standard parts to verify the equipment baseline status, and then extracts multiple fan blades from the current production batch for re-clamping and measurement. If the reference standard part verification results remain abnormal, the equipment is recalibrated first, and then the extracted fan blades are re-measured to determine the impact of equipment baseline changes and product batch changes on the correction trend.

[0143] When either confidence level is in the middle range, or the difference between two confidence levels is insufficient to form a clear attribution, output a result pending confirmation, temporarily suspend the update of the baseline or the production batch, and add a sliding window of measurement records before making a final judgment.

[0144] For example, after a production batch change for a certain type of fan blade, the average correction value of multiple sliding windows continuously increased, and the correction angles of the first and second correction planes were concentrated in similar circumferential areas. Meanwhile, the vibration amplitude and phase of the reference standard part remained within historical allowable ranges, and the spindle speed, drive current, and positioning position did not change significantly. The correction trends of other fan blade models also remained stable. At the same time, the initial imbalance and correction weight of the current production batch of fan blades increased synchronously.

[0145] In this scenario, the reliability of equipment baseline drift is low, while the reliability of product batch offset is high. The system identifies the current anomaly as a product batch offset, maintains the original equipment baseline, and outputs a production batch warning. This process avoids erroneous replacement of the baseline standard parts due to changes in product batch quality distribution.

[0146] In another scenario, multiple fan blade models exhibit unidirectional corrective offsets within a similar timeframe. The measured vibration amplitude and phase of the reference standard component simultaneously deviate from historical reference records, and multiple parameters, including spindle drive current fluctuations, speed standard deviation, and positioning deviation, increase concurrently. Even if a particular model is in a production batch changeover phase, the system prioritizes identifying the current anomaly as equipment reference drift based on reference standard component deviation, dynamic feature fingerprint deviation, and cross-model common changes, triggering reference verification and equipment recalibration.

[0147] The working principle and technical effect of the above technical solution are as follows: This embodiment uses the benchmark standard verification results, dynamic feature fingerprint, initial imbalance, dual-plane correction data and correction trend statistics formed in Embodiment 1, without changing the measurement, trial weight, correction and data archiving process of Embodiment 1, and only adds the abnormality source differentiation step after the abnormality of the correction trend is found in Embodiment 1.

[0148] By associating the persistence of the same correction trend and the consistency of the dual-plane direction with the equipment-side state parameters and the product-side state parameters, respectively, the reliability of equipment reference drift and product batch offset is established based on the same time period and the same abnormal trend. The equipment-side judgment reflects the common changes of the reference standard parts, the spindle operating status, and multiple models, while the product-side judgment reflects the specific changes of the production batch, the initial imbalance state, the correction position, and the model range.

[0149] Therefore, the system can distinguish between abnormal correction trends, such as equipment reference drift, product batch offset, random fluctuations, and mixed equipment and product anomalies, and respectively perform equipment recalibration, production batch warning, continued observation, or joint verification. This reduces the probability of misjudging product batch changes as equipment reference drift and frequently updating the reference, while also reducing the risk of misjudging systematic changes on the equipment side as random fluctuations in the product and continuously generating measurement deviations. This improves the accuracy of reference updates, the timeliness of production batch anomaly identification, and the stability and traceability of the fan dynamic balance correction process.

[0150] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A fan dynamic balance correction method based on multi-source sensor fusion, characterized in that, The method includes: S1. Use independently stored reference standard parts to quickly verify and measure the equipment to obtain the reference deviation value before calibration; S2. Correct the equipment calibration process by using the benchmark deviation value to obtain a verified reliable benchmark. Use the reliable benchmark to perform initial imbalance measurement and analysis on the fan blade under test and record the main shaft operating parameters simultaneously to obtain the initial imbalance amount and dynamic feature fingerprint. S3. Perform trial weight measurements on each correction plane using the initial imbalance and generate a correction scheme. At the same time, use dynamic feature fingerprints to adaptively match the sensing response coefficients of the verification process to obtain accurate measurement data. S4. The pass / fail determination is completed by comparing the accurate measurement data with the acceptance value, and the complete data is archived. The reference standard parts are periodically verified by the accumulated correction trend of the same model in the archived data, and a systematic deviation warning signal is obtained and the reference is updated.

2. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 1, characterized in that, S1 includes: By taking out the sealed standard parts and installing them in the spindle chuck, and aligning the laser light with the reference point, the spindle positioning and light alignment status are confirmed, and a signal indicating that the standard parts installation is complete is obtained. By starting the equipment and running it at a preset speed, and with the vibration sensor collecting data synchronously, the standard parts are quickly verified and measured to obtain the measured vibration characteristics. By comparing and analyzing the measured vibration characteristics with historical benchmark records, the current degree of deviation of the equipment is quantitatively evaluated, and the quasi-deviation value is obtained.

3. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 2, characterized in that, The method involves comparing and analyzing the measured vibration characteristics with historical benchmark records to quantitatively assess the current degree of equipment deviation and obtain a benchmark deviation value, including: By performing difference calculation between the amplitude data in the measured vibration characteristics and the standard amplitude at the corresponding rotational speed in the historical benchmark record, the vibration energy deviation is quantified to obtain the amplitude deviation amount; By calculating the angle difference between the phase data in the measured vibration characteristics and the standard phase at the corresponding rotational speed in the historical benchmark record, the degree of phase drift is quantified, and the phase deviation is obtained. By combining the amplitude deviation and phase deviation into a two-dimensional deviation vector, the degree and direction of the equipment's current deviation from the standard state are comprehensively characterized, and a reference deviation vector is obtained.

4. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 1, characterized in that, S2 includes: By embedding the benchmark deviation value as a compensation coefficient into the calibration algorithm, the original calibration parameters are corrected to obtain a reliable benchmark. The first measurement is initiated by a reliable benchmark. The vibration signal of the fan blade under test is collected and the spindle speed, drive current and position are recorded synchronously to obtain the first vibration data and operating parameters. By rotating the fan blades 180 degrees and re-aligning them with the light, a second measurement was initiated to collect the reverse vibration signal and obtain the opposing vibration data. By comparing and calculating the initial vibration data with the opposing vibration data, the weight and orientation of the unbalance are analyzed to obtain the initial unbalance. By integrating the spindle speed, drive current, and position recorded synchronously in two measurements into a combination of operating parameters, the dynamic characteristics of the fan blade spindle system are characterized, and a dynamic feature fingerprint is obtained.

5. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 4, characterized in that, The dynamic characteristics of the fan blade spindle system are characterized by integrating the spindle speed, drive current, and position recorded synchronously in two measurements into a combination of operating parameters, thereby obtaining a dynamic feature fingerprint, including: By superimposing and comparing the drive current curves recorded in the initial measurement and the flip measurement, the load fluctuation pattern of the motor when the fan blades rotate is extracted to obtain the current fluctuation characteristics. By calculating the mean and standard deviation of the spindle speed data recorded in the initial measurement and the overturning measurement, the speed stability is quantified, and the speed stability characteristics are obtained. By comparing the repeatability of the position data recorded in the initial measurement and the flip measurement, the deviation range of the spindle's stop position each time is quantified to obtain the positioning repeatability characteristics. By combining and encoding current fluctuation characteristics, rotational speed stability characteristics, and positioning repeatability characteristics into a multi-dimensional feature vector, the dynamic response characteristics of the fan blade under the current spindle state are uniquely identified, thus obtaining a dynamic feature fingerprint.

6. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 1, characterized in that, S3 includes: The test weight position of the first correction plane is determined by the initial unbalance. A predetermined test weight is added at zero degrees and two measurements are completed in both directions. The vibration response change of the first plane is recorded to obtain the data before and after the test weight of the first plane. After removing the test weight of the first plane, the position of the test weight of the second corrected plane is determined based on the initial unbalance. The same test weight is added at zero degrees and two measurements are completed in both directions. The changes in the vibration response of the second plane are recorded to obtain the data before and after the second plane test weight. By comprehensively analyzing the differences in data before and after each trial weight on the two planes, the corrected weight, angle, and direction are analyzed to obtain a correction scheme. By comparing and matching the dynamic feature fingerprint with the current spindle operating parameters, the sensing response coefficient of the verification process is adjusted to obtain the measurement conditions after adaptive matching. After performing the operation according to the correction scheme, the measurement is started under matching conditions to collect the corrected vibration data and obtain the corrected accurate measurement data.

7. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 6, characterized in that, The step of comparing and matching dynamic feature fingerprints with current spindle operating parameters to adjust the sensing response coefficients in the verification process and obtain adaptively matched measurement conditions includes: By comparing the difference between the real-time drive current in the current spindle operating parameters and the current fluctuation characteristics in the dynamic feature fingerprint, the consistency between the current load state and the initial measurement is determined, and the load matching degree is obtained. By comparing the real-time speed fluctuations in the current spindle operating parameters with the speed stability features in the dynamic feature fingerprint, the degree of difference between the current stable speed state and the initial measurement is quantified, and the speed matching degree is obtained; By weighting and combining load matching degree and speed matching degree, the overall similarity between the current measurement condition and the initial measurement condition is quantitatively evaluated to obtain the condition consistency coefficient. The sensor response sensitivity in the verification process is reversed by using the consistency coefficient of the operating conditions. When the consistency coefficient is lower than the threshold, the response gain is increased according to the preset compensation curve. When the consistency coefficient is higher than the threshold, the original gain is maintained, thus obtaining the measurement conditions after adaptive matching.

8. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 1, characterized in that, S4 includes: By comparing the corrected precise measurement data with the system's preset acceptable values ​​item by item, the residual unbalance of each plane is judged to determine whether it is within the allowable range, and the pass / fail judgment result is obtained. By writing all the information, including blade information, initial imbalance, correction scheme, post-correction measurement data and operating parameters, into the database, the complete measurement process data is archived and data records are obtained. By extracting the cumulative correction amount of the same type of fan blade from the data records and statistically analyzing its mean and dispersion, the correction trend is analyzed to obtain the statistical results of the correction trend. By comparing the corrected trend statistics with the preset systematic offset threshold, it is determined whether the benchmark has drifted, an early warning signal is obtained, and the benchmark is updated.

9. The fan dynamic balance correction method based on multi-source sensor fusion according to claim 8, characterized in that, The method involves extracting the cumulative correction amount of the same type of fan blade from data records and statistically analyzing its mean and dispersion to obtain statistical results of the correction trend, including: By filtering all qualified records of the same type of fan blade from the data records, the corrected weight data in each record is arranged in chronological order to obtain the corrected weight time sequence. By calculating the sliding window mean of the corrected weight time series, the central tendency of the correction amount within each window is extracted to obtain the corrected mean series. By calculating the standard deviation of the corrected weight time series using a sliding window, the degree of fluctuation of the correction amount within each window is quantified, and the corrected dispersion series is obtained. By jointly analyzing the corrected mean series and the corrected dispersion series, the directional drift trend and non-directional fluctuation trend of the correction amount over time are separated, and the statistical results of the correction trend including the drift slope and fluctuation amplitude are obtained.

10. A fan dynamic balancing correction system based on multi-source sensor fusion, characterized in that, The system includes: The measurement and analysis module is used to quickly verify and analyze the equipment using independently stored reference standard parts to obtain the reference deviation value before calibration. The imbalance feature analysis module is used to correct the equipment calibration process by using the benchmark deviation value to obtain a verified reliable benchmark. The module then uses the reliable benchmark to perform initial imbalance measurement and analysis on the fan blade under test and simultaneously records the spindle operating parameters to obtain the initial imbalance amount and dynamic feature fingerprint. The adaptive matching module is used to perform trial weight measurements on each correction plane based on the initial imbalance and generate a correction scheme. At the same time, it adaptively matches the sensing response coefficients of the verification stage through dynamic feature fingerprints to obtain accurate measurement data. The periodic verification module is used to complete the qualification judgment by comparing the accurate measurement data with the acceptance value, and archive the complete data. The module performs periodic verification on the reference standard parts by the accumulated correction trend of the same model in the archived data, obtains the systematic deviation warning signal and triggers the reference update.

Images