A calibration system and calibration method for a six-dimensional sensor

Abstract

The application discloses a kind of six-dimensional sensor calibration system and calibration method, belong to sensor calibration technical field, to solve the problem of existing calibration scheme time determination inaccuracy, zero drift monitoring is not accurate, fault identification is not complete, calibration opportunity trigger module is based on entropy weight method screening mechanics, environment, mechanical wear core parameter, calculates trigger index to determine calibration opportunity;Signal drift monitoring module records initial signal, calculates instantaneous drift value and zero drift cumulative amount, and fits slope to judge zero drift anomaly;Frequency domain and energy domain characteristic analysis module extracts harmonic ratio by Fourier transform, calculates energy entropy mutation fluctuation value, and identifies frequency domain / energy domain anomaly;Calibration anomaly early warning module stores multidomain anomaly, and triggers calibration or early warning;The application realizes accurate calibration and anomaly early warning, guarantees sensor measurement accuracy and operational reliability.

Description

Technical Field

[0001] This invention belongs to the field of sensor calibration technology, specifically relating to a calibration system and method for a six-dimensional sensor. Background Technology

[0002] In fields such as industrial automation and robot control, six-dimensional sensors are used to simultaneously measure three-dimensional force and three-dimensional torque. The accuracy of these measurements directly affects the operational stability of the equipment. Existing six-dimensional sensor calibration schemes have the following shortcomings:

[0003] 1. The timing of calibration often relies on a single mechanical parameter or human experience, without taking into account the influence of environmental parameters and mechanical wear parameters, which leads to delayed calibration triggering or misjudgment.

[0004] 2. The monitoring of signal zero drift is limited to the analysis of instantaneous values ​​in a single dimension, lacking dynamic tracking of the cumulative amount of zero drift in each dimension and comprehensive evaluation in multiple dimensions, making it difficult to capture the accuracy degradation caused by subtle drifts;

[0005] 3. Frequency domain anomaly analysis does not incorporate the weight allocation of low, medium and high frequency bands, and lacks quantitative analysis of abrupt changes in energy domain signal complexity, making it unable to effectively identify intermittent faults. To address this, we propose a calibration system and method for a six-dimensional sensor. Summary of the Invention

[0006] The purpose of this invention is to provide a calibration system and method for a six-dimensional sensor to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a calibration system for a six-dimensional sensor, comprising:

[0008] Calibration Timing Trigger Module: Based on the entropy weight method, the weights of the performance parameters affecting the six-dimensional sensor are analyzed, and the core parameters are marked accordingly; the monitoring window is divided, and the characteristic values ​​of the core parameters of mechanical, environmental and mechanical wear types within the monitoring window are analyzed and comprehensively analyzed to obtain the trigger index, and the calibration process is determined accordingly.

[0009] Signal drift monitoring module: Records the initial signals of each dimension within the calibration period and calculates the instantaneous drift value corresponding to each moment. Accumulates the zero drift accumulation of each dimension, constructs the time change curve of the zero drift accumulation of each dimension, fits the curve slope, and performs comprehensive analysis to obtain the three-dimensional force and torque comprehensive drift slope, thereby determining whether there is a zero drift calibration abnormality.

[0010] Frequency domain and energy domain feature analysis module: Perform Fourier transform on the six-dimensional signal, extract the harmonic amplitude ratio of low, medium and high frequency bands and weight it to obtain the weighted harmonic ratio, and determine whether the frequency domain distribution is abnormal; divide the time unit to calculate the energy entropy mutation value, calculate the standard deviation to obtain the energy entropy mutation fluctuation value, and determine whether the energy domain is disordered.

[0011] Calibration anomaly early warning module: Collects anomalies in the force domain, torque domain, frequency domain, and energy domain and stores them in the anomaly pool. If an anomaly is found, recalibration is triggered. If three consecutive anomalies occur, a calibration anomaly early warning is triggered.

[0012] Preferably, the specific process for marking the core parameters is as follows:

[0013] Obtain the mechanical parameters, environmental parameters, and mechanical wear parameters that affect the performance of the six-dimensional sensor during actual operation, and define them as performance-influencing parameters;

[0014] For each performance-affecting parameter j, analyze the probability distribution corresponding to the parameter;

[0015] Calculate the entropy value of performance-affecting parameters using the information entropy formula. ;

[0016] Using the formula: The weights Wj of the performance-affecting parameters are obtained; where G is the total number of performance-affecting parameters. ;

[0017] Filter out performance-impacting parameters with weights greater than the corresponding thresholds and mark them as core parameters.

[0018] Preferably, the specific process for analyzing the characteristic values ​​of core parameters related to mechanics, environment, and mechanical wear within the monitoring window is as follows:

[0019] Core parameters selected from the sets of mechanical parameters, environmental parameters, and mechanical wear parameters;

[0020] During the operation of the six-dimensional sensor, a monitoring period is set;

[0021] For any core parameter, the monitoring period is divided into K equal-length windows, where K is the total number of windows;

[0022] For each window, obtain the window's quantile value and fluctuation coefficient;

[0023] Weights are assigned to the quantile values ​​and fluctuation coefficients of the window, and then the single feature value Vj of the core parameter is obtained by weighted summation.

[0024] For all core mechanical parameters, a weighted sum is performed according to their corresponding weights to obtain mechanical characteristic values.

[0025] By using the same method to analyze the eigenvalues ​​of mechanical parameters, we can similarly analyze the eigenvalues ​​of environmental parameters and mechanical wear parameters to obtain the eigenvalues ​​of environmental parameters and mechanical wear parameters.

[0026] Preferably, the specific process of analyzing the trigger index and determining whether to trigger the calibration procedure is as follows:

[0027] After normalizing and dimensionlessly processing the mechanical, environmental, and mechanical wear characteristic values, a weighted sum is performed to obtain the trigger index.

[0028] If the trigger index corresponding to the six-dimensional sensor during the monitoring period is greater than or equal to the corresponding threshold, the automatic calibration process will be triggered.

[0029] Preferably, the specific process for analyzing the cumulative zero drift in each dimension is as follows:

[0030] Record the initial output signals corresponding to the signals of each dimension of the six-dimensional sensor at the initial calibration moment;

[0031] During the continuous acquisition and calibration period, the signals of each dimension corresponding to each acquisition time of the six-dimensional sensor are collected, and the difference between the signals and the initial signals of the corresponding dimensions is calculated to obtain the instantaneous drift values ​​of each dimension corresponding to each acquisition time.

[0032] The acquisition timestamp corresponding to each instantaneous drift value in each dimension is recorded synchronously to form discrete data pairs of time-instantaneous drift values ​​corresponding to signals in each dimension;

[0033] Based on the instantaneous drift value at each moment, the cumulative zero drift corresponding to each dimension of the signal within the calibration period is calculated using an accumulation algorithm.

[0034] Preferably, the specific process for analyzing the combined drift slope of three-dimensional force and torque to determine whether there is a zero-drift calibration anomaly is as follows:

[0035] With time as the horizontal axis and the cumulative zero drift of each dimension as the vertical axis, construct the corresponding discrete data sequence;

[0036] The discrete data points are smoothed by using a linear interpolation algorithm to fill the numerical gaps between adjacent acquisition times and generate curves showing the continuous change of the cumulative zero drift in each dimension over time.

[0037] The slope of the curve of the cumulative zero drift in each dimension changing continuously with time is obtained by linear fitting using the least squares method.

[0038] After assigning different preset weight coefficients to the slopes of the three-dimensional curves corresponding to the force signal, the weighted sum is performed to obtain the comprehensive drift slope of the three-dimensional force signal.

[0039] If the overall drift slope of the three-dimensional force signal is greater than or equal to the corresponding threshold, it is determined that the six-dimensional sensor has an abnormal zero-drift calibration of the force signal.

[0040] Similarly, calculate the comprehensive drift slope corresponding to the three-dimensional torque signal. If it is greater than or equal to the corresponding preset threshold, it is determined that the six-dimensional sensor has a torque signal zero drift calibration abnormality and is marked as a torque signal zero drift abnormality.

[0041] Preferably, the specific process for analyzing the weighted harmonic ratio to determine whether there is an abnormal frequency domain distribution is as follows:

[0042] A fast Fourier transform is performed on the six-dimensional signal during the calibration cycle to extract the ratio of harmonic amplitude to fundamental amplitude in preset low-frequency, mid-frequency, and high-frequency bands; thus obtaining the low-frequency harmonic amplitude ratio, mid-frequency harmonic amplitude ratio, and high-frequency harmonic amplitude ratio.

[0043] By assigning different preset weighting coefficients to the low-frequency harmonic amplitude ratio, the mid-frequency harmonic amplitude ratio, and the high-frequency harmonic amplitude ratio, the weighted sum is obtained.

[0044] If the weighted harmonic ratio corresponding to the calibration period is greater than or equal to the corresponding threshold, then the frequency domain signal distortion is marked as abnormal.

[0045] Preferably, the specific process for analyzing the sudden fluctuations in energy entropy to determine whether there is energy domain disorder is as follows:

[0046] The calibration period is divided into several time units, and the energy entropy value of the signal in each time unit is calculated.

[0047] The difference between the energy entropy values ​​corresponding to two adjacent time units will be calculated to obtain the energy entropy mutation value corresponding to each adjacent time unit.

[0048] The absolute values ​​of energy entropy mutations corresponding to all adjacent time units within the calibration period are used as sample data, and the standard deviation formula is used to calculate the energy entropy mutation fluctuation value.

[0049] If the energy entropy fluctuation value corresponding to the calibration period is greater than or equal to the corresponding threshold, then the energy domain disorder is marked as abnormal.

[0050] Preferably, the working process of the calibration anomaly early warning module is as follows:

[0051] During the calibration process, if any abnormalities occur, such as zero drift in the force signal, zero drift in the torque signal, distortion in the frequency domain signal, or disorder in the energy domain signal, they will be marked as force domain zero drift abnormality, torque domain zero drift abnormality, frequency domain distortion abnormality, and energy domain disorder abnormality, respectively.

[0052] Construct a calibration anomaly pool to store all the above-mentioned anomalies during the calibration process.

[0053] If there are abnormal items in the calibration anomaly pool, the calibration anomaly is marked and the recalibration process is triggered;

[0054] If three consecutive calibrations fail, a calibration failure warning will be triggered, and each corresponding failure item in the three calibration failure pool will be sent to the operations and maintenance department.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] (1) The calibration system and calibration method of the six-dimensional sensor, through the calibration timing trigger module, integrates three types of parameters, namely mechanics, environment and mechanical wear, based on the entropy weight method, objectively calculates the parameter weights and screens the core parameters, and then realizes the quantitative judgment of the calibration timing by dynamically dividing the monitoring window and weighting the trigger index. This method abandons the reliance on a single parameter or human experience, comprehensively reflects the risk of sensor performance degradation, reduces the calibration mis-triggering and missed triggering caused by parameter bias or subjective error, and improves the accuracy and scientific nature of calibration timing judgment.

[0057] (2) The calibration system and calibration method of the six-dimensional sensor record the initial signal, calculate the instantaneous drift value and accumulate it to obtain the zero drift accumulation. Combined with linear interpolation smoothing and least squares fitting slope, the system achieves high-precision quantification and dynamic tracking of zero drift in each dimension. At the same time, by weighted calculation of the three-dimensional force and torque combined drift slope, the system avoids the limitation of single-dimensional evaluation, can capture the zero drift acceleration trend in advance, and form a closed loop with the calibration timing trigger module to prevent the measurement error caused by zero drift accumulation and ensure the long-term operating accuracy of the sensor.

[0058] (3) The calibration system and calibration method of this six-dimensional sensor combine the frequency domain and energy domain feature analysis module with fast Fourier transform and energy entropy analysis. It accurately identifies frequency domain anomalies such as cable interference and power supply ripple through weighted harmonic ratio and captures intermittent faults through energy entropy mutation fluctuation value. The calibration anomaly early warning module centrally manages multi-domain anomalies and handles them in a hierarchical manner. The two work together to build a multi-dimensional fault monitoring system that covers steady-state and intermittent faults, avoids the problem of one-sided fault identification in traditional solutions, provides accurate basis for operation and maintenance, and improves the reliability of sensor operation. Attached Figure Description

[0059] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example: Please refer to Figure 1 The present invention provides a calibration system for a six-dimensional sensor, comprising: a calibration timing triggering module, a signal drift monitoring module, a frequency domain and energy domain feature analysis module, and a calibration anomaly early warning module;

[0062] Calibration Timing Trigger Module: Based on the entropy weight method, the module analyzes the weights of performance parameters affecting the six-dimensional sensor and identifies core parameters accordingly. It then divides the monitoring window, analyzes the characteristic values ​​of core parameters related to mechanics, environment, and mechanical wear within the window, and performs a comprehensive analysis to obtain a trigger index. Based on this index, it determines whether to trigger the calibration process. The specific process is as follows:

[0063] Obtain the mechanical parameters, environmental parameters, and mechanical wear parameters that affect the performance of the six-dimensional sensor during actual operation, and define them as performance-influencing parameters;

[0064] We collected the mechanical parameters, environmental parameters, and mechanical wear parameters that affect the performance of the six-dimensional sensor during actual operation and constructed the corresponding parameter set.

[0065] Among them, the mechanical parameter set includes three-dimensional peak force. Parameters such as torque alternation frequency, force signal standard deviation, and torque signal skewness;

[0066] Environmental parameter set: covering parameters such as temperature change rate, vibration acceleration, humidity, and air pressure;

[0067] Mechanical wear parameter set: including parameters such as loading mechanism clearance, connecting component looseness, and elastomer deformation accumulation;

[0068] Collect historical sample data for each performance-influencing parameter;

[0069] For each performance-influencing parameter j, multiple sets of historical operating data were collected under different operating conditions and time periods to form a sample set. ;

[0070] Normalize the various performance-influencing parameters to eliminate the influence of dimensions;

[0071] The standardized data with parameter j is divided into m equal-width intervals, and the frequency of samples in each interval is counted. ;

[0072] Using the formula: , thus obtaining the probability distribution pij of the i-th sample at the j-th parameter;

[0073] When the frequency Cij in a certain interval is 0, adding 1 to the numerator and denominator avoids the calculation error of ln(0) and ensures the probability normalization condition. ;

[0074] Using the formula: The entropy value of the performance-affecting parameters is obtained. , where i is the sample label, j is the parameter label; pij is the probability distribution of the j-th parameter in the i-th sample, n is the total number of samples, k is the adjustment coefficient, k=1 / ln(n);

[0075] Using the formula: We obtain the weights Wj of the performance-affecting parameters; where G is the total number of performance-affecting parameters. G;

[0076] Preset weight thresholds, filter out performance-impacting parameters whose weights are greater than the corresponding thresholds, and mark them as core parameters;

[0077] The core parameters selected from the set of mechanical parameters are denoted as the set. There are a parameters in total; where m a The core mechanical parameter labeled a;

[0078] The core parameters selected from the set of environmental parameters are denoted as the set. There are b parameters in total; The core environmental parameter for label b;

[0079] The core parameters selected from the set of mechanical wear parameters are denoted as the set. There are c parameters in total; The core parameter for mechanical wear is labeled c;

[0080] Where a+b+c=G;

[0081] During the operation of the six-dimensional sensor, a monitoring period is set, and for any core parameter, the monitoring period is divided into K equal-length windows. K can be adaptively divided according to the historical data fluctuation frequency of the parameter. That is, the more intense the fluctuation, the more windows K are required to ensure that each window can reflect the local change characteristics of the parameter.

[0082] For each window, obtain the window's quantile value and fluctuation coefficient;

[0083] Among them, quantile value: Based on the preset quantile threshold P, the P quantile value of the data in the window is calculated, which reflects the typical level of the data in the window;

[0084] Volatility coefficient: The ratio of the standard deviation to the mean of the data within a calculation window, reflecting the degree of data dispersion;

[0085] The entropy weight analysis logic is used to assign objective weights to the window quantile and fluctuation coefficient, and then the single feature value Vj of the core parameter is obtained by weighted summation; (refer to the calculation logic of the performance impact parameter weights).

[0086] For core mechanical parameters to The weighted calculation is performed according to its corresponding weight (based on the weight Wj of the performance impact parameter), that is: We obtain the mechanical eigenvalues ​​Vm; where represents the weights corresponding to the core mechanical parameters; g represents the labels of the core mechanical parameters.

[0087] By using the same method to analyze the eigenvalues ​​of mechanical parameters, we can similarly analyze the eigenvalues ​​of environmental parameters and mechanical wear parameters to obtain the environmental eigenvalue Ve and the mechanical wear eigenvalue Va.

[0088] After normalizing and dimensionlessly removing the mechanical characteristic value Vm, the environmental characteristic value Ve, and the mechanical wear characteristic value Va, the formula is used: The trigger index S is obtained; where q1, q2, and q3 are preset weighting coefficients; this index directly reflects the comprehensive risk of sensor performance degradation;

[0089] A preset trigger index threshold is set. If the trigger index corresponding to the six-dimensional sensor during the monitoring period is greater than or equal to the corresponding threshold, the automatic calibration process will be triggered.

[0090] It should be noted that by comprehensively considering three types of parameters—mechanical, environmental, and mechanical wear—and objectively calculating their weights using the entropy weight method, the one-sidedness of evaluating a single parameter is avoided. For example, by incorporating parameters such as three-dimensional force peak value, temperature change rate, and cumulative elastic body deformation into the analysis, the performance degradation risk of the sensor under different operating conditions can be comprehensively reflected, making the trigger index closer to the actual operating state.

[0091] When calculating the weighted trigger index, the monitoring window is adaptively divided based on the fluctuation frequency of historical data. The number of windows is increased for high-frequency fluctuating parameters to ensure that the local change characteristics of the parameters are captured. This dynamic analysis method can reduce false triggers caused by instantaneous fluctuations in parameters and improve the accuracy of calibration timing judgment.

[0092] The entropy weight method is used to calculate the parameter weights, which are based on the probability distribution and entropy value of the data itself, rather than subjective experience, thus avoiding interference from human factors in the selection of core parameters.

[0093] An objective weighting calculation logic is used for window quantiles and fluctuation coefficients, and characteristic values ​​are derived through mathematical formulas, reducing the bias caused by subjective settings; this data-driven evaluation mechanism makes the calibration trigger conditions more repeatable and scientific.

[0094] The trigger index transforms multi-dimensional characteristics into quantitative indicators, directly reflecting the comprehensive risk of sensor performance degradation. When the trigger index exceeds the threshold, calibration is actively triggered, changing post-fault calibration to preventive calibration, which can eliminate measurement errors caused by potential problems such as signal drift and frequency domain distortion in advance.

[0095] Signal drift monitoring module: Records the initial signals of each dimension within the calibration period and calculates the instantaneous drift value at each moment. The accumulated zero drift values ​​for each dimension are obtained, and a time-varying curve of the accumulated zero drift values ​​for each dimension is constructed. The slope of the curve is fitted, and comprehensive analysis is performed to obtain the combined drift slope of three-dimensional force and torque. Based on this, it is determined whether there is a zero drift calibration anomaly. The specific process is as follows:

[0096] Record the initial output signals corresponding to the force / torque signals in each dimension of the six-dimensional sensor at the initial calibration moment; including the initial three-dimensional force signal and the initial three-dimensional torque signal;

[0097] At preset time intervals, signals of each dimension corresponding to each acquisition time of the six-dimensional sensor are continuously acquired within the calibration period, and the difference between the acquisition time and the initial signal of the corresponding dimension is calculated to obtain the instantaneous drift value of each dimension corresponding to each acquisition time.

[0098] The acquisition timestamp corresponding to each instantaneous drift value in each dimension is recorded synchronously to form discrete data pairs of time-instantaneous drift values ​​corresponding to signals in each dimension;

[0099] Based on the instantaneous drift value at each moment, the cumulative zero drift corresponding to each dimension of the signal within the calibration period is calculated using an accumulation algorithm;

[0100] Furthermore, for each dimension of the force signal, the product of the instantaneous drift value and the sampling interval is accumulated in time order to obtain the cumulative zero drift of the force signal in each dimension.

[0101] Right now: The cumulative zero-drift force signal along the x-axis is obtained. , where h is each acquisition time and H is the total number of acquisition times within the calibration period; For the collection interval, The instantaneous drift value of the force signal corresponding to the x-axis dimension of the six-dimensional sensor at the acquisition time h;

[0102] Similarly, calculate and ;

[0103] Based on the method of calculating the cumulative zero drift of force signals in each dimension, the cumulative zero drift of torque signals in each dimension can be calculated similarly.

[0104] With time as the horizontal axis and the cumulative zero drift of each dimension as the vertical axis, construct the corresponding discrete data sequences (a total of six discrete data sequences).

[0105] The discrete data points are smoothed by using a linear interpolation algorithm to fill the numerical gaps between adjacent acquisition times and generate curves showing the continuous change of the cumulative zero drift of each dimension (force / torque) over time.

[0106] The slope of the curve of the cumulative zero drift of each dimension (force / torque) changing continuously with time is obtained by linear fitting using the least squares method;

[0107] If the slope of the cumulative zero drift curve in each dimension is positive, the corresponding slope can be amplified and corrected based on the preset amplification rules. The purpose is to amplify and accelerate the effect of drift by correction, so as to avoid underreporting due to the gradual change in slope.

[0108] After assigning different preset weight coefficients to the slopes of the three-dimensional curves corresponding to the force signal, the weighted sum is performed to obtain the comprehensive drift slope of the three-dimensional force signal.

[0109] At the same time, a comprehensive drift threshold for the three-dimensional force signal is preset. If the comprehensive drift slope of the three-dimensional force signal is greater than or equal to the corresponding threshold, it is determined that the six-dimensional sensor has an abnormal zero drift calibration of the force signal.

[0110] The slope of the three-dimensional curves corresponding to the force signal can be set with reference to the weight of the performance influence parameter corresponding to the force signal in the corresponding dimension.

[0111] Similarly, calculate the comprehensive drift slope corresponding to the three-dimensional torque signal. If it is greater than or equal to the corresponding preset threshold, it is determined that the six-dimensional sensor has a torque signal zero drift calibration abnormality and is marked as a torque signal zero drift abnormality.

[0112] It should be noted that by recording the initial signal and calculating the instantaneous drift value in real time, and combining the accumulation algorithm to obtain the cumulative zero drift, high-precision quantification of the zero drift of the six-dimensional sensor is achieved.

[0113] Linear interpolation smoothing and least squares fitting of slope transform discrete data into continuous curves, further improving the accuracy of drift trend analysis.

[0114] The zero-drift accumulation of three-dimensional force and three-dimensional torque signals were analyzed separately, six discrete data sequences were constructed and the comprehensive drift slope was calculated, thus avoiding the limitations of single-dimensional evaluation.

[0115] The real-time acquisition and timestamp recording mechanism enables dynamic tracking of zero drift changes within the calibration cycle, providing time-series data support for sensor performance degradation analysis; by continuously monitoring the slope of the cumulative zero drift, the trend of accelerated zero drift growth can be detected in advance.

[0116] The closed-loop monitoring logic (data acquisition - processing - analysis - judgment) of this module is linked with the calibration timing trigger module. When a zero drift anomaly is detected, recalibration can be triggered, forming a virtuous cycle of monitoring - calibration - re-monitoring, ensuring that the sensor maintains the stability of measurement accuracy during long-term operation.

[0117] Frequency and Energy Domain Feature Analysis Module: Performs Fourier transform on the six-dimensional signal, extracts the harmonic amplitude ratios of low, mid, and high frequency bands, and weights them to obtain the weighted harmonic ratio, thereby determining whether there is anomaly in the frequency domain distribution; divides the signal into time units to calculate energy entropy mutation values, and calculates the standard deviation to obtain energy entropy mutation fluctuation values, thereby determining whether there is energy domain disorder. The specific process is as follows:

[0118] A fast Fourier transform is performed on the six-dimensional signal during the calibration cycle to extract the ratio of harmonic amplitude to fundamental amplitude in preset low-frequency, mid-frequency, and high-frequency bands; the low-frequency harmonic amplitude ratio, mid-frequency harmonic amplitude ratio, and high-frequency harmonic amplitude ratio are obtained; where the fundamental amplitude is the amplitude corresponding to the signal's main frequency;

[0119] By assigning different preset weighting coefficients to the low-frequency harmonic amplitude ratio, the mid-frequency harmonic amplitude ratio, and the high-frequency harmonic amplitude ratio, the weighted sum is obtained.

[0120] The calculation of weighted harmonic ratio can highlight the characterization of common faults such as cable interference and power ripple in the low and mid frequency bands, and improve the fault identification of frequency domain characteristics.

[0121] A preset weighted harmonic ratio threshold is set. If the weighted harmonic ratio corresponding to the calibration period is greater than or equal to the corresponding threshold, the frequency domain signal distortion is marked as abnormal, indicating that there is an abnormal frequency domain distribution in the signal.

[0122] The calibration period is divided into several time units, and the energy entropy value of the signal in each time unit is calculated. The higher the energy entropy value, the more dispersed the signal energy is in each frequency / amplitude range, and the higher the signal complexity; conversely, the lower the energy entropy value, the purer the signal.

[0123] The difference between the energy entropy values ​​corresponding to two adjacent time units will be calculated to obtain the energy entropy mutation value corresponding to each adjacent time unit.

[0124] Among them, the energy entropy value is calculated by the probability entropy of the signal energy distribution, which reflects the signal complexity and is consistent with the probability distribution calculation logic of the "performance impact parameters" mentioned above.

[0125] The absolute values ​​of energy entropy mutations corresponding to all adjacent time units within the calibration period are used as sample data, and the standard deviation formula is used to calculate the energy entropy mutation fluctuation value.

[0126] The smaller the fluctuation of energy entropy, the closer the amplitude of signal complexity changes in all adjacent time units is to the average value, with no significant drastic changes and strong signal stability; conversely, the larger the fluctuation, the more likely it is that the amplitude of changes in some adjacent time units is much greater than the average value, and the signal complexity fluctuates frequently and significantly, possibly indicating intermittent faults.

[0127] A preset threshold for energy entropy fluctuation is set. If the energy entropy fluctuation value corresponding to the calibration cycle is greater than or equal to the corresponding threshold, the energy domain disorder is marked as abnormal.

[0128] It should be noted that by extracting the harmonic amplitude ratios of low, medium and high frequency bands through fast Fourier transform and combining them with weighted summation to obtain the weighted harmonic ratio, typical faults such as cable interference and power ripple can be specifically identified; thus, quantitative judgment of abnormal frequency domain distribution is achieved.

[0129] When the weighted harmonic ratio exceeds the threshold, the frequency domain signal distortion is directly marked as abnormal, providing a clear frequency domain characteristic basis for calibration triggering.

[0130] Based on the analysis of energy entropy and sudden fluctuation values, the dynamic changes in signal energy distribution can be effectively monitored. Energy entropy reflects signal complexity, while sudden fluctuation values ​​quantify the magnitude of energy entropy changes in adjacent time units through standard deviation, which can sensitively capture intermittent faults.

[0131] The time unit division and probability entropy calculation logic enable energy domain analysis to have both time domain and frequency domain characteristics; dividing the calibration period into multiple time units can track the temporal changes in energy distribution, and combined with the probability entropy calculation method, ensures the objectivity of energy entropy value calculation and avoids misjudgment caused by short-term signal fluctuations.

[0132] The combined analysis of frequency and energy domains overcomes the limitations of single signal domain analysis; frequency domain features focus on identifying the frequency components of steady-state faults, while energy domain features focus on monitoring the complexity changes of dynamic processes. The combination of the two can comprehensively cover various anomalies in sensor operation.

[0133] This module works in conjunction with signal drift monitoring and calibration anomaly early warning to build a multi-dimensional fault monitoring system that integrates time-domain drift, frequency-domain distortion, and energy disturbance. When anomalies occur in the frequency or energy domain, the calibration process or anomaly early warning can be triggered, achieving closed-loop management from fault feature extraction to calibration execution and ensuring the long-term stable operation of the six-dimensional sensor under complex working conditions.

[0134] The calibration anomaly early warning module collects anomalies in the force domain, torque domain, frequency domain, and energy domain and stores them in an anomaly pool. If an anomaly is detected, recalibration is triggered. Three consecutive anomalies trigger a calibration anomaly early warning. The specific process is as follows:

[0135] During the calibration process, if any abnormalities occur, such as zero drift in the force signal, zero drift in the torque signal, distortion in the frequency domain signal, or disorder in the energy domain signal, they will be marked as force domain zero drift abnormality, torque domain zero drift abnormality, frequency domain distortion abnormality, and energy domain disorder abnormality, respectively.

[0136] Construct a calibration anomaly pool to store all the above-mentioned anomalies during the calibration process.

[0137] If there are abnormal items in the calibration anomaly pool, the calibration anomaly is marked and the recalibration process is triggered;

[0138] If three consecutive calibrations fail, a calibration failure warning will be triggered, and each corresponding failure item in the three calibration failure pool will be sent to the operations and maintenance department.

[0139] It should be noted that outliers in the force domain, torque domain, frequency domain, and energy domain are stored in an outlier pool to form a centralized management mechanism. When an outlier occurs in any domain, recalibration can be triggered to avoid incomplete calibration caused by missing a single outlier.

[0140] A graded strategy of single-abnormal recalibration plus three consecutive abnormal warnings can be adopted to distinguish between occasional faults and systemic problems.

[0141] The specific anomalies recorded in the anomaly pool provide precise fault location information for operations and maintenance.

[0142] A calibration method for a six-dimensional sensor, comprising:

[0143] Step 1: Analyze the weights of the performance parameters affecting the six-dimensional sensor based on the entropy weight method, and mark the core parameters accordingly; divide the monitoring window, analyze the characteristic values ​​of the core parameters related to mechanics, environment, and mechanical wear within the monitoring window, and conduct a comprehensive analysis to obtain the trigger index, and determine whether to trigger the calibration process based on this index;

[0144] Step 2: Record the initial signals of each dimension within the calibration period and calculate the instantaneous drift value corresponding to each moment. Accumulate the zero drift accumulation of each dimension, construct the time change curve of the zero drift accumulation of each dimension, fit the curve slope, and conduct comprehensive analysis to obtain the three-dimensional force and torque comprehensive drift slope. Based on this, determine whether there is a zero drift calibration anomaly.

[0145] Step 3: Perform a Fourier transform on the six-dimensional signal, extract the harmonic amplitude ratios of the low, mid and high frequency bands, and weight them to obtain the weighted harmonic ratio. Based on this, determine whether there is an abnormal frequency domain distribution; divide the time unit to calculate the energy entropy mutation value, and calculate the standard deviation to obtain the energy entropy mutation fluctuation value. Based on this, determine whether there is energy domain disorder.

[0146] Step 4: Collect abnormal items in the force domain, torque domain, frequency domain, and energy domain and store them in the abnormality pool. If there is an abnormality, recalibration will be triggered. If there are three consecutive abnormalities, a calibration abnormality warning will be triggered.

[0147] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A calibration system for a six-dimensional sensor, characterized in that: include: Calibration Timing Trigger Module: Based on the entropy weight method, analyze the weights of the parameters affecting the performance of the six-dimensional sensor, and mark the core parameters accordingly; The monitoring window is divided, and the characteristic values ​​of core parameters related to mechanics, environment, and mechanical wear within the monitoring window are analyzed and comprehensively analyzed to obtain the trigger index, and the calibration process is determined accordingly. Signal drift monitoring module: Records the initial signals of each dimension within the calibration period and calculates the instantaneous drift value corresponding to each moment. Accumulates the zero drift accumulation of each dimension, constructs the time change curve of the zero drift accumulation of each dimension, fits the curve slope, and performs comprehensive analysis to obtain the three-dimensional force and torque comprehensive drift slope, thereby determining whether there is a zero drift calibration abnormality. Frequency domain and energy domain feature analysis module: Perform Fourier transform on the six-dimensional signal, extract the harmonic amplitude ratio of low, medium and high frequency bands and weight it to obtain the weighted harmonic ratio, and determine whether there is an abnormal frequency domain distribution based on this. Divide the time unit to calculate the energy entropy mutation value, and calculate the standard deviation to obtain the energy entropy mutation fluctuation value, thereby determining whether the energy domain is disordered. Calibration anomaly early warning module: Collects anomalies in the force domain, torque domain, frequency domain, and energy domain and stores them in the anomaly pool. If an anomaly is found, recalibration is triggered. If three consecutive anomalies occur, a calibration anomaly early warning is triggered.

2. The calibration system for a six-dimensional sensor according to claim 1, characterized in that: The specific process of marking the core parameters is as follows: Obtain the mechanical parameters, environmental parameters, and mechanical wear parameters that affect the performance of the six-dimensional sensor during actual operation, and define them as performance-influencing parameters; Construct parameter sets corresponding to mechanical parameters, environmental parameters, and mechanical wear parameters; For each performance-affecting parameter j, analyze the probability distribution corresponding to the parameter; Calculate the entropy value of performance-affecting parameters using the information entropy formula. ; Using the formula: We obtain the weights Wj of the performance-affecting parameters; where G is the total number of performance-affecting parameters; j=1,2,...,G; Filter out performance-impacting parameters with weights greater than the corresponding preset thresholds and mark them as core parameters.

3. The calibration system for a six-dimensional sensor according to claim 2, characterized in that: The specific process for analyzing the characteristic values ​​of core parameters related to mechanics, environment, and mechanical wear within the monitoring window is as follows: Core parameters selected from the sets of mechanical parameters, environmental parameters, and mechanical wear parameters; During the operation of the six-dimensional sensor, a monitoring period is set; For any core parameter, the monitoring period is divided into K equal-length windows, where K is the total number of windows; For each window, obtain the window's quantile value and fluctuation coefficient; Weights are assigned to the quantile values ​​and fluctuation coefficients of the window, and then the single feature value Vj of the core parameter is obtained by weighted summation. For all core mechanical parameters, a weighted sum is performed according to their corresponding weights to obtain mechanical characteristic values. By using the same method to analyze the eigenvalues ​​of mechanical parameters, we can similarly analyze the eigenvalues ​​of environmental parameters and mechanical wear parameters to obtain the eigenvalues ​​of environmental parameters and mechanical wear parameters.

4. The calibration system for a six-dimensional sensor according to claim 3, characterized in that: The specific process of analyzing the trigger index and determining whether to trigger the calibration procedure is as follows: After normalizing and dimensionlessly processing the mechanical, environmental, and mechanical wear characteristic values, a weighted sum is performed to obtain the trigger index. If the trigger index corresponding to the six-dimensional sensor during the monitoring period is greater than or equal to the corresponding threshold, the calibration process will be automatically triggered.

5. The calibration system for a six-dimensional sensor according to claim 4, characterized in that: The specific process for analyzing the cumulative zero drift in each dimension is as follows: Record the initial output signals corresponding to the signals of each dimension of the six-dimensional sensor at the initial calibration moment; During the continuous acquisition and calibration period, the signals of each dimension corresponding to each acquisition time of the six-dimensional sensor are collected, and the difference between the signals and the initial signals of the corresponding dimensions is calculated to obtain the instantaneous drift values ​​of each dimension corresponding to each acquisition time. The acquisition timestamp corresponding to each instantaneous drift value in each dimension is recorded synchronously to form discrete data pairs of time-instantaneous drift values ​​corresponding to each dimension signal; Based on the instantaneous drift value at each moment, the cumulative zero drift corresponding to each dimension of the signal within the calibration period is calculated using an accumulation algorithm.

6. The calibration system for a six-dimensional sensor according to claim 5, characterized in that: The specific process for analyzing the combined drift slope of three-dimensional force and torque to determine whether there is a zero-drift calibration anomaly is as follows: With time as the horizontal axis and the cumulative zero drift of each dimension as the vertical axis, construct the corresponding discrete data sequence; The discrete data points are smoothed by using a linear interpolation algorithm to fill the numerical gaps between adjacent acquisition times and generate curves showing the continuous change of the cumulative zero drift in each dimension over time. The slope of the curve of the cumulative zero drift in each dimension changing continuously with time is obtained by linear fitting using the least squares method. After assigning different preset weight coefficients to the slopes of the three-dimensional curves corresponding to the force signal, the weighted sum is performed to obtain the comprehensive drift slope of the three-dimensional force signal. If the overall drift slope of the three-dimensional force signal is greater than or equal to the corresponding threshold, it is determined that the six-dimensional sensor has an abnormal zero-drift calibration of the force signal. Similarly, calculate the comprehensive drift slope corresponding to the three-dimensional torque signal. If it is greater than or equal to the corresponding preset threshold, it is determined that the six-dimensional sensor has a torque signal zero drift calibration abnormality and is marked as a torque signal zero drift abnormality.

7. The calibration system for a six-dimensional sensor according to claim 6, characterized in that: The specific process for analyzing the weighted harmonic ratio to determine whether there is an abnormal frequency domain distribution is as follows: A fast Fourier transform is performed on the six-dimensional signal during the calibration cycle to extract the ratio of harmonic amplitude to fundamental amplitude in the preset low-frequency, mid-frequency, and high-frequency bands. The harmonic amplitude ratios for the low-frequency band, mid-frequency band, and high-frequency band are obtained. By assigning different preset weighting coefficients to the low-frequency harmonic amplitude ratio, the mid-frequency harmonic amplitude ratio, and the high-frequency harmonic amplitude ratio, the weighted sum is obtained. If the weighted harmonic ratio corresponding to the calibration period is greater than or equal to the corresponding threshold, then the frequency domain signal distortion is marked as abnormal.

8. The calibration system for a six-dimensional sensor according to claim 7, characterized in that: The specific process for analyzing abrupt fluctuations in energy entropy to determine whether there is energy domain disorder is as follows: The calibration period is divided into several time units, and the energy entropy value of the signal in each time unit is calculated. The difference between the energy entropy values ​​corresponding to two adjacent time units is calculated to obtain the energy entropy mutation value corresponding to each adjacent time unit. The absolute values ​​of energy entropy mutations corresponding to all adjacent time units within the calibration period are used as sample data, and the standard deviation formula is used to calculate the energy entropy mutation fluctuation value. If the energy entropy fluctuation value corresponding to the calibration period is greater than or equal to the corresponding threshold, then the energy domain disorder is marked as abnormal.

9. The calibration system and calibration method for a six-dimensional sensor according to claim 8, characterized in that: The working process of the calibration anomaly early warning module is as follows: During the calibration process, if any abnormalities occur, such as zero drift in the force signal, zero drift in the torque signal, distortion in the frequency domain signal, or disorder in the energy domain signal, they will be marked as force domain zero drift abnormality, torque domain zero drift abnormality, frequency domain distortion abnormality, and energy domain disorder abnormality, respectively. Construct a calibration anomaly pool to store all the above-mentioned anomalies during the calibration process. If there are abnormal items in the calibration anomaly pool, the calibration anomaly is marked and the recalibration process is triggered; If three consecutive calibrations fail, a calibration failure warning will be triggered, and each corresponding failure item in the three calibration failure pool will be sent to the operations and maintenance department.

10. A calibration method for a six-dimensional sensor, applied to the calibration system for a six-dimensional sensor as proposed in any one of claims 1-9, characterized in that, include: Step 1: Analyze the weights of the performance parameters affecting the six-dimensional sensor based on the entropy weight method, and mark the core parameters accordingly; The monitoring window is divided, and the characteristic values ​​of core parameters related to mechanics, environment, and mechanical wear within the monitoring window are analyzed and comprehensively analyzed to obtain the trigger index, and the calibration process is determined accordingly. Step 2: Record the initial signals of each dimension within the calibration period and calculate the instantaneous drift value corresponding to each moment. Accumulate the zero drift accumulation of each dimension, construct the time change curve of the zero drift accumulation of each dimension, fit the curve slope, and conduct comprehensive analysis to obtain the three-dimensional force and torque comprehensive drift slope. Based on this, determine whether there is a zero drift calibration anomaly. Step 3: Perform Fourier transform on the six-dimensional signal, extract the harmonic amplitude ratio of low, mid and high frequency bands and weight them to obtain the weighted harmonic ratio, and determine whether there is an abnormal frequency domain distribution based on this. Divide the time unit to calculate the energy entropy mutation value, and calculate the standard deviation to obtain the energy entropy mutation fluctuation value, thereby determining whether the energy domain is disordered. Step 4: Collect abnormal items in the force domain, torque domain, frequency domain, and energy domain and store them in the abnormality pool. If there is an abnormality, recalibration will be triggered. If there are three consecutive abnormalities, a calibration abnormality warning will be triggered.

Images