An ultrasonic robot abnormal situation intelligent detection and handling method

Abstract

The application discloses an ultrasonic robot abnormal situation intelligent detection and processing method, and relates to the field of ultrasonic robots, and comprises the following steps: under the operation state of the ultrasonic robot, real-time collection of ultrasonic image data of a target detection area and robot self operation state parameters is carried out based on timestamp alignment; the ultrasonic image data is subjected to environmental and equipment noise interference elimination based on pixel neighborhood correlation, a dynamic screening interval is constructed based on the preset normal working condition fluctuation range of the operation state parameters, values out of range are preliminarily removed, and the correlation alignment of the pre-processing result is completed based on the collection time sequence of the two types of data; the application can accurately collect and process data to reduce interference and remove abnormal values, and can also dynamically update the judgment standard, so that image distortion and parameter out-of-limit problems can be accurately identified.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic robot technology, specifically to an intelligent detection and handling method for abnormal situations in ultrasonic robots. Background Technology

[0002] Ultrasonic robots integrate high-precision ultrasound probes, multi-degree-of-freedom robotic arms, and real-time image processing algorithms. They acquire tissue images by generating high-frequency sound waves through ultrasound transducers, and, combined with electromagnetic or optical positioning technology, achieve sub-millimeter-level motion control of the robotic arm, ensuring stable probe contact with the detection site. Some models incorporate AI-assisted modules that can automatically identify lesion areas and optimize scanning paths. They also support remote control, balancing imaging clarity and operational flexibility in scenarios such as minimally invasive surgical navigation and bedside monitoring in intensive care units, reducing human error.

[0003] However, most existing ultrasound robots lack intelligent detection and corresponding handling methods for abnormal situations, and often only have simple error reporting functions.

[0004] Therefore, we propose an intelligent detection and handling method for abnormal situations in ultrasonic robots. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides an intelligent detection and handling method for abnormal situations of ultrasonic robots, which can effectively solve the problems of the existing technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions;

[0007] This invention discloses an intelligent detection and processing method for abnormal conditions of an ultrasonic robot, comprising:

[0008] During the operation of the ultrasonic robot, real-time acquisition of ultrasonic image data of the target detection area and the robot's own operating status parameters is performed based on timestamp alignment. Environmental and equipment noise interference is eliminated from the ultrasonic image data based on pixel neighborhood correlation. Dynamic screening intervals are constructed for the operating status parameters based on their preset normal operating condition fluctuation range to initially eliminate values ​​outside the range. Simultaneously, the preprocessing results are correlated and aligned based on the acquisition time sequence of the two types of data. Based on the preprocessed parameter data, texture gradient features and grayscale mean deviation features of the ultrasonic images, and instantaneous fluctuation amplitude features and continuous change trend features of the operating status parameters are extracted. Fusion weights are assigned according to the correlation between each feature and the abnormal situation to construct a multi-dimensional abnormal feature matrix. This multi-dimensional abnormal feature matrix is ​​input into a preset intelligent detection model. The intelligent detection model dynamically updates the abnormal judgment threshold based on the ultrasonic robot's historical normal operating condition dataset, performs dimension-by-dimensional abnormal identification on the feature matrix, and outputs the abnormality type. An adaptive processing strategy is generated based on the output abnormality type and the preset severity grading standard.

[0009] For minor anomalies, the operating parameters of the corresponding modules will be automatically adjusted.

[0010] The module associated with a moderately abnormal fault trigger operates in a degraded mode.

[0011] In case of severe anomalies, immediately suspend robot operation and locate the faulty components;

[0012] After the ultrasonic robot has executed the processing strategy, the parameter data is re-acquired and analyzed to verify whether the anomaly has been eliminated. Simultaneously, the anomaly information, processing strategy and verification results are stored in the robot's local database.

[0013] The abnormality types include abnormal ultrasound image distortion and abnormal operating parameters exceeding limits.

[0014] Furthermore, when eliminating environmental and device noise interference based on pixel neighborhood correlation, the following applies:

[0015] For any original pixel f(x,y) in an ultrasound image, where (x,y) are the two-dimensional coordinates of the pixel in the image and f(x,y) are the original grayscale value of the pixel, all pixels within a rectangular neighborhood centered at (x,y) and of a preset size are selected. a and b are the preset neighborhood half-widths;

[0016] The weight of each neighboring pixel is calculated based on the Euclidean distance between the neighboring pixels and the center pixel. , That is, the Euclidean distance between neighboring pixels and the center pixel;

[0017] Calculate the grayscale value of the pixel after denoising, and then iterate through all pixels of the ultrasound image to complete the noise removal:

[0018] ;

[0019] In the formula: The grayscale value of the pixel (x, y) after denoising.

[0020] Furthermore, when constructing a dynamic screening interval based on the preset normal operating condition fluctuation range of the operating status parameters, a normal operating condition dataset consistent with the current operating condition during the historical operation of the ultrasonic robot is obtained. m represents the number of historical normal data samples under the same operating conditions. These are the running status parameter values ​​from the first sample; other parameters in the set are defined similarly.

[0021] Seeking The mean and standard deviation are used to construct a dynamic screening interval based on the stability coefficient of the current operating parameters. ,in This represents the mean and standard deviation. That is, the stability coefficient;

[0022] in, >0, the more stable the fluctuation of the current running status parameters, the better. The smaller the value, the better. The larger the value, and The operating parameters are dynamically adjusted based on the rate of change of the standard deviation of the operating conditions over the past three historical operating conditions. It is directly proportional to the rate of change of standard deviation.

[0023] Furthermore, when extracting the texture gradient features of the ultrasound image, the Sobel operator is used to calculate the horizontal gradient of each pixel in the ultrasound image. gradient in the vertical direction :

[0024] ;

[0025] ;

[0026] In the formula: The grayscale values ​​of the pixels adjacent to the (x, y) pixel.

[0027] in, Characterizing the rate of change of grayscale in the horizontal direction. Characterizes the rate of change of grayscale in the vertical direction;

[0028] Finally, calculate the texture gradient magnitude of this pixel. The matrix formed by the gradient magnitudes of all pixels in the ultrasound image is used as the texture gradient feature of the ultrasound image.

[0029] Furthermore, when extracting the instantaneous fluctuation amplitude characteristics of the operating state parameters, the sampling window size of the operating state parameters is set to n, where n is a preset positive integer, to obtain the operating state parameter value p(t) at the current sampling time t and the parameter values ​​at the previous n-1 sampling times. ;

[0030] The mean value of the parameters within the sampling window is calculated to determine the instantaneous fluctuation amplitude characteristics of the operating state parameters at the current moment. , This represents the mean of the parameters within the sampling window. The larger the value, the greater the deviation of the current parameter from the recent average level.

[0031] Furthermore, the multi-dimensional anomaly feature matrix, when constructed, follows the following rules:

[0032] Normalization processing was performed on the texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features of operating status parameters, and continuous change trend features of ultrasound images.

[0033] The normalization formula is ,in Represents the normalized eigenvalues. These are the original eigenvalues. These are the minimum and maximum values ​​of this feature in the historical normal operating condition dataset;

[0034] Based on the correlation between each feature and the abnormal situation of the ultrasonic robot, a fusion weight is pre-assigned to each normalized feature, and each normalized feature is multiplied by its corresponding fusion weight.

[0035] Texture gradient features after weight processing Gray-scale mean deviation characteristics Instantaneous fluctuation amplitude characteristics Continuous change trend characteristics Stack the columns to construct a multi-dimensional anomaly feature matrix. The number of rows in the matrix equals the number of feature dimensions, and the number of columns equals the number of features at the current sampling time.

[0036] in, This represents the texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features of operating state parameters, and continuous change trend features of the normalized ultrasound image. These are the fusion weights for the corresponding features, and the sum of all fusion weights is 1. T represents the transpose of the matrix.

[0037] Furthermore, the intelligent detection model dynamically updates the anomaly detection threshold based on the historical normal operating condition dataset of the ultrasonic robot:

[0038] The update cycle for the anomaly detection threshold is set to T, where T is a preset time period, and the unit is consistent with the ultrasonic robot's running time unit. After each update cycle, the operational data set generated by the ultrasonic robot during normal operation within that cycle is collected. Where q is the number of normal data samples within that period. This is the multi-dimensional feature vector corresponding to the first sampling, and the other parameters in the set are defined as follows: Similarly;

[0039] For the k-th feature dimension in the multi-dimensional anomaly feature matrix, extract the k-th eigenvalues ​​of all feature vectors in dataset H to form a feature subset. ;

[0040] Calculate feature subsets confidence quantiles and will This serves as the anomaly detection threshold for the updated feature dimension.

[0041] If the number of samples q of the normal operating condition dataset in any update cycle is less than the preset minimum sample size, the anomaly judgment threshold of the previous cycle will be used, and the dataset will be updated again after supplementing the data in the next cycle.

[0042] Furthermore, the preset severity grading criteria are as follows:

[0043] For each anomaly type in the output, calculate the feature deviation corresponding to that anomaly. ,in Indicates the current abnormal feature value. This represents the normal reference value for this feature. This indicates the normal fluctuation range of the feature, taken from the difference between the upper and lower limits of the dynamic screening interval;

[0044] Based on the degree of impact of the anomaly type on the core function of the ultrasound robot, an influence weight W is preset:

[0045] Anomalies affecting core detection functions are defined as W > 0.7, anomalies affecting auxiliary functions are defined as W < 0.4, and anomalies affecting general functions are defined as W ∈ [0.4, 0.7].

[0046] Calculate the severity index and according to Classification by level:

[0047] when < It was judged as a mild abnormality at that time. ≤ < When it is judged as moderately abnormal, ≥ It was determined to be a severe abnormality at that time, among which , The preset grading threshold is 0 < < <1.

[0048] Furthermore, when automatically adjusting the operating parameters of the corresponding operating module for a minor anomaly, the associated operating module of the ultrasonic robot corresponding to the minor anomaly and the parameters to be adjusted are first determined, and the parameter adjustment amount is calculated based on the deviation of the anomaly characteristics. ,in This indicates the preset adjustment coefficient. ∈ (0,1), That is, the degree of deviation of the abnormal features. This indicates the rated operating value of the parameter to be adjusted;

[0049] If the abnormal feature deviates from the normal range in a positive direction, then the parameter to be adjusted will be updated to... If the deviation is negative, then update to ,in This indicates the current real-time value of the parameter to be adjusted before this automatic adjustment operation for minor anomalies was performed, i.e., the instantaneous running value of the parameter before the adjustment;

[0050] After adjustment, maintain the parameter stable for the preset observation period, and then re-collect parameter data to verify the anomaly.

[0051] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:

[0052] This invention provides an intelligent detection and processing method for abnormal situations in an ultrasonic robot. During the execution of the method, when the ultrasonic robot is running, it collects ultrasonic image data and its own operating status parameters in real time based on timestamp alignment to ensure the effectiveness of data correlation. The ultrasonic image is used to eliminate environmental and equipment noise by using pixel neighborhood correlation. The operating parameters are constructed according to the historical normal operating condition fluctuation range to initially remove out-of-range values ​​and improve data quality. Then, the texture gradient and gray-level mean deviation features of the ultrasonic image and the instantaneous fluctuation amplitude and continuous change trend features of the operating parameters are extracted. A multi-dimensional abnormal feature matrix is ​​constructed by assigning weights according to the correlation between features and abnormalities to comprehensively capture abnormal signals.

[0053] Simultaneously, the intelligent detection model dynamically updates the anomaly judgment threshold based on historical normal operating condition datasets, improving the accuracy of anomaly identification. Adaptive processing strategies are generated according to the anomaly type and severity. Mild cases automatically adjust operating parameters, moderate cases trigger function downgrade operation, and severe cases suspend operation and locate fault-related components. After processing, data is collected and analyzed again to verify whether the anomaly has been eliminated and relevant information is stored, effectively improving the accuracy of anomaly detection and the timeliness of processing of ultrasonic robots, ensuring their stable operation. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0055] Figure 1 This is a flowchart illustrating an intelligent detection and handling method for abnormal situations in an ultrasonic robot. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0057] The present invention will be further described below with reference to embodiments.

[0058] Example:

[0059] This embodiment provides an intelligent detection and handling method for abnormal situations in an ultrasonic robot, such as... Figure 1 As shown, it includes:

[0060] While the ultrasonic robot is in operation, it collects real-time ultrasonic image data of the target detection area and the robot's own operating status parameters based on timestamp alignment.

[0061] For ultrasound image data, environmental and equipment noise interference is eliminated based on pixel neighborhood correlation. For operating status parameters, a dynamic screening interval is constructed based on their preset normal operating condition fluctuation range to initially eliminate values ​​that are out of range. At the same time, the correlation and alignment of preprocessing results are completed based on the acquisition time sequence of the two types of data.

[0062] After the ultrasound image data undergoes pixel neighborhood correlation denoising and the operating status parameters are dynamically screened to remove out-of-range values, a one-to-one mapping relationship is established between the ultrasound image preprocessing results and the operating status parameter preprocessing results at the same acquisition time, using the timestamps of the two types of data acquisition as unique association indices. For acquired data with slight timestamp discrepancies, linear interpolation is used to correct temporal synchronization deviations, ensuring that each set of ultrasound image preprocessing data can accurately match the corresponding operating status parameter preprocessing data, forming a spatiotemporally consistent associated dataset. Its core purpose is to eliminate potential temporal misalignment issues that may occur during the acquisition and preprocessing of the two types of data, providing data consistency assurance for the subsequent collaborative extraction and fusion of multi-dimensional abnormal features.

[0063] When eliminating environmental and device noise interference based on pixel neighborhood correlation, the following applies:

[0064] For any original pixel f(x,y) in an ultrasound image, where (x,y) are the two-dimensional coordinates of the pixel in the image and f(x,y) are the original grayscale value of the pixel, all pixels within a rectangular neighborhood centered at (x,y) and of a preset size are selected. a and b are the preset neighborhood half-widths;

[0065] The weight of each neighboring pixel is calculated based on the Euclidean distance between the neighboring pixels and the center pixel. , That is, the Euclidean distance between neighboring pixels and the center pixel, where That is, neighboring pixels The weighting for the center pixel is adjusted; the closer the pixel is to the center, the greater the weighting.

[0066] Calculate the grayscale value of the pixel after denoising, and then iterate through all pixels of the ultrasound image to complete the noise removal:

[0067] ;

[0068] In the formula: The grayscale value of the pixel (x, y) after denoising;

[0069] The above formula takes pixel neighborhood correlation as its core. By selecting a rectangular neighborhood centered on the target pixel, different weights are assigned to the neighboring pixels and the center pixel according to the Euclidean distance. The closer the distance, the greater the weight. Then, the gray value of the pixel after denoising is calculated by weighted averaging. This allows the denoising process to fit the spatial correlation characteristics between image pixels. While eliminating environmental and equipment noise interference, it preserves the detailed information of the ultrasound image to the greatest extent, ensuring the accuracy of subsequent feature extraction and avoiding feature distortion caused by noise.

[0070] Where a≥1, the value of a is larger when the noise distribution in the ultrasound image is dense or the horizontal resolution of the image is high, and the value of a is smaller when the noise distribution is sparse or the horizontal resolution of the image is low; b≥1, the value of b is larger when the noise distribution in the ultrasound image is dense or the vertical resolution of the image is high, and the value of b is smaller when the noise distribution is sparse or the vertical resolution of the image is low.

[0071] When constructing a dynamic screening interval based on the preset normal operating condition fluctuation range of the operating status parameters, a normal operating condition dataset consistent with the current operating condition during the historical operation of the ultrasonic robot is obtained. m represents the number of historical normal data samples under the same operating conditions. These are the running status parameter values ​​from the first sample; other parameters in the set are defined similarly.

[0072] Seeking The mean and standard deviation are used to construct a dynamic screening interval based on the stability coefficient of the current operating parameters. ,in This represents the mean and standard deviation. That is, the stability coefficient;

[0073] in, >0, the more stable the fluctuation of the current running status parameters, the better. The smaller the value, the better. The larger the value, and The adjustment is dynamically based on the rate of change of the standard deviation of the operating status parameter under the same operating conditions in the past three historical operating conditions, and is proportional to the rate of change.

[0074] Based on the preprocessed parameter data, texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features, and continuous change trend features of the operating status parameters of the ultrasound images are extracted respectively. The fusion weights are assigned according to the correlation between each feature and the abnormal situation to construct a multi-dimensional abnormal feature matrix.

[0075] When extracting texture gradient features from ultrasound images, the Sobel operator is used to calculate the horizontal gradient of each pixel in the ultrasound image. gradient in the vertical direction :

[0076] ;

[0077] ;

[0078] In the formula: The grayscale values ​​of the pixels adjacent to the (x, y) pixel.

[0079] in, Characterizing the rate of change of grayscale in the horizontal direction. Characterizes the rate of change of grayscale in the vertical direction;

[0080] Finally, calculate the texture gradient magnitude of this pixel. The matrix formed by the gradient magnitudes of all pixels in the ultrasound image is used as the texture gradient feature of the ultrasound image.

[0081] The above formula uses the Sobel operator to calculate the gradients of the pixel in the horizontal and vertical directions respectively. These two gradients can accurately represent the gray-level change rate in the corresponding directions. Then, the gradients in the two directions are fused into a gradient magnitude by the square root of the sum of squares. This method fully integrates the texture change information of the pixel in the two key directions of horizontal and vertical. The resulting gradient magnitude matrix can completely and accurately reflect the texture gradient features of the ultrasound image, providing richer and more reliable image feature basis for subsequent anomaly detection.

[0082] When extracting the instantaneous fluctuation amplitude characteristics of the operating status parameters, the sampling window size for the operating status parameters is set to n, where n is a preset positive integer. The operating status parameter value p(t) at the current sampling time t and the parameter values ​​at the previous n-1 sampling times are obtained. ;

[0083] The mean value of the parameters within the sampling window is calculated to determine the instantaneous fluctuation amplitude characteristics of the operating state parameters at the current moment. , This represents the mean of the parameters within the sampling window. The larger the value, the greater the deviation of the current parameter from the recent average level;

[0084] The above formula, by setting a sampling window, incorporates the parameter values ​​at the current sampling time and several previous sampling times. It first calculates the mean of the parameters within the window, and then uses the absolute value of the current parameter value compared to the mean to represent the instantaneous fluctuation amplitude. This design focuses on the changes in parameters over a recent period. By comparing with the recent average level, it intuitively and accurately reflects the degree of deviation of the current parameter, allowing the extracted instantaneous fluctuation amplitude features to more realistically reflect the dynamic abnormal trends of parameter operation.

[0085] When constructing a multidimensional anomaly feature matrix, the following rules apply:

[0086] Normalization processing was performed on the texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features of operating status parameters, and continuous change trend features of ultrasound images.

[0087] The grayscale mean deviation feature is the absolute value of the difference between the grayscale mean of the current ultrasound image and the grayscale mean of historical normal ultrasound images. The continuous change trend feature is characterized by the absolute value of the slope obtained by linear fitting of the operating status parameters.

[0088] The normalization formula is ,in Represents the normalized eigenvalues. These are the original eigenvalues. These are the minimum and maximum values ​​of this feature in the historical normal operating condition dataset;

[0089] Based on the correlation between each feature and the abnormal situation of the ultrasonic robot, a fusion weight is pre-assigned to each normalized feature, and each normalized feature is multiplied by its corresponding fusion weight.

[0090] Texture gradient features after weight processing Gray-scale mean deviation characteristics Instantaneous fluctuation amplitude characteristics Continuous change trend characteristics Stack the columns to construct a multi-dimensional anomaly feature matrix. The number of rows in the matrix equals the number of feature dimensions, and the number of columns equals the number of features at the current sampling time.

[0091] When constructing the above matrix, each feature is first normalized, then fusion weights are assigned based on the correlation between each feature and the anomaly, and multiplied with the corresponding normalized features. Finally, the matrix is ​​stacked column by column to form a matrix. This method combines image features with running parameter features, and highlights the role of highly correlated features in anomaly detection through weight allocation. At the same time, the matrix transpose operation reasonably adjusts the arrangement of feature dimensions and the number of features at the sampling time, so that the constructed matrix can comprehensively and selectively reflect multi-dimensional anomaly information, providing the intelligent detection model with clearly structured and complete input data, and improving the accuracy and efficiency of the model in anomaly identification.

[0092] in, This represents the texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features of operating state parameters, and continuous change trend features of the normalized ultrasound image. These are the fusion weights for the corresponding features, and the sum of all fusion weights is 1. T represents the transpose of the matrix.

[0093] Fusion weights of texture gradient features ∈ (0,1), the higher the proportion of image distortion anomalies in the historical operation of the ultrasonic robot, the better. The larger the value, the higher the proportion of accessibility-related exceptions. The smaller the value;

[0094] Fusion weights of gray-scale mean deviation features If ∈ (0,1), and brightness-related image distortion occurs frequently during the historical operation of the ultrasonic robot, then... The larger the value, the higher the proportion of non-luminance-related anomalies. The smaller the value;

[0095] Fusion weights of instantaneous fluctuation amplitude characteristics ∈ (0, 1), the higher the proportion of sudden out-of-limit anomalies in the parameters during the historical operation of the ultrasonic robot, the better. The larger the value, the higher the proportion of anomalies involving slow parameter changes. The smaller the value;

[0096] Fusion weights of continuous trend characteristics If the ultrasonic robot exhibits frequent gradual parameter change anomalies in its historical operation (∈ (0, 1)), then... The larger the value, the higher the proportion of sudden anomalies in the parameter. The smaller the value;

[0097] The multi-dimensional anomaly feature matrix is ​​input into the preset intelligent detection model. The intelligent detection model dynamically updates the anomaly judgment threshold based on the historical normal working condition dataset of the ultrasonic robot, performs anomaly identification on the feature matrix dimension by dimension, and outputs the anomaly type.

[0098] The intelligent detection model dynamically updates the anomaly detection threshold based on the historical normal operating condition dataset of the ultrasonic robot.

[0099] The update cycle for the anomaly detection threshold is set to T, where T is a preset time period, and the unit is consistent with the ultrasonic robot's running time unit. After each update cycle, the operational data set generated by the ultrasonic robot during normal operation within that cycle is collected. Where q is the number of normal data samples within that period. This is the multi-dimensional feature vector corresponding to the first sampling, and the other parameters in the set are defined as follows: Similarly;

[0100] For the k-th feature dimension in the multi-dimensional anomaly feature matrix, extract the k-th eigenvalues ​​of all feature vectors in dataset H to form a feature subset. ;

[0101] Calculate feature subsets confidence quantiles and will This serves as the anomaly detection threshold for the updated feature dimension.

[0102] If the number of samples q of the normal operating condition dataset in any update cycle is less than the preset minimum sample size, the anomaly judgment threshold of the previous cycle will be used, and the dataset will be updated again after supplementing the data in the next cycle.

[0103] Among them, confidence quantiles During computation, for feature subsets First, sort them in ascending order to get ,in , This represents the feature value of the m-th sample after sorting, where m = 1, 2, ..., q;

[0104] Set the default reliability as Calculate the quantile position:

[0105] ;

[0106] If pos is an integer between 1 and q, then the confidence quantiles = ;

[0107] If pos is not an integer, let pos = m + f, where m is the integer part of pos and f is the fractional part of pos, and 0 < f < 1. ;

[0108] Based on the anomaly type output and the preset severity grading criteria, an adaptive handling strategy is generated:

[0109] For minor anomalies, the operating parameters of the corresponding modules will be automatically adjusted.

[0110] The module associated with a moderately abnormal fault trigger operates in a degraded mode.

[0111] In case of severe anomalies, immediately suspend robot operation and locate the faulty components;

[0112] The preset severity grading criteria are:

[0113] For each anomaly type in the output, calculate the feature deviation corresponding to that anomaly. ,in Indicates the current abnormal feature value. This represents the normal reference value for this feature, taken from the feature mean of a historical normal operating condition dataset. This indicates the normal fluctuation range of the feature; it should be noted that for features related to operating status parameters, The value is taken from the difference between the upper and lower limits of the dynamic screening interval; for ultrasound image features, The fluctuation range of this feature in the historical normal operating condition dataset can be taken as an example. For instance, the F-value of the historical normal operating condition dataset can be calculated using the normalization formula mentioned above. max -F min ;

[0114] Based on the degree of impact of the anomaly type on the core function of the ultrasound robot, an influence weight W is preset:

[0115] If the anomaly affects the core detection function, the preset impact weight W corresponding to the anomaly type is greater than 0.7; if the anomaly affects the auxiliary function, the preset impact weight W corresponding to the anomaly type is less than 0.4; if the anomaly affects the general function, the preset impact weight W corresponding to the anomaly type is [0.4, 0.7].

[0116] Calculate the severity index and according to Classification by level:

[0117] when < It was judged as a mild abnormality at that time. ≤ < When it is judged as moderately abnormal, ≥ It was determined to be a severe abnormality at that time, among which , The preset grading threshold is 0 < < <1;

[0118] The core functions include ultrasonic signal acquisition and imaging quality control, ultrasonic probe scanning accuracy control, and robot motion axis positioning accuracy control. The auxiliary functions include automatic ultrasonic parameter calibration, robot cooling system operation, and temporary storage of detection data. The general functions include equipment status indicator display, operation panel backlight brightness adjustment, and dustproof function of equipment shell heat dissipation holes.

[0119] When automatically adjusting the operating parameters of the corresponding operating module for minor anomalies, the associated operating module of the ultrasonic robot and the parameters to be adjusted are first determined. The parameter adjustment amount is then calculated based on the deviation of the anomaly characteristics. ,in This indicates the preset adjustment coefficient. ∈ (0,1), That is, the degree of deviation of the abnormal features. This indicates the rated operating value of the parameter to be adjusted;

[0120] The above parameter adjustment formula uses a preset adjustment coefficient, anomaly deviation degree, and the rated working value of the parameter to be adjusted as variables. The adjustment amount is determined by multiplying these three factors. The adjustment amount is then subtracted or added according to the direction of the anomaly deviation to update the parameter. The setting of the adjustment coefficient can avoid excessive adjustment range leading to parameter fluctuations. The introduction of deviation degree ensures that the adjustment amount matches the severity of the anomaly, ensuring that the adjustment can effectively correct the anomaly without causing additional interference to the normal operation of the robot. At the same time, the setting of the preset observation time after adjustment can further verify the adjustment effect and ensure the effectiveness and stability of parameter adjustment.

[0121] Among these, the smaller the impact of the parameters to be adjusted on the operational stability of the core detection function of the ultrasonic robot, the better. The larger the value, the lower the value. The smaller the value;

[0122] If the abnormal feature deviates from the normal range in a positive direction, then the parameter to be adjusted will be updated to... If the deviation is negative, then update to ,in This indicates the current real-time value of the parameter to be adjusted before this automatic adjustment operation for minor anomalies was performed, i.e., the instantaneous running value of the parameter before the adjustment;

[0123] After adjustment, maintain the parameter stable for the preset observation period, and then re-collect parameter data to verify the anomaly.

[0124] After executing the processing strategy, the ultrasonic robot re-collects and analyzes parameter data to verify whether the anomaly has been eliminated. Simultaneously, the anomaly information, processing strategy, and verification results are stored in the robot's local database. If the verification finds the anomaly has not been eliminated, an iterative processing flow is initiated: First, the number of times the current processing strategy has been executed is determined. If it is the first execution, the processing intensity is increased—for minor anomalies, the parameter adjustment coefficient is increased (K is increased by 0.2 times) and then readjusted; for moderate anomalies, the upper limit of the functional module's operation is further reduced (reduced by 10% on top of the original downgrade). If the same anomaly has been processed twice and still not eliminated, the anomaly level is upgraded by one level: minor anomalies are treated as moderate anomalies, and moderate anomalies are treated as severe anomalies. The robot's operation is immediately suspended, and a manual intervention mechanism is triggered. Simultaneously, an "anomaly escalation" flag and all executed processing strategies are recorded in the local database for subsequent fault analysis. The entire iterative process executes a maximum of three rounds. If the anomaly still exists after three rounds, the robot is forcibly stopped, and a detailed fault report is generated.

[0125] The abnormality types include abnormal ultrasound image distortion and abnormal operating parameters exceeding limits.

[0126] In this embodiment, during actual operation, the ultrasonic robot can accurately collect and process data to reduce interference and eliminate outliers. It can also dynamically update judgment criteria to accurately identify image distortion and parameter exceeding limits. It responds flexibly according to the severity of the anomaly: automatic parameter adjustment for minor anomalies, downgraded operation for moderate anomalies, and immediate shutdown and location for severe anomalies. After processing, the information is verified and stored, effectively ensuring stable robot operation, improving detection accuracy, reducing the impact of malfunctions, and extending equipment lifespan.

[0127] The following examples demonstrate the application scenarios of the methods described in the above embodiments:

[0128] Example 1:

[0129] Medical institution XX uses an ultrasound robot to perform ultrasound examinations of the abdominal liver region, applying the aforementioned methods for abnormality detection and treatment. The specific process is as follows:

[0130] I. Data Acquisition and Preprocessing

[0131] During ultrasound robot operation, real-time ultrasound image data of the liver region, along with core robot operating parameters (including probe scanning speed and ultrasound signal gain), are acquired according to timestamp alignment. For ultrasound image preprocessing, due to the dense distribution of image noise and moderate resolution, a neighborhood half-width of a=2 and b=2 are set. Through pixel neighborhood correlation calculation, the original grayscale value of pixel (100, 120) with a grayscale value of 180 is denoised and its grayscale value is adjusted to 172, completing the full image noise elimination. For operating parameter preprocessing, historical normal datasets of probe scanning speed under the same operating conditions are extracted, yielding a mean of 15 mm / s and a standard deviation of 0.8 mm / s. Since recent parameter fluctuations are stable, a stability coefficient of k=1.2 is used, constructing a dynamic screening range of 13.92-16.08 mm / s. One out-of-range sampling value of 16.5 mm / s is removed. Finally, the images are correlated and aligned with the parameter preprocessing results according to the acquisition sequence.

[0132] II. Feature Extraction and Construction of Anomaly Feature Matrix

[0133] Image and parameter feature extraction: The Sobel operator is used to calculate the image texture gradient features. The gradient magnitude of pixel (80, 90) is 15, and the magnitudes of all pixels constitute the texture gradient feature matrix. The current image grayscale mean is 165, and the absolute value of the difference between it and the historical normal mean is 158 is 7, which is used as the grayscale mean deviation feature. When extracting the operating parameter features, the sampling window is set to n=5, the current probe scanning speed is 14.8mm / s, and the mean within the window is calculated to be 14.9mm / s based on the previous 4 sampling values, resulting in an instantaneous fluctuation amplitude of 0.1mm / s. Linear fitting of the scanning speed of the last 10 scans yields an absolute slope of 0.05mm / s², which is used as the continuous change trend feature. After normalizing the above features (texture gradient feature normalization value 0.5, grayscale mean deviation 0.58, instantaneous fluctuation amplitude 0.2, continuous change trend 0.25), the fusion weights are allocated according to the proportion of historical anomalies (W_G=0.35, W_B=0.2, W_fluc=0.25, W_trend=0.2, sum 1), and the weights are multiplied and then stacked and transposed in columns to construct a 4-row, 1-column multidimensional anomaly feature matrix.

[0134] III. Intelligent Detection and Adaptive Processing

[0135] The feature matrix is ​​input into the intelligent detection model. The model updates the thresholds for each dimension based on historical normal data. In a certain sampling, the instantaneous fluctuation of the probe scanning speed rises to 0.3 mm / s, which is 0.6 after normalization and 0.15 after weighting. This exceeds the abnormal threshold of 0.12 for this dimension and is judged as "abnormal operating parameters exceeding limits". Severity calculation: The deviation of this abnormal feature D≈0.046. The probe scanning speed affects the core detection function. Taking the influence weight W=0.8, the severity index SI≈0.037 (less than the preset δ1=0.1) is obtained, which is judged as a mild abnormality. Automatic parameter adjustment: The probe drive module is associated. The parameter to be adjusted is the probe speed adjustment coefficient (rated value 1.2). Since the parameter affects the core function, the adjustment coefficient K=0.3 is taken, and the adjustment amount is calculated to be 0.016. Since the parameter is positively deviated, the current value of 1.22 is updated to 1.204, and stable operation is maintained for 5 minutes for observation.

[0136] IV. Validation and Storage

[0137] The parameters were reacquired, and the instantaneous fluctuation of the probe scanning speed was reduced to 0.12 mm / s, eliminating the anomaly. Simultaneously, the results of "operating parameters slightly abnormal" and "adjusted speed adjustment coefficient to 1.204" and the verification results were stored in the robot's local database.

[0138] Example 2:

[0139] When the ultrasound robot performs thyroid ultrasound examination, it first collects ultrasound image data of the thyroid examination area in real time according to the timestamp alignment, as well as the robot's core operating status parameters (including probe scanning speed and ultrasound transmission power).

[0140] In the preprocessing stage, considering that the horizontal and vertical resolution of thyroid ultrasound images are at a moderate level and the noise distribution is moderate, a rectangular neighborhood with a half-width of a=2 and b=2 is selected with each pixel as the center. Different weights are assigned according to the Euclidean distance between the neighboring pixels and the center pixel (the closer the distance, the greater the weight). The denoised gray value of each pixel is calculated. For example, a pixel with an original gray value of 150 has its gray value adjusted to 148 after denoising, thus completing the noise elimination of the entire image. For the operating status parameters, taking the probe scanning speed as an example, a normal dataset under the same historical operating conditions is first obtained. The mean of this dataset is calculated to be 20 mm / s and the standard deviation is 1.2 mm / s. Considering that the fluctuation of this parameter has been relatively stable recently, the stability coefficient k=1.2 is set, and the dynamic screening range is constructed from 18.56 mm / s to 21.44 mm / s. The out-of-range value of 22 mm / s collected in a certain acquisition is removed. At the same time, the image denoising results are correlated and aligned with the parameter screening results according to the acquisition time sequence.

[0141] Subsequently, feature extraction was performed. For ultrasound images, the Sobel operator was used to calculate the horizontal and vertical gradients of each pixel, and then the gradient magnitudes were fused to obtain the gradient magnitudes. For example, if a pixel has a horizontal gradient of 8 and a vertical gradient of 6, the corresponding texture gradient magnitude is 10. The gradient magnitudes of all pixels together constitute the texture gradient feature. The grayscale mean of the current thyroid ultrasound image was calculated to be 130, and the absolute value of the difference between this and the historical normal thyroid ultrasound image grayscale mean of 125 was 5, which was used as the grayscale mean deviation feature. For the operating status parameters, for ultrasound transmission power, the sampling window size was set to n=5, and the current power value of 35W and the previous four sampling values ​​of 34W, 33W, 35W, and 34W were obtained. The mean within the window was calculated to be 34.2W, and the instantaneous fluctuation amplitude at the current moment was 0.8W. Linear fitting was performed on the recent ultrasound transmission power data, and the absolute value of the slope was 0.3, which was used as the continuous trend feature.

[0142] When constructing the multi-dimensional anomaly feature matrix, the above four types of features are first normalized: the historical normal range of texture gradient magnitude is 0~20, and the current feature value of 10 is normalized to 0.5; the historical normal range of grayscale mean deviation is 0 to 10, and the current feature value of 5 is normalized to 0.5; the historical normal range of instantaneous fluctuation amplitude is 0 to 3, and the current feature value of 0.8 is normalized to 0.27; the historical normal range of continuous trend slope is 0 to 1, and the current feature value of 0.3 is normalized to 0.3. Based on historical detection data, the proportion of image distortion anomalies in thyroid detection is slightly higher, so the weight of texture gradient feature W_G=0.3 is set; brightness-related image distortion is less common, so the weight of grayscale mean deviation feature W_B=0.2 is set; parameter sudden over-limit anomalies are less common, so the weight of instantaneous fluctuation amplitude feature W_fluc=0.2 is set; parameter gradual anomalies occur with a certain frequency, so the weight of continuous trend feature W_trend=0.3 (the sum of the four weights is 1). Each normalized feature is multiplied by its corresponding weight to obtain weighted feature values ​​of 0.15, 0.1, 0.054, and 0.09, respectively. These values ​​are then stacked column-wise to construct a multidimensional anomaly feature matrix.

[0143] In the intelligent detection phase, the anomaly judgment threshold update cycle is set to T=1 hour. After each cycle, the operational data set generated by the robot's normal operation within that cycle is collected (the number of samples in this cycle is q=50, meeting the preset minimum sample size requirement). For the "instantaneous fluctuation amplitude weighted feature" dimension in the feature matrix, all feature values ​​of this dimension in the dataset are extracted and sorted in ascending order. The quantile position is calculated using a preset confidence level α=0.95, resulting in an anomaly judgment threshold of 0.08 for this dimension. In a certain detection, the original value of the instantaneous fluctuation amplitude of the ultrasonic emission power was 2.7W, the normalized feature value was 0.9, and after multiplying by the weight, it was 0.18, exceeding the threshold of 0.08. The intelligent detection model outputs the anomaly type as "operational parameter exceeding limit anomaly".

[0144] When classifying the severity, the feature deviation D is first calculated: the absolute value of the difference between the current abnormal feature value (2.7W) and the normal reference value (historical average 0.6W) is 2.1W, the normal fluctuation range of the feature (the difference between the upper limit 3W and the lower limit 0W of the dynamic screening interval) is 3W, and finally D=0.7; since the ultrasonic emission power affects the robot's core detection function (probe scanning accuracy control), the influence weight is set to W=0.8, and the severity index SI is calculated to be 0.7×0.8=0.56. The preset classification thresholds are δ1=0.3 and δ2=0.6. Since 0.3≤0.56<0.6, the abnormality is judged to be a moderate abnormality.

[0145] In the adaptive processing stage, the ultrasonic transmission power control module is degraded to a lower operating mode when a moderate anomaly is triggered, and the upper limit of power output is adjusted from the original 40W to 38W to reduce the power fluctuation range.

[0146] After processing, the ultrasonic transmission power data was re-acquired, and the instantaneous fluctuation amplitude was found to have decreased to 0.6W, which is within the normal range, verifying that the anomaly has been eliminated. Simultaneously, the specific information of this "operating parameter exceeding limits anomaly," the handling strategy for "power control module function degradation," and the verification results of "anomaly elimination" were all stored in the robot's local database, completing the anomaly detection and handling process.

[0147] In summary, the method described in the above embodiments, during the operation of the ultrasonic robot, collects ultrasonic image data and its own operating status parameters in real time based on timestamp alignment to ensure the effectiveness of data correlation. It uses pixel neighborhood correlation to eliminate environmental and equipment noise in the ultrasonic images, and constructs a dynamic screening interval for operating parameters based on historical normal operating condition fluctuation ranges to initially eliminate out-of-range values, improving data quality. It then extracts ultrasonic image texture gradients, grayscale mean deviation features, and instantaneous fluctuation amplitude and continuous change trend features of operating parameters. A multi-dimensional abnormal feature matrix is ​​constructed by assigning weights according to the correlation between features and anomalies to comprehensively capture abnormal signals. Simultaneously, an intelligent detection model is applied to dynamically update the anomaly judgment threshold based on historical normal operating condition datasets, improving the accuracy of anomaly identification. An adaptive processing strategy is generated according to the anomaly type and severity: mild anomalies automatically adjust operating parameters, moderate anomalies trigger function degradation, and severe anomalies suspend operation and locate fault-related components. After processing, data is re-collected, analyzed, and verified to confirm whether the anomaly has been eliminated, and relevant information is stored. This effectively improves the accuracy and timeliness of anomaly detection in the ultrasonic robot, ensuring its stable operation.

[0148] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent detection and handling of abnormal situations in an ultrasonic robot, characterized in that, include: While the ultrasonic robot is in operation, it collects real-time ultrasonic image data of the target detection area and the robot's own operating status parameters based on timestamp alignment. For ultrasound image data, environmental and equipment noise interference is eliminated based on pixel neighborhood correlation. For operating status parameters, a dynamic screening interval is constructed based on their preset normal operating condition fluctuation range to initially eliminate values ​​that are out of range. At the same time, the correlation and alignment of preprocessing results are completed based on the acquisition time sequence of the two types of data. Based on the preprocessed parameter data, texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features, and continuous change trend features of the operating status parameters of the ultrasound images are extracted respectively. The fusion weights are assigned according to the correlation between each feature and the abnormal situation to construct a multi-dimensional abnormal feature matrix. The multi-dimensional anomaly feature matrix is ​​input into a preset intelligent detection model. The intelligent detection model dynamically updates the anomaly judgment threshold based on the historical normal working condition dataset of the ultrasonic robot, performs anomaly identification on the feature matrix dimension by dimension, and outputs the anomaly type. Based on the anomaly type output and the preset severity grading criteria, an adaptive handling strategy is generated: For minor anomalies, the operating parameters of the corresponding modules will be automatically adjusted. The module associated with a moderately abnormal fault trigger operates in a degraded mode. In case of severe anomalies, immediately suspend robot operation and locate the faulty components; After the ultrasonic robot has executed the processing strategy, the parameter data is re-acquired and analyzed to verify whether the anomaly has been eliminated. Simultaneously, the anomaly information, processing strategy and verification results are stored in the robot's local database. Among them, the abnormal types include abnormal ultrasound image distortion and abnormal operating parameters exceeding limits; When eliminating environmental and device noise interference based on pixel neighborhood correlation, the following applies: For any original pixel f(x,y) in an ultrasound image, where (x,y) are the two-dimensional coordinates of the pixel in the image and f(x,y) are the original grayscale value of the pixel, all pixels within a rectangular neighborhood centered at (x,y) and of a preset size are selected. a and b are the preset neighborhood half-widths; The weight of each neighboring pixel is calculated based on the Euclidean distance between the neighboring pixels and the center pixel. , That is, the Euclidean distance between neighboring pixels and the center pixel; Calculate the grayscale value of the pixel after denoising, and then iterate through all pixels of the ultrasound image to complete the noise removal: ； In the formula: The grayscale value of pixel (x, y) after denoising. When constructing a dynamic screening interval based on the preset normal operating condition fluctuation range of the operating status parameters, a normal operating condition dataset consistent with the current operating condition during the historical operation of the ultrasonic robot is obtained. m represents the number of historical normal data samples under the same operating conditions. These are the running status parameter values ​​from the first sample; other parameters in the set are defined similarly. Seeking The mean and standard deviation are used to construct a dynamic screening interval based on the stability coefficient of the current operating parameters. ,in This represents the mean and standard deviation. That is, the stability coefficient; in, >0, the more stable the fluctuation of the current running status parameters, the better. The smaller the value, the better. The larger the value, and The operating parameters are dynamically adjusted based on the rate of change of the standard deviation of the operating conditions over the past three historical operating conditions. It is directly proportional to the rate of change of the standard deviation; When extracting the instantaneous fluctuation amplitude characteristics of the operating state parameters, the sampling window size of the operating state parameters is set to n, where n is a preset positive integer. The operating state parameter value p(t) at the current sampling time t and the parameter values ​​at the previous n-1 sampling times are obtained. ; The mean value of the parameters within the sampling window is calculated to determine the instantaneous fluctuation amplitude characteristics of the operating state parameters at the current moment. , This represents the mean of the parameters within the sampling window. The larger the value, the greater the deviation of the current parameter from the recent average level.

2. The intelligent detection and processing method for abnormal situations in an ultrasonic robot according to claim 1, characterized in that, When extracting the texture gradient features of the ultrasound image, the Sobel operator is used to calculate the horizontal gradient of each pixel in the ultrasound image. gradient in the vertical direction : ； ； In the formula: The grayscale values ​​of the pixels adjacent to the (x, y) pixel. in, Characterizing the rate of change of grayscale in the horizontal direction. Characterizes the rate of change of grayscale in the vertical direction; Finally, calculate the texture gradient magnitude of this pixel. The matrix formed by the gradient magnitudes of all pixels in the ultrasound image is used as the texture gradient feature of the ultrasound image.

3. The intelligent detection and processing method for abnormal situations in an ultrasonic robot according to claim 1, characterized in that, The multi-dimensional anomaly feature matrix, when constructed, follows the following: Normalization processing was performed on the texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features of operating status parameters, and continuous change trend features of ultrasound images. The normalization formula is ,in Represents the normalized eigenvalues. These are the original eigenvalues. These are the minimum and maximum values ​​of this feature in the historical normal operating condition dataset; Based on the correlation between each feature and the abnormal situation of the ultrasonic robot, a fusion weight is pre-assigned to each normalized feature, and each normalized feature is multiplied by its corresponding fusion weight. Texture gradient features after weight processing Gray-scale mean deviation characteristics Instantaneous fluctuation amplitude characteristics Continuous change trend characteristics Stack the columns to construct a multi-dimensional anomaly feature matrix. The number of rows in the matrix equals the number of feature dimensions, and the number of columns equals the number of features at the current sampling time. in, This represents the texture gradient features, gray-level mean deviation features, instantaneous fluctuation amplitude features of operating state parameters, and continuous change trend features of the normalized ultrasound image. These are the fusion weights for the corresponding features, and the sum of all fusion weights is 1. T represents the transpose of the matrix.

4. The intelligent detection and processing method for abnormal situations of an ultrasonic robot according to claim 1, characterized in that, The intelligent detection model dynamically updates the anomaly detection threshold based on the historical normal operating condition dataset of the ultrasonic robot: The update cycle for the anomaly detection threshold is set to T, where T is a preset time period, and the unit is consistent with the ultrasonic robot's running time unit. After each update cycle, the operational data set generated by the ultrasonic robot during normal operation within that cycle is collected. Where q is the number of normal data samples within that period. This is the multi-dimensional feature vector corresponding to the first sampling, and the other parameters in the set are defined as follows: Similarly; For the k-th feature dimension in the multi-dimensional anomaly feature matrix, extract the k-th eigenvalues ​​of all feature vectors in dataset H to form a feature subset. ; Calculate feature subsets confidence quantiles and will This serves as the anomaly detection threshold for the updated feature dimension. If the number of samples q of the normal operating condition dataset in any update cycle is less than the preset minimum sample size, the anomaly judgment threshold of the previous cycle will be used, and the dataset will be updated again after supplementing the data in the next cycle.

5. The intelligent detection and processing method for abnormal situations in an ultrasonic robot according to claim 1, characterized in that, The preset severity grading criteria are as follows: For each anomaly type in the output, calculate the feature deviation corresponding to that anomaly. ,in Indicates the current abnormal feature value. This represents the normal reference value for this feature. This indicates the normal fluctuation range of the feature, taken from the difference between the upper and lower limits of the dynamic screening interval; Based on the degree of impact of the anomaly type on the core function of the ultrasound robot, an influence weight W is preset: If the anomaly affects the core detection function, the preset impact weight W corresponding to the anomaly type is greater than 0.7; if the anomaly affects the auxiliary function, the preset impact weight W corresponding to the anomaly type is less than 0.4; if the anomaly affects the general function, the preset impact weight W corresponding to the anomaly type is [0.4, 0.7]. Calculate the severity index and according to Classification by level: when < It was judged as a mild abnormality at that time. ≤ < When it is judged as moderately abnormal, ≥ It was determined to be a severe abnormality, among which , The preset grading threshold is 0 < < <1.

6. The intelligent detection and processing method for abnormal situations in an ultrasonic robot according to claim 1, characterized in that, When automatically adjusting the operating parameters of the corresponding operating module for minor anomalies, the associated operating module of the ultrasonic robot corresponding to the minor anomaly and the parameters to be adjusted are first determined, and the parameter adjustment amount is calculated based on the deviation of the anomaly characteristics. ,in This indicates the preset adjustment coefficient. ∈ (0,1), That is, the degree of deviation of the abnormal features. This indicates the rated operating value of the parameter to be adjusted; If the abnormal feature deviates from the normal range in a positive direction, then the parameter to be adjusted will be updated to... ; If it is a negative deviation, then update to ,in This indicates the current real-time value of the parameter to be adjusted before this automatic adjustment operation for minor anomalies was performed, i.e., the instantaneous running value of the parameter before the adjustment; After adjustment, maintain the parameter stable for the preset observation period, and then re-collect parameter data to verify the anomaly.

Images