Intelligent evaluation method and device for multi-modal test data

Abstract

The application discloses a kind of intelligent evaluation method and device of multimodal test data, the method comprises: the multimodal test data set in equipment performance test process is obtained by acquisition;The multimodal test data set includes the multi-source data subset of each technical index measured in equipment performance test process;The preprocessing data set is obtained by pre-processing to the multimodal test data set;The multi-source data subset of each technical index in the preprocessing data set is respectively fused and evaluated, and the evaluation value of each technical index is obtained;The evaluation value of all technical indexes is subjected to performance comprehensive evaluation processing, and the test evaluation value of equipment performance is obtained.The intelligent evaluation method can flexibly adjust the weight distribution according to the characteristics of different technical indexes, thereby significantly improving the accuracy and reliability of the evaluation result, and providing strong support for the comprehensive evaluation of equipment performance.

Description

Technical Field

[0001] This invention relates to the fields of industrial data processing and equipment system performance evaluation, specifically to an intelligent evaluation method and device for multimodal test data. Background Technology

[0002] In the field of equipment performance testing, with the continuous development of technology and the increasing complexity of testing, the data generated during the testing process exhibits multimodal and heterogeneous characteristics. Traditional test data evaluation methods mainly rely on the analysis of single-modal data (such as numerical data), ignoring the rich information contained in other modal data such as images and text. This single-modal data evaluation approach has many limitations. For example, relying solely on numerical data is insufficient to fully reflect the performance of equipment under complex operating conditions, while image data and text data (such as records and descriptions by test personnel) often contain key fault information and abnormal situations, but this information is often overlooked in traditional evaluation methods. Furthermore, existing methods lack systematic processing of multi-source heterogeneous data in the data preprocessing stage, leading to problems such as data noise, inconsistent formats, and abnormal data that seriously affect the accuracy of the evaluation results.

[0003] Therefore, how to effectively integrate multimodal test data and achieve a comprehensive and accurate evaluation of equipment performance through intelligent evaluation methods is a technical problem that urgently needs to be solved in the field of equipment performance testing. Summary of the Invention

[0004] This invention primarily addresses the problem of how to effectively integrate multimodal test data and achieve a comprehensive and accurate evaluation of equipment performance through intelligent evaluation methods. This invention discloses an intelligent evaluation method and device for multimodal test data.

[0005] In a first aspect, this invention discloses an intelligent evaluation method for multimodal test data, comprising:

[0006] S1, Collect a multimodal test dataset during the equipment performance test process; the multimodal test dataset includes a subset of multi-source data for each technical indicator measured during the equipment performance test; the subset of multi-source data includes image data sequences, text data sequences, and test data sequences;

[0007] S2, preprocess the multimodal test dataset to obtain a preprocessed dataset;

[0008] S3, perform fusion evaluation on the multi-source data subsets of each technical indicator in the preprocessed dataset to obtain the evaluation value of each technical indicator;

[0009] S4 performs a comprehensive performance evaluation on all technical indicators to obtain the test evaluation value of the equipment performance.

[0010] The preprocessing of the multi-source heterogeneous data set to obtain a preprocessed dataset includes:

[0011] S21, perform data cleaning processing on the multi-source heterogeneous data set to obtain a first data set;

[0012] S22, perform format detection processing on the first data set to obtain the second data set;

[0013] S23, perform data pattern matching detection processing on the second dataset to obtain a preprocessed dataset.

[0014] The process of performing data pattern matching and detection on the second dataset to obtain a preprocessed dataset includes:

[0015] S231, For each type of data attribute in the second data set, using the data collection information of the data as the independent variable and the data value of the data as the dependent variable, perform autoregressive-moving average modeling to obtain the regression model of the data attribute.

[0016] S232, using the regression model, calculate and process the independent variables to obtain regression data values; determine whether the absolute value of the difference between the regression data value and the corresponding dependent variable value is greater than a set first regression discrimination threshold; if it is greater than the first regression discrimination threshold, delete the data from the second data set; if it is less than or equal to the first regression discrimination threshold, do not process the data.

[0017] S233, perform fusion processing on all the data after execution data pattern matching detection processing in the second data set to obtain a preprocessed dataset.

[0018] The step of fusing and evaluating a subset of multi-source data for each technical indicator in the preprocessed dataset to obtain an evaluation value for each technical indicator includes:

[0019] S31, using the training dataset corresponding to each technical indicator, train the preset evaluation network to obtain the evaluation network corresponding to each technical indicator.

[0020] S32, input the image data sequence of each technical indicator in the multi-source data subset of the preprocessed dataset into the evaluation network corresponding to the technical indicator to obtain the corresponding image evaluation value;

[0021] S33, perform text evaluation processing on the text data sequence in the multi-source data subset of each technical indicator in the preprocessed dataset to obtain the corresponding text evaluation value;

[0022] S34, Perform numerical evaluation processing on the test data sequence in the multi-source data subset of each technical indicator in the preprocessed dataset to obtain the corresponding numerical evaluation value.

[0023] S35 performs a weighted summation of the image evaluation value, text evaluation value, and numerical evaluation value corresponding to each technical indicator to obtain the evaluation value of each technical indicator.

[0024] The evaluation network includes a first processing unit, a second processing unit, and a fully connected unit;

[0025] The output of the first processing unit is connected to the input of the second processing unit; the output of the second processing unit is connected to the input of the fully connected layer.

[0026] The input to the first processing unit is a sequence of image data;

[0027] The first processing unit includes a convolutional layer, a pooling layer, and a fully connected layer;

[0028] The convolutional layer utilizes a two-dimensional convolutional kernel w∈R 3×3 Feature extraction is performed on the image data sequence to obtain the feature matrix C. n ;

[0029] The pooling layer uses the max pooling method for feature extraction. The calculation expression for the max pooling method is as follows:

[0030] p u =Max 2×2 [C n ],

[0031] Where u represents the number of pooling operations, and Max... 2×2 This represents the max pooling operation for a 2×2 matrix, p u Features extracted;

[0032] The fully connected layer, for the feature p u A dimensional transformation is performed to obtain variables that match the input dimension of the second processing unit; the output of the fully connected layer is connected to the input of the second processing unit.

[0033] The second processing unit is used to perform time-series attribute extraction processing on the input data to obtain an output value, and input the output value into the fully connected unit;

[0034] The fully connected unit is used to calculate and process the output value of the second processing unit to obtain the image evaluation value.

[0035] The comprehensive performance evaluation of all technical indicators yields the test evaluation values ​​for equipment performance, including:

[0036] S41, Obtain the standard evaluation value for each technical indicator;

[0037] S42 performs a difference fusion calculation on the evaluation values ​​of all technical indicators and the corresponding standard evaluation values ​​to obtain the test evaluation values ​​of equipment performance.

[0038] The expression for the difference fusion calculation process is:

[0039]

[0040] Where qe is the test evaluation value of equipment performance, q i Let e ​​be the evaluation value of the i-th technical indicator. i Let be the standard evaluation value corresponding to the i-th technical indicator, and N be the number of technical indicators.

[0041] According to a second aspect of the present invention, an intelligent evaluation device for multimodal test data is disclosed, the device comprising:

[0042] Memory containing executable program code;

[0043] A processor coupled to the memory;

[0044] The processor calls the executable program code stored in the memory to execute the intelligent evaluation method for multimodal test data.

[0045] In a third aspect, the present invention discloses a computer-storable medium storing computer instructions, which, when invoked by a computer, are used to execute the intelligent evaluation method for multimodal test data.

[0046] In a fourth aspect, the present invention discloses an information data processing terminal, which is used to implement the intelligent evaluation method for the multimodal test data.

[0047] The beneficial effects of this invention are as follows:

[0048] The intelligent evaluation method for multimodal test data of the present invention effectively solves the problems of single data modality and insufficient preprocessing in the prior art by collecting multimodal datasets, including images, text and numerical data, during the equipment performance test process, and performing systematic preprocessing and fusion evaluation on them.

[0049] For each technical indicator's multi-source data subset, this invention utilizes evaluation networks, text evaluation, and numerical evaluation methods respectively, and obtains a comprehensive evaluation value through weighted summation, achieving deep fusion and complementarity of multimodal data. This fusion evaluation method not only fully leverages the intuitiveness of image data, the descriptiveness of text data, and the precision of numerical data, but also flexibly adjusts weight allocation according to the characteristics of different technical indicators, thereby significantly improving the accuracy and reliability of the evaluation results. Finally, by comprehensively processing the evaluation values ​​of all technical indicators, an overall experimental evaluation value of equipment performance is obtained, providing strong support for the comprehensive evaluation of equipment performance. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention. Detailed Implementation

[0051] To better understand the content of this invention, an embodiment is provided here.

[0052] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0053] In a first aspect, this invention discloses an intelligent evaluation method for multimodal test data, comprising:

[0054] S1, Collect a multimodal test dataset during the equipment performance test process; the multimodal test dataset includes a subset of multi-source data for each technical indicator measured during the equipment performance test; the subset of multi-source data includes image data sequences, text data sequences, and test data sequences;

[0055] S2, preprocess the multimodal test dataset to obtain a preprocessed dataset;

[0056] S3, perform fusion evaluation on the multi-source data subsets of each technical indicator in the preprocessed dataset to obtain the evaluation value of each technical indicator;

[0057] S4 performs a comprehensive performance evaluation on all technical indicators to obtain the test evaluation value of the equipment performance.

[0058] The preprocessing of the multi-source heterogeneous data set to obtain a preprocessed dataset includes:

[0059] The multi-source heterogeneous data set is cleaned to obtain a first data set;

[0060] The first data set is subjected to format detection processing to obtain the second data set;

[0061] Perform data pattern matching detection on the second dataset to obtain a preprocessed dataset;

[0062] The process of performing data pattern matching and detection on the second dataset to obtain a preprocessed dataset includes:

[0063] For each data attribute in the second dataset, using the data collection information of the data as the independent variable and the data value of the data as the dependent variable, autoregressive-moving average modeling is performed to obtain the regression model of the data attribute.

[0064] Using the regression model, the independent variables are calculated and processed to obtain regression data values; it is determined whether the absolute value of the difference between the regression data value and the corresponding dependent variable value is greater than a set first regression discrimination threshold; if it is greater than the first regression discrimination threshold, the data is deleted from the second data set; if it is less than or equal to the first regression discrimination threshold, the data is not processed.

[0065] The data from all the data processed by the execution data pattern matching detection in the second dataset are fused to obtain a preprocessed dataset.

[0066] This invention employs a multi-step preprocessing process involving data cleaning, format detection, and data pattern matching detection to effectively remove noisy and abnormal data, ensuring the quality and consistency of input data.

[0067] The data cleaning process includes filling in missing values, smoothing noisy data, and smoothing or deleting outliers. Smoothing noisy data involves first identifying the noisy data, and then smoothing it based on the data preceding and following it. The noisy data refers to values ​​whose values ​​are less than the sensor's detection sensitivity for the observed data, or greater than the sensor's measurement upper limit for the observed data. Outlier identification can be performed using a Kalman filter. The filling values ​​for missing values ​​can be determined by averaging the measurements within a certain sampling interval before and after the missing value.

[0068] The format detection process includes: detecting whether the data format of each data attribute in the first data set is consistent with a preset format, and deleting the data with inconsistent formats from the first data set.

[0069] The step of fusing and evaluating a subset of multi-source data for each technical indicator in the preprocessed dataset to obtain an evaluation value for each technical indicator includes:

[0070] S31, using the training dataset corresponding to each technical indicator, train the preset evaluation network to obtain the evaluation network corresponding to each technical indicator.

[0071] S32, input the image data sequence of each technical indicator in the multi-source data subset of the preprocessed dataset into the evaluation network corresponding to the technical indicator to obtain the corresponding image evaluation value;

[0072] S33, perform text evaluation processing on the text data sequence in the multi-source data subset of each technical indicator in the preprocessed dataset to obtain the corresponding text evaluation value;

[0073] S34, Perform numerical evaluation processing on the test data sequence in the multi-source data subset of each technical indicator in the preprocessed dataset to obtain the corresponding numerical evaluation value.

[0074] S35, perform weighted summation of the image evaluation value, text evaluation value and numerical evaluation value corresponding to each technical indicator to obtain the evaluation value of each technical indicator;

[0075] The evaluation network includes a first processing unit, a second processing unit, and a fully connected unit;

[0076] The output of the first processing unit is connected to the input of the second processing unit; the output of the second processing unit is connected to the input of the fully connected layer.

[0077] The input to the first processing unit is a sequence of image data;

[0078] The first processing unit includes a convolutional layer, a pooling layer, and a fully connected layer;

[0079] The convolutional layer utilizes a two-dimensional convolutional kernel w∈R 3×3 Feature extraction is performed on the image data sequence to obtain the feature matrix C. n ;

[0080] The pooling layer uses the max pooling method for feature extraction. The calculation expression for the max pooling method is as follows:

[0081] p u =Max 2×2 [C n ],

[0082] Where u represents the number of pooling operations, and Max... 2×2 This represents the max pooling operation for a 2×2 matrix, p u Features extracted;

[0083] The fully connected layer, for the feature p u A dimensional transformation is performed to obtain variables that match the input dimension of the second processing unit; the output of the fully connected layer is connected to the input of the second processing unit.

[0084] The second processing unit is used to perform time-series attribute extraction processing on the input data to obtain an output value, and input the output value into the fully connected unit;

[0085] The fully connected unit is used to calculate and process the output value of the second processing unit to obtain the image evaluation value;

[0086] The convolutional layer consists of 13 convolutional modules cascaded together; the max pooling layer consists of 5 max pooling modules cascaded together; the fully connected layer consists of 3 fully connected modules cascaded together; the convolutional layer is connected to the pooling layer, and the pooling layer is connected to the fully connected layer.

[0087] The calculation expression for the convolutional layer is:

[0088]

[0089] Where n represents the number of convolution operations, m represents the number of convolution kernels, and p i Let represent the i-th feature matrix obtained, f represent the non-linear activation function, · represent the corresponding operation between the shared weights of the convolution kernel and the feature matrix, w represent the weights of the convolution kernel, b represent the bias value, and Ri represent the bias value. 3×3 Represents a 3×3 real matrix;

[0090] The dimensional transformation process can be implemented using the Reshape function.

[0091] The fully connected unit can be implemented using an FC layer.

[0092] The processing procedure of the second processing unit is expressed as follows:

[0093] o t =g(V st +V′ sT′ ),

[0094]

[0095] Among them, s t s′ represents the output of the positive timing input of the second processing unit at time t. t U represents the output of the inverse timing input of the second processing unit at time t. Xt U′ represents the initial input of the forward timing of the second processing unit. Xt This represents the initial input quantity for the reverse timing of the second processing unit. This represents the positive timing input of the second processing unit at the previous time step. For the reverse timing input at the next time step, o tLet t represent the standard output of the second processing unit at time t, and g and f be the activation function and sigmoid function, respectively.

[0096] The output of the second processing unit includes s t ,s′ t o t .

[0097] The input quantity of the forward time series is the extracted feature p. u ;

[0098] The input quantity of the reverse timing sequence is the extracted feature p. u The reverse order;

[0099] The second processing unit can be implemented using a bidirectional LSTM network.

[0100] The comprehensive performance evaluation of all technical indicators yields the test evaluation values ​​for equipment performance, including:

[0101] S41, Obtain the standard evaluation value for each technical indicator;

[0102] S42 performs a difference fusion calculation on the evaluation values ​​of all technical indicators and the corresponding standard evaluation values ​​to obtain the test evaluation values ​​of equipment performance.

[0103] The expression for the difference fusion calculation process is:

[0104]

[0105] Where qe is the test evaluation value of equipment performance, q i Let e ​​be the evaluation value of the i-th technical indicator. i Let be the standard evaluation value corresponding to the i-th technical indicator, and N be the number of technical indicators.

[0106] The aforementioned difference fusion calculation process, by fusing the evaluation values ​​of all technical indicators with the standard evaluation values, can comprehensively and accurately reflect the overall performance of the equipment. This expression quantitatively integrates the differences between each technical indicator and the standard value in a weighted manner, avoiding the one-sidedness of single-indicator evaluation and fully considering the multi-dimensional characteristics of equipment performance. Furthermore, the calculation process based on this expression is simple and efficient, capable of rapidly processing large amounts of technical indicator data. It can promptly generate accurate test evaluation values ​​after equipment performance testing, providing a reliable basis for equipment optimization, quality inspection, and acceptance decisions, greatly enhancing the scientific rigor and practicality of equipment performance evaluation.

[0107] The text evaluation process includes:

[0108] Obtain a standard text data sequence;

[0109] For text data sequences and standard text data sequences, autoregressive-moving average modeling was performed to obtain the first regression model and the second regression model, respectively.

[0110] Extract the coefficient vectors of the two regression models, calculate their cross-correlation matrix, and then calculate the maximum eigenvalue of the cross-correlation matrix.

[0111] The maximum feature value is evaluated and calculated to obtain the text evaluation value;

[0112] The expression for the evaluation calculation process is:

[0113]

[0114] Where pg is the text evaluation value, M is the length of the text data sequence, and f max For the largest eigenvalue, a i For the i-th item in the text data sequence, la i It is the i-th item in the standard text data sequence.

[0115] This expression extracts features from text data sequences and standard text data sequences using autoregressive moving average modeling, and constructs an evaluation calculation logic based on feature coefficients. It not only meticulously captures the differences between text data sequences and standard sequences in various aspects, calculating the relative degree of difference, but also incorporates the maximum eigenvalue of the regression model features and uses an exponential function to weight the differences, highlighting the impact of significant differences. This enables precise evaluation of the quality and accuracy of text data sequences. In the evaluation scenario of text data (such as test reports and operating instructions) involved in equipment performance testing, this expression can effectively determine whether the text data conforms to standard specifications, providing a scientific and effective quantitative means for the reliability evaluation of text information during equipment testing, and improving the accuracy and comprehensiveness of text evaluation in multimodal test data.

[0116] The numerical evaluation process includes:

[0117] Obtain the standard measurement values ​​corresponding to the technical specifications;

[0118] The test data sequence is processed and compared with standard measurement values ​​to obtain numerical evaluation values;

[0119] The expression for calculating the numerical evaluation value is as follows:

[0120]

[0121] Where P is the length of the test data sequence, R is the numerical evaluation value, and p iLet p0 be the i-th element of the test data sequence, and p0 be the standard measurement value.

[0122] The training dataset corresponding to each technical indicator can be constructed based on historical image data and indicator evaluation values ​​during equipment testing.

[0123] The weight values ​​for the weighted summation in S35 can be 0.2, 0.4, and 0.3 for the image evaluation value, text evaluation value, and numerical evaluation value, respectively.

[0124] The data attributes include images, text, and test data;

[0125] The data collection information refers to the collection time information.

[0126] The text data sequence is obtained by converting the collected text into a data vector using the word2vec function.

[0127] The step of training a pre-defined evaluation network using the training dataset corresponding to each technical indicator to obtain the evaluation network corresponding to each technical indicator includes:

[0128] Obtain the training dataset corresponding to the technical indicators; the training dataset includes training data and corresponding label information; the label information is image evaluation value; the training data is a sequence of image data.

[0129] Initialize the number of training iterations;

[0130] The training data in the training dataset is used as input data and input into the evaluation network;

[0131] The input data is processed using the evaluation network to obtain predicted values;

[0132] The difference between the predicted value and the label information corresponding to the input data is calculated to obtain the difference value.

[0133] Determine whether the difference value satisfies the convergence condition to obtain the first determination result;

[0134] When the first judgment result is negative, it is determined whether the training iteration count value is equal to the training count threshold to obtain the second judgment result;

[0135] When the second judgment result is negative, the model training state is determined to be that the termination training condition is not met.

[0136] When the second judgment result is yes, it is determined that the model training state meets the termination training condition;

[0137] When the first judgment result is yes, it is determined that the model training state meets the termination training condition;

[0138] When the model training state does not meet the termination training condition, the parameters of the evaluation network are updated using the parameter update model, the training iteration count is increased by 1, and the training data in the training dataset is used as input data to the evaluation network.

[0139] When the model training state meets the termination training condition, the training process of the evaluation network is completed, and the trained evaluation network is obtained.

[0140] The difference value satisfies the convergence condition when it is less than a preset convergence threshold; the difference value does not satisfy the convergence condition when it is not less than a preset convergence threshold.

[0141] The difference calculation process can be implemented using a loss function.

[0142] The loss function can be the cross-entropy loss function.

[0143] The parameter update model is as follows:

[0144]

[0145] θ←θ+v;

[0146] In the formula, Let v be the difference value calculated for the i-th training data in the training dataset, θ be the parameter update value, η be the parameters of the evaluation network, α be the initial parameter learning rate, and α be the momentum angle parameter, where 0 ≤ α ≤ π / 4. This indicates taking the partial derivative with respect to the variable θ;

[0147] The comprehensive performance evaluation of all technical indicators yields the test evaluation values ​​for equipment performance, including:

[0148] Obtain the standard evaluation value for each technical indicator;

[0149] A comprehensive performance evaluation calculation is performed on the evaluation values ​​and standard evaluation values ​​of all technical indicators to obtain the test evaluation value of the equipment performance;

[0150] The expression for the comprehensive performance evaluation calculation is as follows:

[0151]

[0152] Where zi is the evaluation value of the i-th technical indicator, bzi is the standard evaluation value of the i-th technical indicator, K is the number of technical indicators, and S is the test evaluation value of equipment performance.

[0153] According to a second aspect of the present invention, an intelligent evaluation device for multimodal test data is disclosed, the device comprising:

[0154] Memory containing executable program code;

[0155] A processor coupled to the memory;

[0156] The processor calls the executable program code stored in the memory to execute the intelligent evaluation method for multimodal test data.

[0157] In a third aspect, the present invention discloses a computer-storable medium storing computer instructions, which, when invoked by a computer, are used to execute the intelligent evaluation method for multimodal test data.

[0158] In a fourth aspect, the present invention discloses an information data processing terminal, which is used to implement the intelligent evaluation method for the multimodal test data.

[0159] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. An intelligent evaluation method for multimodal test data, characterized in that, include: S1, collect the multimodal test dataset during the equipment performance test process; The multimodal test dataset includes a subset of multi-source data for each technical indicator measured during equipment performance testing; The multi-source data subset includes image data sequences, text data sequences, and test data sequences; S2, preprocess the multimodal test dataset to obtain a preprocessed dataset, including: S21, perform data cleaning processing on the multi-source heterogeneous data set to obtain a first data set; S22, perform format detection processing on the first data set to obtain the second data set; S23, perform data pattern matching detection processing on the second dataset to obtain a preprocessed dataset, including: S231, For each type of data attribute in the second data set, using the data collection information of the data as the independent variable and the data value of the data as the dependent variable, perform autoregressive-moving average modeling to obtain the regression model of the data attribute. S232, using the regression model, calculate and process the independent variables to obtain regression data values; determine whether the absolute value of the difference between the regression data value and the corresponding dependent variable value is greater than a set first regression discrimination threshold; if it is greater than the first regression discrimination threshold, delete the data from the second data set; if it is less than or equal to the first regression discrimination threshold, do not process the data. S233, perform fusion processing on all data after execution data pattern matching detection processing in the second data set to obtain a preprocessed dataset; S3, perform fusion evaluation on the multi-source data subsets of each technical indicator in the preprocessed dataset to obtain the evaluation value of each technical indicator, including: S31, using the training dataset corresponding to each technical indicator, train the preset evaluation network to obtain the evaluation network corresponding to each technical indicator. S32, input the image data sequence of each technical indicator in the multi-source data subset of the preprocessed dataset into the evaluation network corresponding to the technical indicator to obtain the corresponding image evaluation value; S33, perform text evaluation processing on the text data sequence in the multi-source data subset of each technical indicator in the preprocessed dataset to obtain the corresponding text evaluation value; S34, Perform numerical evaluation processing on the test data sequence in the multi-source data subset of each technical indicator in the preprocessed dataset to obtain the corresponding numerical evaluation value. S35, perform weighted summation of the image evaluation value, text evaluation value and numerical evaluation value corresponding to each technical indicator to obtain the evaluation value of each technical indicator; S4 performs a comprehensive performance evaluation on all technical indicators to obtain the test evaluation values ​​of the equipment performance, including: S41, Obtain the standard evaluation value for each technical indicator; S42, perform difference fusion calculation on the evaluation values ​​of all technical indicators and the corresponding standard evaluation values ​​to obtain the test evaluation values ​​of equipment performance; The expression for the difference fusion calculation process is: Where qe is the test evaluation value of equipment performance, q i Let e ​​be the evaluation value of the i-th technical indicator. i Let be the standard evaluation value corresponding to the i-th technical indicator, and N be the number of technical indicators.

2. The intelligent evaluation method for multimodal test data as described in claim 1, characterized in that, The evaluation network includes a first processing unit, a second processing unit, and a fully connected unit; The output of the first processing unit is connected to the input of the second processing unit; the output of the second processing unit is connected to the input of the fully connected layer. The input to the first processing unit is a sequence of image data; The first processing unit includes a convolutional layer, a pooling layer, and a fully connected layer; The convolutional layer utilizes a two-dimensional convolutional kernel w∈R 3×3 Feature extraction is performed on the image data sequence to obtain the feature matrix C. n ; The pooling layer uses the max pooling method for feature extraction. The calculation expression for the max pooling method is as follows: p u =Max 2×2 [C n ], Where u represents the number of pooling operations, and Max... 2×2 This represents the max pooling operation for a 2×2 matrix, p u Features extracted; The fully connected layer, for the feature p u A dimensional transformation is performed to obtain variables that match the input dimension of the second processing unit; the output of the fully connected layer is connected to the input of the second processing unit. The second processing unit is used to perform time-series attribute extraction processing on the input data to obtain an output value, and input the output value into the fully connected unit; The fully connected unit is used to calculate and process the output value of the second processing unit to obtain the image evaluation value.

3. An intelligent evaluation device for multimodal test data, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the intelligent evaluation method for multimodal test data as described in any one of claims 1 to 2.

4. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by the computer, are used to execute the intelligent evaluation method for multimodal test data as described in any one of claims 1 to 2.

5. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the intelligent evaluation method for multimodal test data as described in any one of claims 1 to 2.

Images