Intelligent processing method and system for radiator salt spray test data

Abstract

The invention relates to the field of radiator testing, in particular to a radiator salt spray test data intelligent processing method and system.The method comprises the steps that test data in the salt spray test process are obtained and preprocessed, and fitting curves of salt spray concentration data and corrosion current density data are generated respectively; acquiring a peak point of the fitting curve of the salt mist concentration data as a segmentation point, matching the two curves through a dynamic time warping algorithm, and calculating a matching degree to determine the segmentation point on the fitting curve of the corrosion current density data; and calculating a correlation value based on the time difference of the matching point pairs and the shortest path distance, dynamically determining weight distribution of the salt mist concentration data and the corrosion current density data in a forgetting gate formula, correcting the forgetting gate formula, and performing test data prediction by using the corrected LSTM model. According to the method, the correlation values of the corresponding segments of the two curves are calculated, and the weights of the salt mist concentration data and the corrosion current density data in the forgetting gate are dynamically adjusted, so that the adaptability of the model to complex data and the prediction accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of radiator testing. In particular, it relates to an intelligent processing method and system for salt spray test data of radiators. Background Art

[0002] As a key heat exchange component in various devices, radiators are applied in multiple fields such as industrial equipment, automobiles, and electronics. By efficiently conducting heat, the heat generated during the operation of the equipment is dissipated into the surrounding environment, which is crucial for ensuring the stable operation of the equipment, preventing overheating failures, and extending the service life.

[0003] However, the actual working environment of radiators is often complex and changeable. Especially in coastal areas and some special industrial environments, high-salt air makes salt spray corrosion a key factor affecting their performance and lifespan. To address this challenge, salt spray tests have emerged. By simulating the salt spray conditions in the natural environment, artificial accelerated corrosion tests are carried out on radiators to evaluate their environmental adaptability in harsh environments. With the help of salt spray tests, the performance of radiators during long-term use can be predicted in advance in the laboratory environment, ensuring that they can still maintain a stable and reliable working state in different environments and meet the operation requirements of various devices in diverse scenarios.

[0004] During the salt spray test process, in the prior art, an LSTM model is usually used to analyze and predict the data parameters of radiators. By learning the time series characteristics in the salt spray test data, the risk of decreased heat dissipation efficiency or poor contact can be detected in advance, and the time for the radiator to reach the critical corrosion depth in the salt spray environment can be estimated. However, the fixed forgetting gate in the LSTM model cannot distinguish between key memories and redundant memories, and cannot adapt to the complex and changeable characteristics in the salt spray test data. Especially when dealing with data in different corrosion stages, it cannot effectively adjust the memory retention and forgetting strategies, resulting in inaccurate prediction results. Summary of the Invention

[0005] To solve the problem that the forgetting gate weight of the LSTM model is fixed and cannot adapt to the complex and changeable characteristics of salt spray test data, especially in different corrosion stages, it cannot effectively adjust the memory retention and forgetting strategies, affecting the prediction accuracy. The present invention provides solutions in the following aspects.

[0006] In a first aspect, an intelligent processing method for radiator salt spray test data includes: obtaining test data during the salt spray test and performing preprocessing, where the test data includes: data sequences of salt spray concentration data and corrosion current density data; fitting curves for the salt spray concentration data and the corrosion current density data in the test data to generate two fitting curves, and using the peak points of the fitting curves as segmentation points, matching the two fitting curves using the dynamic time warping algorithm and calculating the matching degree, determining the segmentation points on the other curve based on the matching degree to complete the corresponding segmentation of the fitting curves; obtaining the correlation values of each pair of corresponding segments according to the time difference and the shortest path distance of each pair of matching points in the corresponding segments, determining the weight distribution of the salt spray concentration data and the corrosion current density data in the forgetting gate formula based on the correlation values, and correcting the forgetting gate formula in the LSTM model to obtain the output result of the forgetting gate, and predicting the test data according to the output result of the forgetting gate.

[0007] By obtaining the test data in the salt spray test and preprocessing it, fitting the curves of the salt spray concentration data and the corrosion current density data, matching the curves using the dynamic time warping algorithm and calculating the matching degree, and determining the segmentation points on the other curve, the corresponding segmentation of the curves is realized. Calculating the correlation value based on the time difference and the shortest path distance of the matching point pairs, dynamically allocating the weights of the salt spray concentration data and the corrosion current density data in the forgetting gate formula accordingly, correcting the forgetting gate formula, and optimizing the prediction accuracy of the LSTM model. The segmentation matching accuracy is improved, and the adaptability and prediction ability of the model to different corrosion stages are enhanced.

[0008] Preferably, the preprocessing includes: Aligning the test data according to the time stamp, filling the missing parts in the test data using the linear interpolation method to ensure the integrity of the data, and normalizing the salt spray concentration data and the corrosion current density data to unify their numerical ranges.

[0009] Preferably, the fitting curves for the salt spray concentration data and the corrosion current density data in the test data to generate two fitting curves includes: Using the least squares method to respectively fit curves for the data sequences of the salt spray concentration data and the corrosion current density data to obtain a salt spray concentration data - time curve and a corrosion current density data - time curve.

[0010] The data sequences of salt spray concentration data and corrosion current density data are respectively curve-fitted by the least squares method to generate a salt spray concentration data-time curve and a corrosion current density data-time curve. Further, the dynamic time warping algorithm is used to match the two fitted curves. By accurately fitting and matching the curves, accurate data support is provided for subsequent analysis of the relationship between salt spray concentration data and corrosion current density data, improving the adaptability and prediction ability of the model to different corrosion stages.

[0011] Preferably, the matching of the two fitted curves using the dynamic time warping algorithm includes: Using the first derivative detection method for the fitted curve of salt spray concentration data, marking all peak points on the curve, and taking the peak points as the segmentation points of the fitted curve, and performing segmentation; Using the dynamic time warping algorithm to match the data points on the two curves, calculating the distances between all data points on the two curves, constructing a distance matrix, starting from the upper left corner of the distance matrix, and gradually calculating the minimum cumulative distance when reaching each data point along the shortest path to form a cumulative distance matrix. Starting from the lower right corner of the cumulative distance matrix, taking the path of the minimum cumulative distance as the optimal path, and obtaining the corresponding matching points and distance values of the data points on the two curves according to the principle of the closest distance, completing the matching of the fitted curves of salt spray concentration data and corrosion current density data.

[0012] Marking the peak points of the salt spray concentration data curve as segmentation points by the first derivative detection method can automatically identify the key feature points of the curve and ensure the rationality of segmentation. The dynamic time warping algorithm calculates the distances between all data points on the two curves, constructs a distance matrix, and finds the optimal path through dynamic programming to achieve accurate matching of the curves, effectively solving the similarity measurement problem caused by different time series lengths or speeds, and improving the accuracy and efficiency of curve fitting and matching.

[0013] Preferably, the matching of the two fitted curves using the dynamic time warping algorithm includes: Taking any segmentation point on the fitted curve of salt spray concentration data as the target segmentation point, in response to the target segmentation point on the fitted curve of salt spray concentration data corresponding to multiple matching points on the fitted curve of corrosion current density data, selecting the matching point corresponding to the maximum value of the matching degree as the segmentation point on the corrosion current density data-time curve.

[0014] By improving the accuracy of segment matching, ensuring the correspondence of key feature points, making subsequent analysis more targeted. By selecting the point with the maximum matching degree as the segmentation point, the corresponding position on the corrosion current density data curve can be accurately located, enhancing the reliability of segment matching. It enhances the adaptability of the model to the complex characteristics of salt spray test data and improves the prediction accuracy.

[0015] Preferably, the matching degree includes: Calculate the second derivative of each matching point on the fitting curve of the corrosion current density data, calculate the Euclidean distance between the segmentation points on the fitting curve of the salt spray concentration data and the matching points on the fitting curve of the corrosion current density data, map the Euclidean distance using a negative exponential function, and normalize the product of the mapped Euclidean distance and the second derivative as the matching degree between the segmentation point and the matching point.

[0016] Preferably, the correlation value includes: Taking any pair of segments as the target segment, calculate the shortest Euclidean distance between all segmentation points and matching points in the target segment, take the reciprocal of the shortest Euclidean distance, and normalize it to obtain the similarity weight; calculate the ratio of the standard deviation of the time difference between each pair of matching points in the target segment to the mean value of the time difference between each pair of matching points in the target segment to obtain the coefficient of variation of the time difference; calculate the average value between the matching degree of the left segmentation point and the matching point and the matching degree of the right segmentation point and the matching point in the target segment to obtain the comprehensive matching degree; Take the product of multiplying the similarity weight, the coefficient of variation of the time difference, and the comprehensive matching degree respectively as the correlation value of the target segment.

[0017] Preferably, the determining the weight allocation of the salt spray concentration data and the corrosion current density data in the forgetting gate formula based on the correlation value includes: Perform a weighted average of the correlation values of all segments as the weight of the salt spray concentration data in the forgetting gate , and take 1 minus the weight of the salt spray concentration data in the forgetting gate as the weight of the corrosion current density data in the forgetting gate formula ; In response to the weight of the salt spray concentration data in the forgetting gate approaching 1, the LSTM model prediction focuses more on the influence of the salt spray concentration data on the corrosion current density data. In response to the weight of the salt spray concentration data in the forgetting gate approaching 0, the LSTM model prediction pays more attention to the historical trend of the corrosion current density data.

[0018] Dynamically allocate weights through the correlation of different segments, so as to more flexibly respond to the complex changes of salt spray test data. When the weight of the salt spray concentration data approaches 1, the model focuses on the influence of the salt spray concentration data on the corrosion current density data and can more sensitively capture the driving effect of external environmental changes on corrosion; while when When approaching 0, the model pays more attention to the historical trend of the corrosion current density data, which is beneficial to maintaining the stability of prediction when the data fluctuates or the environment changes. The adaptive weight adjustment not only improves the prediction accuracy of the model for different corrosion stages, but also enhances the overall generalization ability and robustness of the model, ensuring that the prediction results are closer to the actual corrosion process.

[0019] Preferably, the output result of the forgetting gate includes: The product obtained by multiplying the salt spray concentration data at the current moment of the forgetting gate, the corresponding weight influence matrix, and the weight of the salt spray concentration data in the forgetting gate is used as the contribution value of the salt spray concentration data at the current moment; the product obtained by multiplying the corrosion current density data at the current moment of the forgetting gate, the corresponding weight influence matrix, and the weight of the corrosion current density data in the forgetting gate is used as the contribution value of the corrosion current density data at the current moment; calculate the product between the hidden state at the previous moment and the weight matrix of the hidden state at the current moment to obtain the contribution value of the hidden state at the previous moment; sum the contribution value of the salt spray concentration data, the contribution value of the corrosion current density data, the contribution value of the hidden state at the previous moment, and the bias term respectively, and use The activation function to obtain the output result of the corrected forgetting gate at the current moment.

[0020] In a second aspect, an intelligent processing method and system for radiator salt spray test data include: a processor and a memory, and the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned intelligent processing method for radiator salt spray test data is implemented.

[0021] The present invention has the following effects: 1. The present invention determines the peak point of the salt spray concentration data curve as the segmentation point through the derivative detection method, and matches the two curves through the DTW algorithm, and calculates the matching degree between the matching point and the segmentation point, so as to find the segmentation point with the highest matching degree with the segmentation point of the salt spray concentration data curve on the corrosion current density data curve. It effectively improves the accuracy of segmentation matching, ensures the maximum similarity of segmentation local features, and provides a more reliable data basis for subsequent analysis.

[0022] 2. The present invention dynamically determines the weight distribution of the salt spray concentration data and the corrosion current density data in the forgetting gate formula by calculating the correlation value between the corresponding segments of the two curves, so that the LSTM model can more accurately capture the relationship between the salt spray concentration data and the corrosion current density data, especially in different corrosion stages. Through the weight adjustment strategy, the model can better adapt to the complex and changeable characteristics of the salt spray test data, and improves the adaptability and prediction accuracy of the model for different corrosion stages. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1It is a flowchart of the method from step S1 to step S3 in the intelligent processing method of radiator salt spray test data according to an embodiment of the present invention.

[0024] Figure 2 It is a structural block diagram of an intelligent processing system for radiator salt spray test data according to an embodiment of the present invention. Detailed implementation manners

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments.

[0026] Refer to Figure 1 , an intelligent processing method for radiator salt spray test data includes steps S1 to S3, which are specifically as follows: During the radiator salt spray test, in order to evaluate the degree of performance loss of the radiator in the salt spray environment and predict the time when the radiator reaches the critical corrosion depth in the salt spray environment, it is necessary to align the timestamps of the data sequences of the salt spray concentration data and the corrosion current density data changing with time for subsequent synchronous analysis. The specific steps are as follows: S1: Obtain the test data during the salt spray test and perform preprocessing. Among them, the test data includes: the data sequences of the salt spray concentration data and the corrosion current density data.

[0027] Align the test data according to the timestamps, use the linear interpolation method to fill in the missing parts in the test data to ensure the integrity of the data, and normalize the salt spray concentration data and the corrosion current density data to unify their numerical ranges.

[0028] Through linear interpolation to fill in the missing values and normalization processing, the data becomes more complete and consistent. Timestamp alignment and data preprocessing ensure the synchronization and comparability between different data parameters, which helps to more accurately evaluate the corrosion situation and performance loss of the radiator.

[0029] S2: Fit curves to the salt spray concentration data and the corrosion current density data in the test data to generate two fitted curves, use the peak points of the fitted curves as segmentation points, use the dynamic time warping algorithm to match the two fitted curves and calculate the matching degree, and determine the segmentation points on the other curve based on the matching degree to complete the segmented correspondence of the fitted curves.

[0030] Use the least squares method to fit curves to the data sequences of the salt spray concentration data and the corrosion current density data respectively to obtain the salt spray concentration data - time curve and the corrosion current density data - time curve.

[0031] For the fitting curve of the salt spray concentration data, the first derivative detection method is used to mark all the peak points on the curve, and the peak points are used as the segmentation points of the fitting curve for segmentation. It should be noted that in the salt spray test, the curve of the salt spray concentration data changing with time usually shows the characteristic of fluctuating. In the initial stage of corrosion, the salt spray concentration data often gradually increases. However, after reaching a specific threshold, due to the formation of corrosion products and the influence of other related factors, the salt spray concentration data may show a certain degree of fluctuation. Taking the study of the cyclic salt spray accelerated corrosion behavior of high-strength aluminum alloy as an example, the change trend of the corrosion current density data shows a fluctuating increase first and then a decrease, which indicates that the change of the salt spray concentration data is not constant but complex and variable. Based on this characteristic, the peak points on the salt spray concentration data curve can be accurately determined through the extreme value detection algorithm, and these peak points are used as the segmentation points to conduct a more detailed segmented analysis of the salt spray concentration data curve.

[0032] At the same time, the curve of the corrosion current density data changing with time mostly shows a monotonically increasing trend. However, this change trend also varies due to different salt spray concentration data. Specifically, under the condition of lower salt spray concentration data, the corrosion current density data shows a gradually increasing trend over time; while when the salt spray concentration data is at a higher level, the rising speed of the corrosion current density data will be significantly accelerated. It should be noted that although the corrosion current density data curve generally shows a monotonically increasing characteristic, it does not have obvious peak points like the salt spray concentration data curve.

[0033] To achieve the segmented correspondence of the two curves, the dynamic time warping (DTW) algorithm is introduced in the experiment. Through this algorithm, the segmentation points on the salt spray concentration data curve can be accurately matched with the corresponding positions on the corrosion current density data curve. Specifically, by using the DTW algorithm to match the two fitting curves, the corresponding segmented positions on the corrosion current density data curve can be determined, thus realizing the segmented correspondence between the salt spray concentration data curve and the corrosion current density data curve. This process not only fully considers the complex relationship between the salt spray concentration data and the corrosion current density data, but also lays a solid foundation for subsequent corrosion behavior analysis and prediction.

[0034] Use the dynamic time warping algorithm to match the data points on two curves, calculate the distances between all data points on the two curves, construct a distance matrix, start from the upper left corner of the distance matrix, and gradually calculate the minimum cumulative distance when reaching each data point along the shortest path to form a cumulative distance matrix. Start from the lower right corner of the cumulative distance matrix, take the path of the minimum cumulative distance as the optimal path, and obtain the corresponding matching points and distance values of the data points on the two curves according to the principle of the closest distance, so as to complete the matching of the fitting curves of the salt spray concentration data and the corrosion current density data.

[0035] Take any segmentation point on the fitting curve of the salt spray concentration data as the target segmentation point. In response to the target segmentation point on the fitting curve of the salt spray concentration data corresponding to multiple matching points on the fitting curve of the corrosion current density data, select the matching point corresponding to the maximum value of the matching degree as the segmentation point on the corrosion current density data-time curve.

[0036] The matching degree includes: Calculate the second derivative of each matching point on the corrosion current density data-time curve, calculate the Euclidean distance between the segmentation point on the salt spray concentration data-time curve and the matching point on the corrosion current density data-time curve, map the Euclidean distance using the negative exponential function, and normalize the product of the mapped Euclidean distance and the second derivative as the matching degree between the segmentation point and the matching point.

[0037] Specifically, the matching degree satisfies the following relational expression: ; In the formula, represents the matching degree between the segmentation point and the matching point , represents the second derivative of the th matching point of the segmentation point on the corrosion current density data-time curve, , represents the exponential function with the natural number as the base, represents the Euclidean distance between the segmentation point and the matching point in the DTW matching algorithm, represents the normalization function.

[0038] That is to say, represents the correlation and reliability of the matching points on the corrosion current density data curve, is the second derivative of the matching point on the corrosion current density data curve, reflecting the degree of change of the corrosion rate, The larger it is, the greater the curvature of the corrosion current density data at this point, and the more drastic the change in the corrosion rate. Such points usually correspond to the key turning points in the corrosion process and are closely related to the change in corrosion behavior caused by the change in salt spray concentration data.

[0039] is the Euclidean distance between the segmentation point of the salt spray concentration data curve and the matching point of the corrosion current density data curve. The smaller the Euclidean distance, the closer the two points are in space, and the more consistent the change trend of the corrosion current density data is with the change trend of the salt spray concentration data.

[0040] When is larger and is smaller, is naturally larger, indicating that the matching point is both a key point with a significant change in corrosion rate on the corrosion current density data curve and has a high similarity and correlation with the segmentation point on the salt spray concentration data curve in the time series. Therefore, it is suitable as the corresponding segmentation point. Considering both the change in corrosion rate and the proximity of the two points, it is used to evaluate whether the matching point is suitable as the segmentation point.

[0041] S3: According to the time difference and the shortest path distance between each pair of matching points in the corresponding segments, obtain the correlation value of each pair of corresponding segments. Based on the correlation value, determine the weight allocation of the salt spray concentration data and the corrosion current density data in the forgetting gate formula, and modify the forgetting gate formula. Predict the experimental data according to the output result of the modified forgetting gate in the LSTM model.

[0042] The correlation value includes: Taking any pair of segments as the target segment, calculate the shortest Euclidean distance between all segmentation points and matching points in the target segment, take the reciprocal of the shortest Euclidean distance, and normalize it to obtain the similarity weight; calculate the ratio of the standard deviation of the time difference of each pair of matching points in the target segment to the mean value of the time difference of each pair of matching points in the target segment to obtain the coefficient of variation of the time difference; calculate the average value between the matching degree of the left segmentation point and the matching point and the matching degree of the right segmentation point and the matching point in the target segment to obtain the comprehensive matching degree; Multiply the similarity weight, the coefficient of variation of the time difference, and the comprehensive matching degree respectively, and take the product as the correlation value of the target segment.

[0043] Specifically, the correlation value satisfies the following relational expression: ; In the formula, represents the correlation value between the th pair of segments, represents the The shortest Euclidean distance in the segmented section, represents the total number of matching point pairs in the segmented section, represents the time difference between the th pair of matching point pairs in the segmented section, represents the mean value of the time differences between all matching point pairs in the segmented section, represents the matching degree between the left segmented point and the matching point in the segmented section, represents the matching degree between the right segmented point and the matching point in the segmented section, represents the normalization function, represents the exponential function with the natural number as the base.

[0044] That is to say, the larger the value, the greater the distance between the corresponding segmented sections of the two curves, and the lower their similarity. Therefore, is negatively correlated with the correlation value . is the coefficient of variation of the time differences of the matching point pairs in the corresponding segmented section of the curve, which can reflect the degree of dispersion of the time differences of the matching point pairs. Since the influence of the salt spray concentration data on the corrosion current density data has a lag, the time difference between the matching point pairs is the lag period length of the change of the corrosion current density data caused by the salt spray concentration data. The more consistent the lag period length in the corresponding segmented section, the smaller the coefficient of variation. At this time, the greater the influence of the salt spray concentration data on the change of the corrosion current density data, the stronger the correlation. The larger it is, the greater the influence of the salt spray concentration data on the change of the corrosion current density data in the corresponding segmented section, and the stronger the correlation.

[0045] Based on the correlation value, determine the weight allocation of the salt spray concentration data and the corrosion current density data in the forgetting gate formula, including: Perform a weighted average of the correlation values of all segmented sections as the weight of the salt spray concentration data in the forgetting gate , and use 1 minus the weight of the salt spray concentration data in the forgetting gate as the weight of the corrosion current density data in the forgetting gate formula .

[0046] It should be noted that the influence degree of the salt spray concentration data at different time periods on the change of the corrosion current density data is different. Therefore, based on the weight matrix obtained from the model training, the forgetting gate parameters are adjusted secondarily according to the correlation between the salt spray concentration data and the corrosion current density data. In response to approaching 1, the model pays more attention to the influence of the salt spray concentration data on the corrosion process. In response to When approaching 0, the model relies more on the historical trend of the corrosion current density data for prediction.

[0047] Modify the forgetting gate formula in the LSTM model to obtain the output result of the forgetting gate, including: Multiply the salt spray concentration data at the current moment of the forgetting gate, the corresponding weight influence matrix, and the weight of the salt spray concentration data in the forgetting gate respectively, and use the product as the contribution value of the salt spray concentration data at the current moment; multiply the corrosion current density data at the current moment of the forgetting gate, the corresponding weight influence matrix, and the weight of the corrosion current density data in the forgetting gate respectively, and use the product as the contribution value of the corrosion current density data at the current moment; calculate the product between the hidden state at the previous moment and the weight matrix of the hidden state at the current moment to obtain the contribution value of the hidden state at the previous moment; sum the contribution value of the salt spray concentration data, the contribution value of the corrosion current density data, the contribution value of the hidden state at the previous moment, and the bias term respectively, and use The activation function to obtain the output result of the corrected forgetting gate at the current moment.

[0048] Specifically, the output result of the corrected forgetting gate satisfies the following relational expression: ; In the formula, represents the output result of the forgetting gate, represents the activation function, represents the weight of the salt spray concentration data, represents the weight influence matrix of the salt spray concentration data, represents the salt spray concentration data, represents the weight of the corrosion current density data, is the weight influence matrix of the corrosion current density data, represents the corrosion current density data, represents the weight matrix of the hidden state at the previous moment, represents the hidden state at the previous moment, represents the bias term of the forgetting gate.

[0049] Based on the improved forgetting gate formula, use the LSTM model to predict the corrosion current density data, and calculate the corrosion depth. When the corrosion depth reaches the preset depth corrosion threshold, mark the corresponding time as the time when the radiator reaches the critical corrosion depth in the salt spray environment.

[0050] It should be noted that the prediction process of the LSTM model includes: Preprocess the data sequences of the obtained salt spray concentration data and corrosion current density data, and use them as input data to train a preset LSTM model. The process of training the model is well-known to those skilled in the art and will not be described in detail; The LSTM model structure is as follows: forget gate: determines how much information of the previous cell state to retain, input gate: controls the degree of update of the current input to the cell state, and output gate: determines the hidden state at the current moment.

[0051] Replace the forget gate formula in the existing trained LSTM model with the modified forget gate of the present invention, and input the data sequences of the salt spray concentration data and corrosion current density data to be predicted into the LSTM model with the modified forget gate. The output result is: corrosion current density data. Calculate the corrosion depth according to the predicted result, so as to judge the time to reach the critical corrosion depth in the salt spray environment.

[0052] Exemplarily, preset the depth corrosion threshold to be half of the thickness of the radiator surface material. Specifically, the implementer can set it according to the specifications of the radiator material.

[0053] Specifically, the corrosion depth satisfies the following relational expression: ; In the formula, represents the corrosion depth, represents the corrosion current density data, represents the time of the corrosion test, represents the molar mass of the radiator material, represents the oxidation state of the radiator material in the corrosion reaction, represents the Faraday constant, represents the density of the radiator material.

[0054] It should be noted that the corrosion current density data (unit: ); molar mass (unit: ); the oxidation state of the material in the corrosion reaction (i.e., the number of electrons lost). For example, aluminum is usually +3 in corrosion, then ; Faraday constant( ); the density of the radiator material (unit: ), such as aluminum( ).

[0055] It should be noted that the corrosion depth is not the focus of the present invention. There are various ways to calculate the corrosion depth, which will not be described in detail here.

[0056] The present invention also provides an intelligent processing system for radiator salt spray test data. As shown in Figure 2As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an intelligent processing method for radiator salt spray test data according to the first aspect of the present invention. The system also includes a communication bus, a communication interface, and other components well-known to those skilled in the art. Their settings and functions are known in the art, so they will not be elaborated here.

[0057] It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent shall be subject to the appended claims.

Claims

1. A method for intelligent processing of radiator salt spray test data, characterized in that: include: Acquire test data during the salt spray test and perform preprocessing, wherein the test data includes: a data sequence of salt spray concentration data and corrosion current density data; The salt spray concentration data and corrosion current density data in the test data are fitted with curves to generate two fitting curves, and the peak points of the fitting curves are used as segmentation points. The two fitting curves are matched and the matching degree is calculated using the dynamic time warping algorithm. The segmentation points on the other curve are determined based on the matching degree to complete the corresponding segmentation of the fitting curve. According to the time difference and shortest path distance of each pair of matching points in the corresponding segment, the correlation value of each pair of corresponding segments is obtained. The weight distribution of salt spray concentration data and corrosion current density data in the forget gate formula is determined based on the correlation value, and the forget gate formula in the LSTM model is corrected to obtain the output result of the forget gate. The test data is predicted based on the output result of the forget gate.

2. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The preprocessing comprises: The test data were aligned according to the timestamps, and the missing parts in the test data were filled by linear interpolation to ensure the integrity of the data. The salt spray concentration data and corrosion current density data were normalized to unify their numerical ranges.

3. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The salt spray concentration data and the corrosion current density data in the test data are fitted with curves to generate two fitting curves, including: The least squares method was used to fit curves to the data series of salt spray concentration data and corrosion current density data, respectively, to obtain the salt spray concentration data-time curve and the corrosion current density data-time curve.

4. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The method of matching the two fitting curves by using a dynamic time warping algorithm includes: The fitting curve of the salt spray concentration data is detected by the first-order derivative method, all peak points on the curve are marked, and the peak points are used as segmentation points of the fitting curve, and segmentation is performed; The dynamic time warping algorithm is used to match the data points on the two curves, calculate the distance between all data points in the two curves, and construct a distance matrix. Starting from the upper left corner of the distance matrix, the minimum cumulative distance to each data point is gradually calculated along the shortest path to form a cumulative distance matrix. Starting from the lower right corner of the cumulative distance matrix, the path with the minimum cumulative distance is taken as the optimal path. According to the principle of the shortest distance, the corresponding matching points and distance values ​​of the data points on the two curves are obtained to complete the matching of the fitting curves of the salt spray concentration data and the corrosion current density data.

5. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The method of matching the two fitting curves by using a dynamic time warping algorithm includes: Any segmentation point on the fitting curve of the salt spray concentration data is taken as the target segmentation point. In response to the target segmentation point on the fitting curve of the salt spray concentration data corresponding to multiple matching points on the fitting curve of the corrosion current density data, the matching point corresponding to the maximum matching degree is selected as the segmentation point on the corrosion current density data-time curve.

6. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The degree of matching includes: The second-order derivative of each matching point on the fitting curve of the corrosion current density data is calculated, and the Euclidean distance between the segmentation point on the fitting curve of the salt spray concentration data and the matching point on the fitting curve of the corrosion current density data is calculated. The Euclidean distance is mapped using a negative exponential function, and the product of the mapped Euclidean distance and the second-order derivative is normalized as the degree of matching between the segmentation point and the matching point.

7. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The correlation value includes: Take any pair of segments as the target segment, calculate the shortest Euclidean distance between all segmentation points and matching points in the target segment, take the inverse of the shortest Euclidean distance, and normalize it to obtain the similarity weight; calculate the ratio between the standard deviation of the time difference of each matching point pair in the target segment and the mean of the time difference of each matching point pair in the target segment, and obtain the coefficient of variation of the time difference; calculate the average value between the matching degree of the left segmentation point and the matching point in the target segment and the matching degree of the right segmentation point and the matching point, and obtain the comprehensive matching degree; The product of the similarity weight, the coefficient of variation of the time difference and the comprehensive matching degree is taken as the relevance value of the target segment.

8. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The method of determining the weight distribution of the salt spray concentration data and the corrosion current density data in the forget gate formula based on the correlation value includes: The weighted average of the correlation values ​​of all segments is used as the weight of the salt spray concentration data in the forget gate , 1 minus the weight of the salt spray concentration data in the forget gate is used as the weight of the corrosion current density data in the forget gate formula ; The weight in the forget gate in response to the salt spray concentration data When it approaches 1, the LSTM model prediction focuses more on the impact of salt spray concentration data on corrosion current density data, responding to the weight of salt spray concentration data in the forget gate. When it approaches 0, the LSTM model prediction pays more attention to the historical trend of corrosion current density data.

9. The intelligent processing method for radiator salt spray test data according to claim 1 is characterized in that: The output result of the forget gate includes: The product of the multiplication of the salt spray concentration data of the forget gate at the current moment, the corresponding weight influence matrix and the weight of the salt spray concentration data in the forget gate is used as the contribution value of the salt spray concentration data at the current moment; the product of the multiplication of the corrosion current density data of the forget gate at the current moment, the corresponding weight influence matrix and the weight of the corrosion current density data in the forget gate is used as the contribution value of the corrosion current density data at the current moment; the product between the hidden state at the previous moment and the weight matrix of the hidden state is calculated to obtain the contribution value of the hidden state at the previous moment; the sum of the contribution value of the salt spray concentration data, the contribution value of the corrosion current density data, the contribution value of the hidden state at the previous moment and the bias term is added respectively, and used Activation function to obtain the output result of the corrected forget gate at the current moment.

10. A radiator salt spray test data intelligent processing system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent processing method for radiator salt spray test data according to any one of claims 1 to 9 is implemented.

Images