An intelligent analysis method for engineering plastics solutions based on deep learning

Abstract

The present invention relates to the field of data processing technology, and more particularly to a method for intelligently analyzing engineering plastic solutions based on deep learning. The method comprises the following steps: transmitting engineering plastic data collected from the engineering plastic solution to a location to be processed, dividing the optimized data into a training set, a validation set, and a test set; training an initial model using the training set to output a deep learning model; analyzing the engineering plastic solution using the deep learning model to output analysis results, and optimizing the engineering plastic solution based on the analysis results; determining the analysis stability of the engineering plastic solution based on the variance of the proportion of lost data in the engineering plastic data; adjusting the maximum number of concurrent transmission connections if the analysis stability does not meet requirements; and adjusting the learning rate of the deep learning model if the training effectiveness does not meet requirements. The present invention improves the analysis stability of the engineering plastic solution.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an intelligent analysis method for engineering plastic solutions based on deep learning. Background Art

[0002] The deep learning-based intelligent analysis method for engineering plastics solutions is an advanced system that uses deep learning technology to efficiently and accurately analyze and recommend solutions for the complexities of engineering plastics materials and their applications. By processing a large amount of material performance data, molding process parameters, application environmental conditions, and other information, this method can predict the performance of engineering plastic products, optimize formulation design, and guide production process adjustments, thereby improving product quality and development efficiency while reducing R&D costs.

[0003] Chinese Patent Publication No. CN109558896A discloses an intelligent disease analysis method and system based on ultrasound omics and deep learning. The method includes at least the following steps: acquiring multiple ultrasound data of a lesion site to obtain multimodal ultrasound omics data; inputting the multimodal ultrasound omics data into a trained deep learning neural network, and adjusting the connection weights, convolution ratios, and pooling layers of neurons based on the multimodal ultrasound omics data to obtain adjusted multimodal ultrasound omics data; using classifiers under different modalities to classify each data in the adjusted multimodal ultrasound omics data to obtain classification probabilities for all modalities including each classification; performing weighted averaging on the classification probabilities of all modalities based on the confusion scores between modalities given by the discriminator to obtain a score for each classification; and based on clinical outcome indicators and genomic data, using logistic regression to calculate high-risk indicators based on the scores of each classification, using decision trees or Adaboost to establish a classification model, and using t-tests and Pearson / Spearman correlation analyses to obtain prognosis results, efficacy evaluation results, and auxiliary diagnosis results. It can be seen from this that the intelligent disease analysis method and system based on ultrasound omics and deep learning have the problem that due to the overly strict firewall rules, some normal engineering plastic solution data transmission may be mistaken as external attacks, thereby interrupting the data transmission line, resulting in partial data loss and thus causing a decrease in the analysis stability of the engineering plastic solution. Summary of the Invention

[0004] To this end, the present invention provides an intelligent analysis method for engineering plastic solutions based on deep learning, which is used to overcome the problem in the prior art that due to the overly strict setting of firewall rules, some normal engineering plastic solution data transmission may be mistaken for external attacks, thereby interrupting the data transmission line, resulting in partial data loss and thus causing a decrease in the analysis stability of the engineering plastic solution.

[0005] To achieve the above-mentioned object, the present invention provides an intelligent analysis method for engineering plastic solutions based on deep learning, comprising: transmitting the engineering plastic data collected in the engineering plastic solution to a location to be processed, and sequentially performing cleaning, denoising, conversion and feature extraction operations to output optimized data, and dividing the optimized data into a training set, a validation set and a test set; using the training set to train an initial model to output a deep learning model, and using the validation set and the test set to perform validation and test operations on the deep learning model respectively; using the deep learning model to analyze the engineering plastic solution to output analysis results, and according to the analysis The engineering plastics scheme is optimized based on the results; the amount of lost data of engineering plastics data and the total amount of engineering plastics data in several acquisition cycles are obtained respectively; the analytical stability of the engineering plastics scheme is determined based on the variance of the proportion of lost data of engineering plastics data; if the analytical stability does not meet the requirements, the maximum number of concurrent transmission connections is adjusted, or the training effectiveness of the deep learning model is determined based on the reuse rate of the optimized data; if the training effectiveness does not meet the requirements, the learning rate of the deep learning model is adjusted, or the acquisition frequency of the engineering plastics data is adjusted based on the byte difference between the acquisition value of the engineering plastics data and the actual value.

[0006] Furthermore, determining the analytical stability of the engineering plastic solution includes:

[0007] Compare the variance of the missing data ratio of engineering plastics data with the preset first variance;

[0008] If the variance of the proportion of missing data of the engineering plastic data is greater than the preset first variance, it is determined that the analysis stability of the engineering plastic solution does not meet the requirements.

[0009] Furthermore, the effectiveness of the training of the deep learning model is confirmed, including:

[0010] Comparing the variance of the missing data ratio of the engineering plastic data with the preset first variance and the preset second variance respectively;

[0011] If the variance of the proportion of lost data in the engineering plastic data is greater than the preset first variance and less than or equal to the preset second variance, it is preliminarily determined that the training effectiveness of the deep learning model does not meet the requirements, and whether the training effectiveness of the deep learning model meets the requirements is determined based on the reuse rate of the optimized data.

[0012] Furthermore, adjusting the maximum number of concurrent transmission connections includes:

[0013] Comparing the variance of the missing data ratio of the engineering plastic data with the preset second variance;

[0014] If the variance of the proportion of lost data of the engineering plastic data is greater than a preset second variance, the maximum number of concurrent transmission connections is increased.

[0015] Furthermore, the increase range of the maximum number of concurrent transmission connections is determined by the difference between the variance of the proportion of lost data of the engineering plastics data and a preset second variance.

[0016] Furthermore, adjusting the learning rate of the deep learning model includes:

[0017] Comparing the reuse rate of the optimized data with a preset first reuse rate and a preset second reuse rate respectively;

[0018] If the reuse rate of the optimized data is greater than the preset first reuse rate, it is determined that the training effectiveness of the deep learning model does not meet the requirements;

[0019] If the reuse rate of the optimized data is greater than the preset first reuse rate and less than or equal to the preset second reuse rate, reducing the learning rate of the deep learning model;

[0020] If the reuse rate of the optimized data is greater than the preset second reuse rate, it is preliminarily determined that the quality of the engineering plastic data does not meet the requirements, and whether the quality of the engineering plastic data meets the requirements is determined based on the byte difference between the collected value and the actual value of the engineering plastic data.

[0021] Furthermore, the reduction extent of the learning rate of the deep learning model is determined by optimizing the difference between the reuse rate of the data and a preset first reuse rate.

[0022] Furthermore, the frequency of collecting the engineering plastics data is adjusted, including:

[0023] Comparing the byte difference between the engineering plastic data collection value and the actual value with a preset difference;

[0024] If the byte difference between the collected value and the actual value of the engineering plastic data is greater than the preset difference, it is determined that the quality of the engineering plastic data does not meet the requirements, and the collection frequency of the engineering plastic data is reduced.

[0025] Furthermore, the byte difference between the collected value and the actual value of the engineering plastic data is the difference between the byte amount of the collected engineering plastic data and the byte amount of the actual engineering plastic data.

[0026] Furthermore, the reduction range of the frequency of collecting the engineering plastics data is determined by the difference between the byte difference between the collected value and the actual value of the engineering plastics data and a preset difference.

[0027] Compared with the prior art, the beneficial effect of the present invention is that the method of the present invention adjusts the maximum number of concurrent transmission connections according to the variance of the proportion of lost data of engineering plastics data. Since the firewall rules are set too strictly, some normal engineering plastics solution data transmission may be mistaken for external attacks, thereby interrupting the data transmission line and causing some data loss. By increasing the maximum number of concurrent transmission connections, more of these concurrent data transmission tasks can be allowed to proceed smoothly, more normal data transmission connections can be accepted, and data loss caused by connection number restrictions can be reduced. The learning rate of the deep learning model is adjusted according to the reuse rate of the optimized data. Since it is difficult to obtain plastic sample data in some rare engineering plastics and special scenarios, it leads to errors in model training. The amount of data is small, and the number of training rounds is large during training, resulting in training mismatch and thus over-training of the model. By reducing the learning rate of the deep learning model, the update of the model parameters can be smoother, so that the model can more carefully find the optimal solution on limited data. The collection frequency of engineering plastics data is adjusted according to the byte difference between the collected value and the actual value of the engineering plastics data. Due to the long use time of the collection equipment, it ages, resulting in a decline in performance when collecting data, and deviations in the collected data, which leads to a decline in the quality of engineering plastics data. By reducing the collection frequency of engineering plastics data, the amount of unstable data can be reduced, so that subsequent data processing and analysis can more effectively utilize the relatively stable data part, thereby improving the analytical stability of the engineering plastics solution.

[0028] Furthermore, the method of the present invention adjusts the maximum number of concurrent transmission connections by setting a preset first variance and a preset second variance. Since the firewall rules are set too strictly, some normal engineering plastic solution data transmissions may be mistaken for external attacks, thereby interrupting the data transmission line and causing some data loss. By increasing the maximum number of concurrent transmission connections, more of these concurrent data transmission tasks can be allowed to proceed smoothly, more normal data transmission connections can be accepted, and data loss caused by connection number restrictions can be reduced, thereby further improving the analysis stability of the engineering plastic solution.

[0029] Furthermore, the method of the present invention adjusts the learning rate of the deep learning model by setting a preset first reuse rate and a preset second reuse rate. Since it is difficult to obtain plastic sample data for certain rare engineering plastics and special scenarios, the amount of data used for model training is small and the number of training rounds is large, resulting in training mismatch and thus excessive model training. By reducing the learning rate of the deep learning model, the update of the model parameters can be smoother, allowing the model to more carefully find the optimal solution based on limited data, further improving the analytical stability of the engineering plastic solution.

[0030] Furthermore, the method of the present invention adjusts the frequency of collecting engineering plastics data by setting a preset difference amount. Since the collection equipment is used for a long time and ages, its performance in collecting data decreases, and the collected data deviates, thereby causing the quality of the engineering plastics data to decrease. By reducing the frequency of collecting engineering plastics data, the amount of unstable data can be reduced, so that subsequent data processing and analysis can more effectively utilize the relatively stable data portion, further improving the analytical stability of the engineering plastics solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is an overall flow chart of the intelligent analysis method for engineering plastic solutions based on deep learning according to an embodiment of the present invention;

[0032] Figure 2 This is a logic flow chart of an intelligent analysis method for engineering plastic solutions based on deep learning according to an embodiment of the present invention;

[0033] Figure 3 This is a specific flow chart of the process of adjusting the maximum number of concurrent transmission connections in the intelligent analysis method for engineering plastic solutions based on deep learning in an embodiment of the present invention;

[0034] Figure 4 This is a specific flow chart of the process of adjusting the learning rate of a deep learning model in the intelligent analysis method for engineering plastic solutions based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0036] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0037] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0038] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0039] See also Figure 1 、 Figure 2 、 Figure 3 as well as Figure 4 As shown in the figure, they are respectively an overall flow chart, a logic flow chart, a specific flow chart of the process of adjusting the maximum number of concurrent transmission connections, and a specific flow chart of the process of adjusting the learning rate of the deep learning model of an intelligent analysis method for engineering plastic solutions based on deep learning according to an embodiment of the present invention. The intelligent analysis method for engineering plastic solutions based on deep learning according to the present invention includes:

[0040] Step S1, transmitting the engineering plastic data collected in the engineering plastic solution to a location to be processed, and performing cleaning, denoising, conversion, and feature extraction operations in sequence to output optimized data, and dividing the optimized data into a training set, a validation set, and a test set;

[0041] Step S2: using the training set to train the initial model to output a deep learning model, and using the validation set and the test set to perform validation and testing operations on the deep learning model respectively;

[0042] Step S3, using the deep learning model to analyze the engineering plastic solution to output an analysis result, and optimizing the engineering plastic solution according to the analysis result;

[0043] Step S4, respectively obtaining the amount of lost data of engineering plastics data and the total amount of engineering plastics data in several acquisition cycles;

[0044] Step S5, determining the analytical stability of the engineering plastics solution based on the variance of the missing data ratio of the engineering plastics data;

[0045] Step S6: If the analysis stability does not meet the requirements, the maximum number of concurrent transmission connections is adjusted, or the training effectiveness of the deep learning model is determined based on the multiplexing rate of the optimized data;

[0046] Step S7: If the training effectiveness does not meet the requirements, the learning rate of the deep learning model is adjusted, or the collection frequency of the engineering plastics data is adjusted based on the byte difference between the collected value of the engineering plastics data and the actual value.

[0047] Specifically, the engineering plastics solution includes determining the performance selection of engineering plastics, determining the cost of engineering plastics, and determining the processing technology of engineering plastics.

[0048] Specifically, engineering plastics data include processing temperature of engineering plastics, extrusion rate of engineering plastics, and impurity content of engineering plastics.

[0049] Specifically, the optimized data includes complete engineering plastic tensile strength data, engineering plastic injection pressure data after unified unit system, and engineering plastic processing temperature data with erroneous values ​​removed.

[0050] Specifically, deep learning models include convolutional neural networks, recurrent neural networks, and deep neural networks.

[0051] Specifically, the analysis results include the prediction of heat deformation temperature of engineering plastics, electrical conductivity of engineering plastics, and injection molding temperature of engineering plastics.

[0052] Specifically, the maximum number of concurrent transmission connections is the maximum number of engineering plastic data transmission connections that the system can support at the same time.

[0053] Specifically, the reuse rate of optimized data is the ratio of the amount of optimized data reused to the total data usage.

[0054] Specifically, the learning rate of a deep learning model is a hyperparameter used to control the step size of adjusting model parameters according to the gradient of the loss function during the training process of the deep learning model.

[0055] In implementation, the method of the present invention adjusts the maximum number of concurrent transmission connections according to the variance of the proportion of lost data of engineering plastics data. Due to the overly strict setting of firewall rules, some normal engineering plastic solution data transmission may be mistaken for external attacks, thereby interrupting the data transmission line and causing some data loss. By increasing the maximum number of concurrent transmission connections, more of these concurrent data transmission tasks can be allowed to proceed smoothly, more normal data transmission connections can be accepted, and data loss caused by connection number restrictions can be reduced. The learning rate of the deep learning model is adjusted according to the reuse rate of the optimized data. Since it is difficult to obtain plastic sample data in some rare engineering plastics and special scenarios, the amount of data is small when training the model. A large number of training rounds during training leads to training mismatch, which in turn leads to over-training of the model. By reducing the learning rate of the deep learning model, the update of the model parameters can be smoother, allowing the model to more carefully find the optimal solution on limited data. The collection frequency of engineering plastics data is adjusted according to the byte difference between the collected value and the actual value of the engineering plastics data. Due to the long use time of the collection equipment, it ages, resulting in a decrease in performance when collecting data, and deviations in the collected data, which leads to a decrease in the quality of the engineering plastics data. By reducing the collection frequency of engineering plastics data, the amount of unstable data can be reduced, so that subsequent data processing and analysis can more effectively utilize the relatively stable data part, thereby improving the analytical stability of the engineering plastics solution.

[0056] Specifically, the analytical stability of the engineering plastics solution was determined, including:

[0057] Obtain the amount of missing data and the total amount of engineering plastics data in several collection cycles, and calculate the variance of the proportion of missing data of engineering plastics data;

[0058] Comparing the variance of the missing data ratio of the engineering plastic data with a preset first variance;

[0059] If the variance of the proportion of missing data of the engineering plastic data is greater than the preset first variance, it is determined that the analysis stability of the engineering plastic solution does not meet the requirements.

[0060] Specifically, the effectiveness of the training of the deep learning model is verified, including:

[0061] Comparing the variance of the missing data ratio of the engineering plastic data with the preset first variance and the preset second variance respectively;

[0062] If the variance of the proportion of lost data in the engineering plastic data is greater than the preset first variance and less than or equal to the preset second variance, it is preliminarily determined that the training effectiveness of the deep learning model does not meet the requirements, and whether the training effectiveness of the deep learning model meets the requirements is determined based on the reuse rate of the optimized data.

[0063] It can be understood that the three intervals corresponding to the preset first variance and the preset second variance correspond to three situations respectively:

[0064] The first interval is when the variance of the missing data percentage of the engineering plastics data is less than or equal to the preset first variance, indicating that the analytical stability of the engineering plastics solution meets the requirements.

[0065] The second interval is when the variance of the missing data ratio of engineering plastics data is greater than the preset first variance and less than or equal to the preset second variance. This corresponds to the difficulty in obtaining sample data for some rare engineering plastics and plastics in special scenarios. This results in a small amount of data for model training and a large number of training rounds, resulting in training mismatch and thus overtraining of the model:

[0066] The third interval is when the variance of the proportion of lost data for engineering plastics data is greater than the preset second variance. This indicates that the firewall rules are set too strictly, which may mistake some normal engineering plastic solution data transmission as external attacks, thereby interrupting the data transmission line and causing partial data loss.

[0067] In practice, the preset first variance is generally selected in the range of [0.02, 0.04], and the preset second variance is generally selected in the range of [0.05, 0.07].

[0068] Preferably, the preferred embodiment of the preset first variance is 0.03, and the preferred embodiment of the preset second variance is 0.06.

[0069] Specifically, the variance of the proportion of lost data of engineering plastics data is the variance of the proportion of lost data of engineering plastics data in several acquisition cycles. The calculation method of the variance of the proportion of lost data of engineering plastics data is a conventional technical means well known to those skilled in the art. Therefore, the calculation process of the variance of the proportion of lost data of engineering plastics data will not be repeated here.

[0070] Specifically, the proportion of lost data of engineering plastics data is the ratio of the lost data of engineering plastics data in several collection cycles to the total data volume of engineering plastics data.

[0071] In implementation, the method of the present invention determines the analytical stability of the engineering plastic solution by setting a preset first variance and a preset second variance, thereby reducing the impact of the decreased analytical accuracy of the engineering plastic solution due to inaccurate determination of the analytical stability of the engineering plastic solution, and further improving the analytical stability of the engineering plastic solution.

[0072] Specifically, adjusting the maximum number of concurrent transmission connections includes:

[0073] Comparing the variance of the missing data ratio of the engineering plastic data with the preset second variance;

[0074] If the variance of the proportion of lost data of the engineering plastic data is greater than a preset second variance, the maximum number of concurrent transmission connections is increased.

[0075] Specifically, the increase range of the maximum number of concurrent transmission connections is determined by the difference between the variance of the proportion of lost data of engineering plastics data and a preset second variance.

[0076] Specifically, when the difference between the variance of the proportion of lost data of engineering plastics data and the preset second variance is within 0.02, the maximum number of concurrent transmission connections increases to 1.2 times the original number; when the difference between the variance of the proportion of lost data of engineering plastics data and the preset second variance exceeds 0.02, the maximum number of concurrent transmission connections increases by 60 for every 0.01 that exceeds. For example, the difference between the variance of the proportion of lost data of engineering plastics data and the preset second variance is 0.04. The current maximum number of concurrent transmission connections is 500, and the increased maximum number of concurrent transmission connections is 500×1.2+60×2=720.

[0077] During implementation, the method of the present invention adjusts the maximum number of concurrent transmission connections by setting a preset first variance and a preset second variance. Since the firewall rules are set too strictly, some normal engineering plastic solution data transmissions may be mistaken for external attacks, thereby interrupting the data transmission line and causing some data loss. By increasing the maximum number of concurrent transmission connections, more of these concurrent data transmission tasks can be allowed to proceed smoothly, more normal data transmission connections can be accepted, and data loss caused by connection number restrictions can be reduced, further improving the analysis stability of the engineering plastic solution.

[0078] Specifically, adjusting the learning rate of the deep learning model includes:

[0079] Obtaining the amount of data reused for the optimized data and the total data usage respectively, and calculating the reuse rate of the optimized data;

[0080] Comparing the reuse rate of the optimized data with a preset first reuse rate and a preset second reuse rate respectively;

[0081] If the reuse rate of the optimized data is greater than the preset first reuse rate, it is determined that the training effectiveness of the deep learning model does not meet the requirements;

[0082] If the reuse rate of the optimized data is greater than the preset first reuse rate and less than or equal to the preset second reuse rate, reducing the learning rate of the deep learning model;

[0083] If the reuse rate of the optimized data is greater than the preset second reuse rate, it is preliminarily determined that the quality of the engineering plastic data does not meet the requirements, and whether the quality of the engineering plastic data meets the requirements is determined based on the byte difference between the collected value and the actual value of the engineering plastic data.

[0084] It can be understood that the three intervals corresponding to the preset first multiplexing rate and the preset second multiplexing rate correspond to three situations respectively:

[0085] The first interval is when the reuse rate of optimized data is less than or equal to the preset first reuse rate, corresponding to the situation where the training effectiveness of the deep learning model meets the requirements;

[0086] The second interval is when the reuse rate of optimized data is greater than the preset first reuse rate and less than or equal to the preset second reuse rate. This corresponds to the difficulty in obtaining sample data for certain rare engineering plastics and plastics in special scenarios. This results in a small amount of data for model training and a large number of training rounds, leading to training mismatch and thus overtraining of the model.

[0087] The third interval is when the reuse rate of the optimized data is greater than the preset second reuse rate. This is because the collection equipment has been used for a long time and has aged, resulting in performance degradation when collecting data and deviations in the collected data.

[0088] In practice, the preset first multiplexing rate is generally selected from the range of [43%, 47%], and the preset second multiplexing rate is generally selected from the range of [48%, 52%].

[0089] Preferably, the preferred embodiment of the preset first multiplexing rate is 45%, and the preferred embodiment of the preset second multiplexing rate is 50%.

[0090] In implementation, the method of the present invention determines the training effectiveness of the deep learning model by setting a preset first reuse rate and a preset second reuse rate, thereby reducing the impact of the decline in the analytical stability of the engineering plastic solution due to inaccurate determination of the training effectiveness of the deep learning model, and further improving the analytical stability of the engineering plastic solution.

[0091] Specifically, the reduction range of the learning rate of the deep learning model is determined by optimizing the difference between the reuse rate of the data and the preset first reuse rate.

[0092] Specifically, when the difference between the reuse rate of the optimized data and the preset first reuse rate is within 5%, the learning rate of the deep learning model is reduced to 0.95 times the original; when the difference between the reuse rate of the optimized data and the preset first reuse rate exceeds 5%, the learning rate of the deep learning model is reduced by 0.002 for every 3% exceeding it. For example, the difference between the reuse rate of the optimized data and the preset first reuse rate is 11%, the current learning rate of the deep learning model is 0.04, and the reduced learning rate of the deep learning model is 0.04×0.95-0.002×2=0.034.

[0093] In implementation, the method of the present invention adjusts the learning rate of the deep learning model by setting a preset first reuse rate and a preset second reuse rate. Since it is difficult to obtain plastic sample data for some rare engineering plastics and special scenarios, the amount of data used in training the model is small and the number of training rounds is large, resulting in training mismatch and thus excessive model training. By reducing the learning rate of the deep learning model, the update of the model parameters can be smoother, allowing the model to more carefully find the optimal solution based on limited data, further improving the analytical stability of the engineering plastic solution.

[0094] Specifically, adjusting the frequency of collecting the engineering plastics data includes:

[0095] respectively obtaining the byte amount of the collected engineering plastic data and the byte amount of the actual engineering plastic data, and calculating the byte difference between the collected value and the actual value of the engineering plastic data;

[0096] Comparing the byte difference between the engineering plastic data collection value and the actual value with a preset difference;

[0097] If the byte difference between the collected value and the actual value of the engineering plastic data is greater than the preset difference, it is determined that the quality of the engineering plastic data does not meet the requirements, and the collection frequency of the engineering plastic data is reduced.

[0098] It can be understood that the two intervals corresponding to the preset difference amount correspond to two situations:

[0099] The first interval is when the byte difference between the collected value and the actual value of the engineering plastics data is less than or equal to the preset difference, which corresponds to the situation where the quality of the engineering plastics data meets the requirements;

[0100] The second interval is when the byte difference between the collected value and the actual value of engineering plastic data is greater than the preset difference. This indicates that the collection device has been in use for a long time and has aged, resulting in performance degradation during data collection and deviation in the collected data.

[0101] In practice, the preset difference amount is generally selected in the range of [700Byte, 800Byte].

[0102] Preferably, the preset difference amount is 750 Byte.

[0103] In implementation, the method of the present invention determines the quality of engineering plastic data by setting a preset difference amount, thereby reducing the impact of the decline in analytical stability of the engineering plastic solution due to inaccurate determination of the quality of the engineering plastic data, and further improving the analytical stability of the engineering plastic solution.

[0104] Specifically, the byte difference between the collected value and the actual value of the engineering plastic data is the difference between the byte amount of the collected engineering plastic data and the byte amount of the actual engineering plastic data.

[0105] Specifically, the reduction range of the frequency of collecting the engineering plastics data is determined by the difference between the byte difference between the collected value and the actual value of the engineering plastics data and a preset difference.

[0106] Specifically, when the difference between the byte difference between the collected value of engineering plastics data and the actual value and the preset difference is within 100 bytes, the collection frequency of engineering plastics data is reduced to 0.9 times the original; when the difference between the byte difference between the collected value of engineering plastics data and the actual value and the preset difference exceeds 100 bytes, the collection frequency of engineering plastics data is reduced by 1 time / minute for every 50 bytes exceeding. For example, the difference between the byte difference between the collected value of engineering plastics data and the actual value and the preset difference is 200 bytes. The current collection frequency of engineering plastics data is 10 times / minute, and the reduced collection frequency of engineering plastics data is 10×0.9-1×2=7 times / minute.

[0107] During implementation, the method of the present invention adjusts the frequency of collecting engineering plastics data by setting a preset difference amount. Since the collection equipment has been used for a long time and has aged, its performance in collecting data has declined, and the collected data has deviated, thereby causing the quality of the engineering plastics data to decline. By reducing the frequency of collecting engineering plastics data, the amount of unstable data can be reduced, so that subsequent data processing and analysis can more effectively utilize the relatively stable data portion, further improving the analytical stability of the engineering plastics solution.

[0108] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. An intelligent analysis method for engineering plastic solutions based on deep learning, characterized in that: include: The engineering plastic data collected in the engineering plastic solution is transmitted to the location to be processed, and cleaning, denoising, conversion and feature extraction operations are performed in sequence to output optimized data, and the optimized data is divided into a training set, a validation set and a test set; Using the training set to train the initial model to output a deep learning model, and using the validation set and the test set to perform validation and testing operations on the deep learning model respectively; Analyzing the engineering plastic solution using the deep learning model to output analysis results, and optimizing the engineering plastic solution based on the analysis results; Obtain the amount of lost data and the total amount of engineering plastics data in several acquisition cycles respectively; Determine the analytical stability of the engineering plastics solution based on the variance of the missing data percentage of the engineering plastics data; If the analysis stability does not meet the requirements, the maximum number of concurrent transmission connections is adjusted, or the training effectiveness of the deep learning model is determined based on the multiplexing rate of the optimized data; If the training effectiveness does not meet the requirements, the learning rate of the deep learning model is adjusted, or the frequency of collecting engineering plastics data is adjusted based on the byte difference between the collected value of the engineering plastics data and the actual value; Among them, engineering plastics data include processing temperature of engineering plastics, extrusion rate of engineering plastics, and impurity content of engineering plastics; The analysis results include the prediction of heat distortion temperature of engineering plastics, electrical conductivity of engineering plastics, and injection molding temperature of engineering plastics; Verify the training effectiveness of the deep learning model, including: Comparing the variance of the missing data ratio of the engineering plastic data with a preset first variance and a preset second variance respectively; If the variance of the proportion of missing data of the engineering plastic data is greater than the preset first variance and less than or equal to the preset second variance, it is preliminarily determined that the training effectiveness of the deep learning model does not meet the requirements, and whether the training effectiveness of the deep learning model meets the requirements is determined based on the reuse rate of the optimized data; Adjusting the maximum number of concurrent transmission connections includes: Comparing the variance of the missing data ratio of the engineering plastic data with the preset second variance; If the variance of the proportion of lost data of the engineering plastics data is greater than a preset second variance, increasing the maximum number of concurrent transmission connections; Adjusting the learning rate of the deep learning model includes: Comparing the reuse rate of the optimized data with a preset first reuse rate and a preset second reuse rate respectively; If the reuse rate of the optimized data is greater than the preset first reuse rate, it is determined that the training effectiveness of the deep learning model does not meet the requirements; If the reuse rate of the optimized data is greater than the preset first reuse rate and less than or equal to the preset second reuse rate, reducing the learning rate of the deep learning model; If the reuse rate of the optimized data is greater than the preset second reuse rate, it is preliminarily determined that the quality of the engineering plastics data does not meet the requirements, and whether the quality of the engineering plastics data meets the requirements is determined based on the byte difference between the collected value and the actual value of the engineering plastics data; Adjusting the frequency of collecting the engineering plastics data includes: Comparing the byte difference between the engineering plastic data collection value and the actual value with a preset difference; If the byte difference between the collected value and the actual value of the engineering plastic data is greater than the preset difference, it is determined that the quality of the engineering plastic data does not meet the requirements, and the collection frequency of the engineering plastic data is reduced.

2. The method for intelligent analysis of engineering plastic solutions based on deep learning according to claim 1, characterized in that: Determine the analytical robustness of the engineering plastics solution, including: Compare the variance of the missing data ratio of engineering plastics data with the preset first variance; If the variance of the proportion of missing data of the engineering plastic data is greater than the preset first variance, it is determined that the analysis stability of the engineering plastic solution does not meet the requirements.

3. The method for intelligent analysis of engineering plastic solutions based on deep learning according to claim 2, characterized in that: The increase in the maximum number of concurrent transmission connections is determined by the difference between the variance of the proportion of lost data of engineering plastics data and a preset second variance; The greater the difference between the variance of the ratio of lost data volume of the engineering plastics data and the preset second variance, the greater the increase in the maximum number of concurrent transmission connections.

4. The method for intelligent analysis of engineering plastic solutions based on deep learning according to claim 3, characterized in that: The reduction range of the learning rate of the deep learning model is determined by optimizing the difference between the reuse rate of the data and the preset first reuse rate; Among them, the greater the difference between the reuse rate of the optimized data and the preset first reuse rate, the greater the reduction in the learning rate of the deep learning model.

5. The method for intelligent analysis of engineering plastic solutions based on deep learning according to claim 4, characterized in that: The byte difference between the collected engineering plastic data value and the actual value is the difference between the byte amount of the collected engineering plastic data and the byte amount of the actual engineering plastic data.

6. The method for intelligent analysis of engineering plastic solutions based on deep learning according to claim 5, characterized in that: The reduction range of the acquisition frequency of the engineering plastics data is determined by the difference between the byte difference between the acquired value and the actual value of the engineering plastics data and the preset difference; The greater the difference between the byte difference between the engineering plastic data collection value and the actual value and the preset difference, the greater the reduction in the frequency of engineering plastic data collection.

Images