An efficient testing method and device for the amount of carbon fiber hairiness

Abstract

The present invention relates to an efficient testing method and device for the amount of carbon fiber fly, and relates to the field of carbon fiber fly detection, including: tracing back the carbon fiber preparation data of the test sample, extracting the raw material defect, the time series values of process mismatch parameters and the time series values of equipment misalignment parameters, performing several crack simulations and hairiness simulations to obtain the crack distribution scale and the hairiness distribution scale; generating the fly amount simulation according to the crack distribution scale and the hairiness distribution scale to obtain the predicted fly amount; obtaining the fly amount confidence interval, when it belongs to the fly amount confidence interval, receiving the test fly amount, and if the deviation from the predicted fly amount is less than the deviation threshold, performing a credible identification and feeding it back to the indirect fly amount test end. It solves the technical problems of high direct measurement cost in the prior art, and poor stability and low detection efficiency in indirect measurement due to the need for multiple measurements and the inability to completely avoid accidental risks.

Description

Technical Field

[0001] The present invention relates to the field of carbon fiber hairiness detection, and particularly to an efficient testing method and device for the hairiness amount of carbon fiber. Background Art

[0002] There are two traditional methods for testing the hairiness amount of carbon fiber: direct measurement method and indirect measurement method. The direct measurement method, for example, observes the fiber surface through a high-resolution microscope or an electron microscope, and combines image processing software to count the number and distribution of hairiness. The indirect measurement method, for example, monitors the fracture points during fiber stretching, indirectly evaluates the density of hairiness defects through the fracture frequency and strength, and uses laser to irradiate the fiber surface, analyzes the length and density of hairiness through the scattered light signal, and there is also a standard detection process, such as calculating the remaining amount of hairiness through fiber combing and weighing under a specific tension.

[0003] The direct measurement method has a relatively high accuracy, but the disadvantage is that the implementation cost is relatively high. The indirect measurement method has a low implementation cost, but it depends on parameters with a relatively high correlation with the hairiness amount for testing. The disadvantage is that when the correlation parameters fluctuate, it will lead to a large error in the calculation result of the hairiness amount. Therefore, multiple tests are usually required. However, even so, there is still a risk that multiple retrieved data are accidental values, which results in poor stability and low efficiency of the detection results. Summary of the Invention

[0004] Aiming at the technical problems in the prior art that due to the high cost of direct measurement, and the indirect measurement requires multiple measurements and cannot completely avoid accidental risks, resulting in poor stability and low detection efficiency, the present invention provides an efficient testing method and device for the hairiness amount of carbon fiber to solve this problem.

[0005] The technical solution of the present invention to solve the above technical problems is as follows:

[0006] In the first aspect, the present invention provides an efficient testing method for the hairiness amount of carbon fiber, including:

[0007] Trace back the carbon fiber preparation data of the test sample, and extract the raw material defects, the time series values of the process mismatch parameters, and the time series values of the equipment misalignment parameters;

[0008] Perform several crack simulations and hairiness simulations according to the raw material defects, the time series values of the process mismatch parameters, and the time series values of the equipment misalignment parameters, and obtain the crack distribution scale and the hairiness distribution scale;

[0009] Generate a hairiness amount simulation according to the crack distribution scale and the hairiness distribution scale, and obtain the predicted hairiness amount;

[0010] Retrieve the same-type and same-position production carbon fiber in the preset time zone that meets the crack distribution scale and the hairiness distribution scale, perform confidence analysis, and obtain the confidence interval of the hair fiber amount, where the preset time zone is the time zone obtained by pushing forward the preset duration from the current moment;

[0011] When the predicted hair fiber amount belongs to the confidence interval of the hair fiber amount, receive the test hair fiber amount from the indirect test end of the hair fiber amount. If the deviation from the predicted hair fiber amount is less than the deviation threshold, perform a credible identification and feedback it to the indirect test end of the hair fiber amount.

[0012] In a second aspect, the present invention provides an efficient test device for the hair fiber amount of carbon fiber, including:

[0013] A preparation data module for retrieving the carbon fiber preparation data of the test sample and extracting the raw material defects, the time series values of the process mismatch parameters, and the time series values of the equipment misalignment parameters;

[0014] A crack and hairiness generation simulation module for performing several crack simulation generations and hairiness simulation generations according to the raw material defects, the time series values of the process mismatch parameters, and the time series values of the equipment misalignment parameters, and obtaining the crack distribution scale and the hairiness distribution scale;

[0015] A hair fiber amount simulation generation module for simulating and generating the hair fiber amount according to the crack distribution scale and the hairiness distribution scale, and obtaining the predicted hair fiber amount;

[0016] A confidence interval configuration module for retrieving the same-type and same-position production carbon fiber in the preset time zone that meets the crack distribution scale and the hairiness distribution scale, perform confidence analysis, and obtain the confidence interval of the hair fiber amount, where the preset time zone is the time zone obtained by pushing forward the preset duration from the current moment;

[0017] A credible identification module for, when the predicted hair fiber amount belongs to the confidence interval of the hair fiber amount, receiving the test hair fiber amount from the indirect test end of the hair fiber amount. If the deviation from the predicted hair fiber amount is less than the deviation threshold, perform a credible identification and feedback it to the indirect test end of the hair fiber amount.

[0018] The beneficial effects of the present invention are as follows: By retrospectively testing the preparation data of samples, raw material defects, process mismatch parameter time series values, and equipment misalignment parameter time series values related to the amount of hairiness are determined. Then, based on the stage characteristics of the generation of hairiness, the simulation of the generation of cracks and hairiness in the first stage is carried out first, and then the simulation of the amount of hairiness in the second stage is carried out to obtain the predicted amount of hairiness. Furthermore, a confidence interval is constructed using recent detection data to perform a primary verification on the predicted amount of hairiness. When it meets the confidence interval, the predicted amount of hairiness is used to verify the test result. If the verification passes, a credible label is generated, indicating that the test result can be directly used without the need for multiple tests. Moreover, since the predicted amount of hairiness conforms to the characteristic values in the long term and short term, the probability of accidental risks in the test hairiness of the credible label is reduced, thereby achieving the technical effects of improving efficiency and taking into account the implementation cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic flow chart of an efficient test method for the amount of hairiness of carbon fiber provided by the present invention;

[0020] Figure 2 It is a schematic structural diagram of an efficient test device for the amount of hairiness of carbon fiber provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] 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 only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0022] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0023] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0024] Example 1, as Figure 1 shown, an embodiment of the present invention provides an efficient test method for the amount of carbon fiber hairiness, including the steps of:

[0025] S10: Trace back the carbon fiber preparation data of the test sample, and extract the raw material defects, the time series values of process mismatch parameters, and the time series values of equipment misalignment parameters;

[0026] Specifically, the test sample is a sample cut from a carbon fiber product for which the amount of hairiness needs to be tested; the carbon fiber preparation data refers to the monitoring data during the preparation process of the test sample, including the raw material detection data before production, and the raw material defect data can be extracted. Specifically, the parameters with a relatively strong correlation with the amount of hairiness in the raw material defect data are impurities, bubbles, and the ratio of the shrinkage rate of the skin layer to the core layer.

[0027] In the embodiments of the present application, the impurity scale is used to characterize impurities, and the impurity scale is quantitatively calculated by the impurity weight per unit weight of the raw material. In the embodiments of the present application, the bubble scale is used to characterize bubbles; the bubble scale is quantitatively calculated by the number of bubbles per unit area; the ratio of the shrinkage rate of the skin layer to the core layer is the ratio of the shrinkage rate of the skin layer to the core layer, which is an existing concept and will not be elaborated in detail.

[0028] Furthermore, the carbon fiber preparation data further includes the time series values of process mismatch parameters. The time series values of process mismatch parameters are the time series record values of the process parameters deviated from the user's preset process parameters in the production process. Specifically, in the time series values of process mismatch parameters, the embodiments of the present application select the time series values of draw ratio mismatch and heat setting temperature mismatch with relatively strong correlation with the amount of hairiness for calculation.

[0029] Furthermore, the carbon fiber preparation data further includes the time series values of equipment misalignment parameters. The time series values of equipment misalignment parameters are the time series record values of the equipment control deviation. Specifically, in the time series values of equipment misalignment parameters, the embodiments of the present application select the time series values of the gas flow rate distribution mismatch in the carbonization furnace and the take-up tension mismatch with relatively strong correlation with the amount of hairiness for calculation.

[0030] For the remaining parameters, either they are related to the aforementioned selected parameters or have a weak correlation with the amount of hairiness, and the embodiments of the present application do not select them to avoid increasing redundant computational effort.

[0031] Using the parameters of the test sample, the generation state of its hairiness amount can be simulated, which can be used as the final verification data. Compared with the traditional manual experience evaluation, it has higher stability and objectivity.

[0032] Furthermore, in order to ensure the accurate prediction of the hairiness amount, the embodiments of the present application build a multi-stage simulator in combination with historical data. In traditional data-driven models, causal data of collection mapping is usually acquired and then the corresponding simulator is directly built. However, in the actual production process, there are often multiple stages of transitions between causal data. Especially in the production process of carbon fiber, the hairiness amount is gradually generated. When training the simulator, if the multi-stage process of hairiness amount generation is not considered, the obtained results will surely be inaccurate. Therefore, the embodiments of the present application are based on the stage characteristics of the hairiness amount and generate the following generation path:

[0033] The first stage:

[0034] A [Raw material defect] -> B (Amplification of spinning stage defect);

[0035] C [Process parameter mismatch] -> D (Pre-oxidation / carbonization structure defect);

[0036] E [Equipment wear / operation error] -> F (Accumulation of mechanical damage);

[0037] The second stage:

[0038] B -> G [Surface crack / hairiness];

[0039] D -> G;

[0040] F -> G;

[0041] The third stage:

[0042] G -> H [Excessive hairiness amount].

[0043] It can be seen that the generation of the hairiness amount is in three stages, and the best simulator should be in three stages. However, since the characteristic data of the stages of A [Raw material defect] -> B (Amplification of spinning stage defect), C [Process parameter mismatch] -> D (Pre-oxidation / carbonization structure defect), and E [Equipment wear / operation error] -> F (Accumulation of mechanical damage) are difficult to collect and difficult to quantitatively characterize, the embodiments of the present application temporarily build a two-stage simulator, that is:

[0044] The first stage:

[0045] A [Raw material defect] -> G [Surface crack / hairiness];

[0046] C[Process parameter mismatch] --> G[Surface crack / fuzz];

[0047] E[Equipment wear / operation error] --> G[Surface crack / fuzz];

[0048] The second stage:

[0049] G --> H[Excessive amount of hairiness]。

[0050] It can be seen that in the first stage, simulators for A[Raw material defect] --> G[Surface crack / fuzz], C[Process parameter mismatch] --> G[Surface crack / fuzz], and E[Equipment wear / operation error] --> G[Surface crack / fuzz] need to be trained. In the embodiments of the present application, three groups of simulators are built for the three mapping relationships respectively, and each group of simulators includes a crack simulation generator and a fuzz simulation generator inside; in the second stage, a simulator for G --> H[Excessive amount of hairiness] needs to be trained.

[0051] Specifically, the training methods of any one of the mapping relationship simulators in the first stage are the same. In the embodiments of the present application, the training process of A[Raw material defect] --> G[Surface crack / fuzz] is used as an exemplary illustration, and the other two can be implemented with reference to the training process of A[Raw material defect] --> G[Surface crack / fuzz].

[0052] Details are as follows:

[0053] Furthermore, several crack simulations and fuzz simulations are generated according to the impurity scale, bubble scale, and the ratio of the shrinkage rate of the cortex to the core layer of the raw material defect, and several first crack distribution scales and several first fuzz distribution scales are obtained, including the steps:

[0054] S11: Limited by the production process, production equipment model, and carbon fiber model, collect the carbon fiber preparation record data with the impurity scale, bubble scale, and the ratio of the shrinkage rate of the cortex to the core layer as the only variables, and extract multiple groups of data. Any group of the multiple groups of data includes impurity scale record parameters, bubble scale record parameters, the ratio of the shrinkage rate of the cortex to the core layer record parameters, crack distribution scale detection values, and fuzz distribution scale detection values;

[0055] S12: Supervise and train the crack simulation generator according to the impurity scale record parameters, bubble scale record parameters, the ratio of the shrinkage rate of the cortex to the core layer record parameters, and crack distribution scale detection values of the multiple groups of data;

[0056] S13: Supervise and train the fuzz simulation generator according to the impurity scale record parameters, bubble scale record parameters, the ratio of the shrinkage rate of the cortex to the core layer record parameters, and fuzz distribution scale detection values of the multiple groups of data.

[0057] Specifically, in the chemical production process, minor deviations can cause significant differences. Therefore, in the embodiments of the present application, models are separately built for the production process, production equipment models, and carbon fiber models in a limited scenario, and they only have generalization ability in the limited scenario and are not adaptable to other scenarios. During the training process from A [raw material defects] -> G [surface cracks / fuzz], the impurity scale, bubble scale, and the ratio of the shrinkage rate of the cortex to the core are used as input data. Therefore, it is required to limit that other influencing parameters are all kept within the standard range, and the carbon fiber preparation record data with the impurity scale, bubble scale, and the ratio of the shrinkage rate of the cortex to the core as the only variables are collected, and multiple groups of data are extracted. Any group of the multiple groups of data includes the impurity scale record parameter, bubble scale record parameter, the ratio of the shrinkage rate of the cortex to the core record parameter, the crack distribution scale detection value, and the fuzz distribution scale detection value. The crack distribution scale detection value of each group of data in the embodiments of the present application is preferably characterized by the average number of cracks distributed in a carbon fiber product per unit area, and the fuzz distribution scale detection value is preferably characterized by the average weight of fuzz in a carbon fiber product per unit weight.

[0058] Furthermore, for the crack simulation generator and the fuzz simulation generator, the model architectures are configured as follows:

[0059] Input layer [impurity scale, bubble scale, ratio of shrinkage rate of cortex to core] -> Shared feature extraction layer (preferably a convolutional fully connected neural network in the embodiments of the present application);

[0060] Shared feature extraction layer -> Crack branch: fully connected layer + ReLU;

[0061] Shared feature extraction layer -> Fuzz branch: fully connected layer + Sigmoid;

[0062] Crack branch -> Crack distribution scale;

[0063] Fuzz branch -> Fuzz distribution scale.

[0064] Among them, the crack branch is the crack simulation generator in the embodiments of the present application, preferably using a deep neural network (DNN), with a hidden layer structure of 256-128-64 and an activation function of LeakyReLU (α = 0.01). The fuzz branch is the fuzz simulation generator in the embodiments of the present application, preferably introducing an attention mechanism (AttentionLayer), which can focus on the interaction between bubbles and the shrinkage rate. The loss function of the crack branch selects the mean square error loss function, the convergence error threshold is set to 0.15, and the coefficient of determination is set to 0.85. The fuzz branch selects the mean absolute percentage error, and the convergence error threshold is set to 8%.

[0065] During specific training, multiple groups of data are divided in an 8:2 ratio. 80% of the data is selected as training data, and 20% of the data is selected as validation data. First, 80% of the training data is retrieved, and the impurity scale, bubble scale, and the ratio of the shrinkage rate of the cortex to the core are used as the input data of the input layer. The crack distribution scale detection value and the hairiness distribution scale detection value are used as supervision. Then, the mean square error and the mean absolute percentage error are calculated through the loss function. If the mean square error of at least 85% of the output errors is less than or equal to 0.15 and the mean absolute percentage error is less than or equal to 8% for continuously preset times of training, then 20% of the data is retrieved for continued training. If the mean square error of at least 85% of the output errors is less than or equal to 0.15 and the mean absolute percentage error is less than or equal to 8% for continuously preset times of training, it is considered convergent, and a crack simulation generator and a hairiness simulation generator are obtained.

[0066] Simulators with other two mapping relationships are trained in the same way.

[0067] Furthermore, the process of training the simulator of G-->H [excessive hairiness] in the second stage is as follows:

[0068] Further, based on the crack distribution scale and the hairiness distribution scale, hairiness amount simulation generation is performed to obtain the predicted hairiness amount, including the steps of:

[0069] S14: Limited by the production process, production equipment model, and carbon fiber model, carbon fiber preparation record data is collected. Among them, multiple groups of data are extracted, and any group of the multiple groups of data includes a crack distribution scale identification value, a hairiness distribution scale identification value, and a true value of the detected hairiness amount;

[0070] S15: Based on the crack distribution scale identification value, hairiness distribution scale identification value, and true value of the detected hairiness amount of the multiple groups of data, a hairiness amount generation simulator is supervised and trained;

[0071] S16: Based on the hairiness amount generation simulator, the crack distribution scale and the hairiness distribution scale are trained to obtain the predicted hairiness amount.

[0072] Specifically, the model architecture preferably adopted in the embodiments of the present application is as follows:

[0073] Input layer [First crack distribution scale, Second crack distribution scale, Third crack distribution scale] --> Crack feature branch, 1D convolutional layer;

[0074] Input layer [First hairiness distribution scale, Second hairiness distribution scale, Third hairiness distribution scale] --> Hairiness feature branch, 1D convolutional layer;

[0075] Crack feature branch --> Feature splicing layer;

[0076] Hairiness feature branch --> Feature splicing layer;

[0077] Feature splicing layer --> Fully connected layer, 256 nodes;

[0078] Fully connected layer, 256 nodes --> Dropout layer, rate = 0.2, to prevent overfitting --> I [Output layer: Prediction of hairiness amount].

[0079] The loss function is preferably the mean square error loss function, and the convergence error threshold is 0.1, which is more stringent than before. Using the true value of the detected hairiness amount as the supervised data, and using the crack distribution scale identification value and the hairiness distribution scale identification value to train the foregoing model architecture, a hairiness amount generation simulator is obtained. The convergence and training processes are the same as the convergence conditions of the foregoing crack simulation generator and hairiness simulation generator. Preferably, the crack distribution scale identification value and the hairiness distribution scale identification value implemented in this application are generated by processing the data corresponding to the true value of the detected hairiness amount through the simulator in the foregoing first stage.

[0080] Furthermore, after the training is completed and during specific use, to ensure stability, the embodiments of this application reduce the probability of accidental errors through the method of generating multiple times repeatedly, as detailed below:

[0081] S20: Perform multiple crack simulations and hairiness simulations according to the raw material defects, the process mismatch parameter time series values, and the equipment misalignment parameter time series values to obtain the crack distribution scale and the hairiness distribution scale;

[0082] S30: Perform hairiness amount simulation generation according to the crack distribution scale and the hairiness distribution scale to obtain the predicted hairiness amount;

[0083] Further, performing multiple crack simulations and hairiness simulations according to the raw material defects, the process mismatch parameter time series values, and the equipment misalignment parameter time series values to obtain the crack distribution scale and the hairiness distribution scale includes the steps:

[0084] S21: Perform multiple crack simulations and hairiness simulations according to the impurity scale, bubble scale, and the ratio of the shrinkage rate of the skin layer to the core layer of the raw material defects to obtain several first crack distribution scales and several first hairiness distribution scales;

[0085] S22: Perform multiple crack simulations and hairiness simulations according to the draw ratio mismatch time series value and the heat setting temperature mismatch time series value of the process mismatch parameter time series values to obtain several second crack distribution scales and several second hairiness distribution scales;

[0086] S23: Based on the out-of-calibration sequence values of the carbonization furnace gas flow rate distribution and the wire drawing tension in the device out-of-calibration parameter sequence values, perform several crack simulations and hairiness simulations to obtain several third crack distribution scales and several third hairiness distribution scales;

[0087] S24: Calculate the mean values of the several first crack distribution scales, the several second crack distribution scales, and the several third crack distribution scales respectively, and set them as the crack distribution scale. Calculate the mean values of the several first hairiness distribution scales, the several second hairiness distribution scales, and the several third hairiness distribution scales respectively to obtain the hairiness distribution scale.

[0088] Preferably, in specific use, the output parameters are at least looped 20 times each time, then 20 first crack distribution scales and 20 first hairiness distribution scales can be obtained. Then calculate the mean value of the 20 first crack distribution scales as the first crack distribution scale, calculate the mean value of the 20 first hairiness distribution scales as the first hairiness distribution scale, calculate the second crack distribution scale and the second hairiness distribution scale in the same way, as well as the third crack distribution scale and the third hairiness distribution scale. Add the first crack distribution scale, the second crack distribution scale, and the third crack distribution scale to the crack distribution scale, add the first hairiness distribution scale, the second hairiness distribution scale, and the third hairiness distribution scale to the hairiness distribution scale, and then input the hairiness amount into the simulator to obtain the predicted hairiness amount.

[0089] S40: Retrieve and perform confidence analysis on the same-type and same-position produced carbon fiber in the preset time zone that meets the crack distribution scale and the hairiness distribution scale to obtain the hairiness amount confidence interval, where the preset time zone is the time zone obtained by pushing forward the preset duration from the current moment;

[0090] Specifically, the predicted hairiness amount generated by the aforementioned simulator is obtained by fitting based on long-term data rules. However, the actual hairiness amount value is also affected by the service life of the equipment and the workshop environment. Therefore, it is necessary to refer to the short-term characteristic hairiness amount to verify the credibility of the predicted hairiness amount. Preferably, limit the model of the carbon fiber product and the position number of the production line of the test sample, collect the set of reliable hairiness amount detection record values of the carbon fiber products produced in the preset time zone that have been verified by experts, use the LOF outlier factor analysis to delete outliers from the set of hairiness amount detection record values, and then statistically analyze the distribution interval of the remaining data, and set it as the hairiness amount confidence interval.

[0091] S50: When the predicted hairiness amount belongs to the hairiness amount confidence interval, receive the test hairiness amount from the indirect test end of the hairiness amount. If the deviation from the predicted hairiness amount is less than the deviation threshold, perform a credibility mark and feedback it to the indirect test end of the hairiness amount.

[0092] Specifically, when the predicted hairiness amount belongs to the hairiness amount confidence interval, it indicates that the predicted hairiness amount is applicable not only in the long term but also in the short term. Then, the measured hairiness amount is received from the indirect measurement end of the hairiness amount. If the deviation from the predicted hairiness amount is less than the deviation threshold preset by the user, a trustworthy label is given and feedback is sent to the indirect measurement end of the hairiness amount. If the deviation is greater than or equal to the deviation threshold preset by the user, a suspicious label is given and feedback is sent to the indirect measurement end of the hairiness amount. Subsequently, the data with suspicious labels is deleted, and only the data with trustworthy labels is used for mean calculation. The obtained measured hairiness amount will be relatively accurate. Moreover, due to the trustworthy label, the tester can reduce the number of measurements, thereby improving the testing efficiency.

[0093] Further, it also includes:

[0094] When the predicted hairiness amount does not belong to the hairiness amount confidence interval, obtain the central value of the hairiness amount within the hairiness amount confidence interval;

[0095] Calculate the first deviation distance between the predicted hairiness amount and the hairiness amount confidence interval;

[0096] When the first deviation distance is less than the deviation threshold, perform mean calculation on the predicted hairiness amount and the central value of the hairiness amount to obtain a corrected hairiness amount, and update the predicted hairiness amount;

[0097] When the first deviation distance is greater than or equal to the deviation threshold, return to the hairiness amount simulation generation step to execute a loop.

[0098] Specifically, the central value of the hairiness amount refers to the mode value or the mean value of the hairiness amount. The first deviation distance refers to the distance between the predicted hairiness amount and the nearest boundary of the hairiness amount confidence interval. The deviation threshold is the deviation threshold preset by the user mentioned above. When the first deviation distance is less than the deviation threshold, it indicates that although there is a deviation, it is still within the degree of being referenceable. Therefore, perform mean calculation on the predicted hairiness amount and the central value of the hairiness amount to obtain a corrected hairiness amount, and update the predicted hairiness amount. When the first deviation distance is greater than or equal to the deviation threshold, return to the hairiness amount simulation generation step to execute a loop. If the generation results of the preset number of times are all greater than or equal to the deviation threshold, then output the central value of the hairiness amount as the predicted hairiness amount.

[0099] Further, when the first deviation distance is less than the deviation threshold, performing mean calculation on the predicted hairiness amount and the central value of the hairiness amount to obtain a corrected hairiness amount includes:

[0100] Calculate the ratio of the first deviation distance to the deviation threshold, and set it as the first fitting weight;

[0101] Use 1 minus the first fitting weight to obtain the second fitting weight;

[0102] Weight the concentrated value of the hairiness amount according to the first fitting weight, and weight the predicted hairiness amount according to the second fitting weight to obtain the corrected hairiness amount.

[0103] When the predicted hairiness amount can be used as a reference, the farther it deviates, the less available it is, and the weight should be smaller. Therefore, calculate the ratio of the first deviation distance to the deviation threshold, and set it as the first fitting weight; use 1 minus the first fitting weight to obtain the second fitting weight; weight the concentrated value of the hairiness amount according to the first fitting weight, and weight the predicted hairiness amount according to the second fitting weight to obtain the corrected hairiness amount, which can improve the reliability of the mean calculation result.

[0104] It should be noted that the mean fitting rule is an exemplary description proposed in the embodiments of the present application. In actual applications, it can be set according to actual needs.

[0105] Further, perform confidence analysis on the same-type and same-position produced carbon fibers in the preset time zone that meet the crack distribution scale and the hairiness distribution scale to obtain the hairiness amount confidence interval, including:

[0106] Retrieve the first-level historical sample set of the same-type and same-position produced carbon fibers in the preset time zone that meet the crack distribution scale and the hairiness distribution scale from the test historical samples;

[0107] Extract the first-level crack distribution scale and the first-level hairiness distribution scale of each first-level historical sample in the first-level historical sample set, and retrieve the second-level historical sample set that meets the first-level crack distribution scale and the first-level hairiness distribution scale;

[0108] Perform central tendency analysis on the hairiness amount detection values of the second-level historical sample set and the first-level historical sample set to obtain the hairiness amount confidence interval.

[0109] Specifically, in order to avoid insufficient analysis data, the embodiment of the present application adopts a two-level sampling strategy for data collection. That is, the tester configures the crack distribution scale deviation threshold and the hairiness distribution scale deviation threshold according to requirements. When the crack distribution scale deviation between the sample crack distribution scale and the crack distribution scale is less than the crack distribution scale deviation threshold, and the hairiness distribution scale deviation between the sample hairiness distribution scale and the hairiness distribution scale is less than the hairiness distribution scale deviation threshold, it is considered that the crack distribution scale and the hairiness distribution scale are satisfied. First, retrieve the test historical samples that meet the preset time zone, same model, and same position number. Then, extract the first-level historical sample set of the same-type and same-position-produced carbon fiber in the preset time zone that meets the crack distribution scale and the hairiness distribution scale. Based on the first-level crack distribution scale and the first-level hairiness distribution scale of each sample in the first-level historical sample set, retrieve the second-level historical sample set that meets the first-level crack distribution scale and the first-level hairiness distribution scale. This secondary sampling will not deviate too much from the reference conditions, while increasing the data volume and improving the objectivity of the analysis results. Conduct a central tendency analysis on the hairiness amount detection values of the second-level historical sample set and the first-level historical sample set to obtain the confidence interval of the hairiness amount.

[0110] An efficient testing method for the hairiness amount of carbon fiber provided by the embodiment of the present invention has at least the following technical effects:

[0111] By tracing back the preparation data of the test samples, determine the raw material defects, process mismatch parameter time series values, and equipment misalignment parameter time series values related to the hairiness amount. Then, based on the stage characteristics of the hairiness amount generation, first perform the simulation of the generation of cracks and hairiness in the first stage, and then perform the simulation of the hairiness amount in the second stage to obtain the predicted hairiness amount. Furthermore, use the recent detection data to construct a confidence interval to perform a primary verification on the predicted hairiness amount. When it meets the confidence interval, use the predicted hairiness amount to verify the test result. If the verification passes, generate a credible label, indicating that the test result is directly available without the need for multiple tests. And because the predicted hairiness amount is a characteristic value that conforms to long-term and short-term characteristics, it reduces the probability of accidental risks of the hairiness amount of the credible label test, thus achieving the technical effects of improving efficiency and taking into account the implementation cost.

[0112] Embodiment 2, as Figure 2 shown, based on the same inventive concept as the efficient testing method for the hairiness amount of carbon fiber provided in Embodiment 1, the embodiment of the present invention also provides an efficient testing device for the hairiness amount of carbon fiber, including:

[0113] A preparation data module, configured to trace back the carbon fiber preparation data of the test samples and extract the raw material defects, process mismatch parameter time series values, and equipment misalignment parameter time series values;

[0114] A crack and hairiness generation simulation module, configured to perform a plurality of crack simulations and hairiness simulations based on the raw material defects, the process mismatch parameter time series values, and the equipment misalignment parameter time series values, and obtain a crack distribution scale and a hairiness distribution scale;

[0115] A filament amount simulation generation module, configured to perform a filament amount simulation generation based on the crack distribution scale and the hairiness distribution scale, and obtain a predicted filament amount;

[0116] A confidence interval configuration module, configured to retrieve and perform a confidence analysis on the same-type and same-position production carbon fiber in a preset time zone that satisfies the crack distribution scale and the hairiness distribution scale, and obtain a filament amount confidence interval, where the preset time zone is a time zone obtained by pushing forward a preset duration from the current moment;

[0117] A credible identification module, configured to, when the predicted filament amount belongs to the filament amount confidence interval, receive a test filament amount from an indirect filament amount test end, and if the deviation from the predicted filament amount is less than a deviation threshold, perform a credible identification and feedback it to the indirect filament amount test end.

[0118] Further, it further includes a predicted filament amount correction module, and the execution steps include:

[0119] When the predicted filament amount does not belong to the filament amount confidence interval, obtain the filament amount central value of the filament amount confidence interval;

[0120] Calculate a first deviation distance between the predicted filament amount and the filament amount confidence interval;

[0121] When the first deviation distance is less than the deviation threshold, perform a mean calculation on the predicted filament amount and the filament amount central value to obtain a corrected filament amount, and update the predicted filament amount;

[0122] When the first deviation distance is greater than or equal to the deviation threshold, return to the filament amount simulation generation step to execute a loop.

[0123] Further, the execution steps of the crack and hairiness generation simulation module include:

[0124] Perform a plurality of crack simulations and hairiness simulations according to the impurity scale, bubble scale, and the shrinkage rate ratio of the cortex to the core layer of the raw material defects, and obtain a plurality of first crack distribution scales and a plurality of first hairiness distribution scales;

[0125] Perform a plurality of crack simulations and hairiness simulations according to the draw ratio mismatch time series value and the heat setting temperature mismatch time series value of the process mismatch parameter time series values, and obtain a plurality of second crack distribution scales and a plurality of second hairiness distribution scales;

[0126] Based on the out-of-alignment time-series values of the carbonization furnace gas flow rate distribution and the wire take-up tension in the device, perform several crack simulations and hairiness simulations to obtain several third crack distribution scales and several third hairiness distribution scales;

[0127] Calculate the means of the several first crack distribution scales, the several second crack distribution scales, and the several third crack distribution scales, and set them as the crack distribution scale. Calculate the means of the several first hairiness distribution scales, the several second hairiness distribution scales, and the several third hairiness distribution scales to obtain the hairiness distribution scale.

[0128] Further, the steps executed by the crack and hairiness generation simulation module include:

[0129] Limited by the production process, production equipment model, and carbon fiber model, collect carbon fiber preparation record data with the impurity scale, bubble scale, and the ratio of skin layer to core layer shrinkage rate as the only variables, and extract multiple groups of data. Any group of the multiple groups of data includes impurity scale record parameters, bubble scale record parameters, skin layer to core layer shrinkage rate ratio record parameters, crack distribution scale detection values, and hairiness distribution scale detection values;

[0130] Supervise and train the crack simulation generator according to the impurity scale record parameters, bubble scale record parameters, skin layer to core layer shrinkage rate ratio record parameters, and crack distribution scale detection values of the multiple groups of data;

[0131] Supervise and train the hairiness simulation generator according to the impurity scale record parameters, bubble scale record parameters, skin layer to core layer shrinkage rate ratio record parameters, and hairiness distribution scale detection values of the multiple groups of data.

[0132] Further, the steps executed by the hairiness amount simulation generation module include:

[0133] Limited by the production process, production equipment model, and carbon fiber model, collect carbon fiber preparation record data, and extract multiple groups of data. Any group of the multiple groups of data includes crack distribution scale identification values, hairiness distribution scale identification values, and hairiness amount detection true values;

[0134] Supervise and train the hairiness amount generation simulator according to the crack distribution scale identification values, hairiness distribution scale identification values, and hairiness amount detection true values of the multiple groups of data;

[0135] Train the crack distribution scale and the hairiness distribution scale according to the hairiness amount generation simulator to obtain the predicted hairiness amount.

[0136] Further, the steps executed by the confidence interval configuration module include:

[0137] Retrieving a primary historical sample set of carbon fibers of the same type and position produced in a preset time zone that meet the crack distribution scale and the hairiness distribution scale from the test historical samples;

[0138] Extracting the primary crack distribution scale and the primary hairiness distribution scale of each primary historical sample of the primary historical sample set, and retrieving the secondary historical sample set that satisfies the primary crack distribution scale and the primary hairiness distribution scale;

[0139] A central tendency analysis is performed on the hair quantity detection values ​​of the secondary historical sample set and the primary historical sample set to obtain the hair quantity confidence interval.

[0140] Furthermore, the execution steps of the predicted hair amount correction module include:

[0141] Calculating a ratio of the first deviation distance to the deviation threshold, and setting the ratio as a first fitting weight;

[0142] Subtracting the first fitting weight from 1 to obtain a second fitting weight;

[0143] The concentrated value of the hair amount is weighted according to the first fitting weight, and the predicted hair amount is weighted according to the second fitting weight to obtain the corrected hair amount.

[0144] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0145] It should be understood by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1apparatus for the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction apparatus that implements the functions specified in one Figure 1 or more processes and / or boxes Figure 1 or one or more boxes.

[0148] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one Figure 1 or more processes and / or boxes Figure 1 or one or more boxes.

[0149] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept.

[0150] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. An efficient method for testing the amount of carbon fiber filaments, characterized in that: include: Back-test the carbon fiber preparation data of the sample to extract the raw material defects, process mismatch parameter timing values ​​and equipment misalignment parameter timing values; Performing crack simulation generation and hairiness simulation generation several times according to the raw material defects, the process mismatch parameter timing values, and the equipment misalignment parameter timing values ​​to obtain a crack distribution scale and a hairiness distribution scale; Simulating and generating the amount of hairiness according to the crack distribution scale and the hairiness distribution scale to obtain a predicted amount of hairiness; Retrieving carbon fibers of the same type and position produced in a preset time zone that meets the crack distribution scale and the hairiness distribution scale, performing confidence analysis, and obtaining a confidence interval of the hairiness amount, wherein the preset time zone is a time zone obtained by pushing a preset time length forward from the current moment; When the predicted hair amount belongs to the hair amount confidence interval, the tested hair amount is received from the hair amount indirect testing end. If the deviation from the predicted hair amount is less than the deviation threshold, a credible mark is performed and fed back to the hair amount indirect testing end.

2. The method according to claim 1, characterized in that Also includes: When the predicted hair amount does not belong to the hair amount confidence interval, obtaining a hair amount concentration value of the hair amount confidence interval; Calculating a first deviation distance between the predicted hair amount and the hair amount confidence interval; When the first deviation distance is less than the deviation threshold, average calculation is performed on the predicted hair amount and the concentrated value of the hair amount to obtain a corrected hair amount, and the predicted hair amount is updated; When the first deviation distance is greater than or equal to the deviation threshold, the process returns to the step of fuzz amount simulation generation to execute a loop.

3. The method according to claim 1, characterized in that According to the raw material defects, the process mismatch parameter timing values ​​and the equipment misalignment parameter timing values, several crack simulations and hairiness simulations are performed to obtain the crack distribution scale and the hairiness distribution scale, including: According to the impurity scale, bubble scale, and skin-to-core shrinkage ratio of the raw material defects, several crack simulations and hairiness simulations are performed to obtain several first crack distribution scales and several first hairiness distribution scales; According to the draft ratio mismatch timing value and the heat setting temperature mismatch timing value of the process mismatch parameter timing value, a plurality of crack simulation generation and hairiness simulation generation are performed to obtain a plurality of second crack distribution scales and a plurality of second hairiness distribution scales; According to the carbonization furnace gas velocity distribution misalignment timing value and the wire drawing tension misalignment timing value of the equipment misalignment parameter timing value, several crack simulation generation and hairiness simulation generation are performed to obtain several third crack distribution scales and several third hairiness distribution scales; The average values ​​of the first crack distribution scales, the second crack distribution scales and the third crack distribution scales are calculated respectively, set as the crack distribution scale, and the average values ​​of the first hairiness distribution scales, the second hairiness distribution scales and the third hairiness distribution scales are calculated respectively to obtain the hairiness distribution scale.

4. The method according to claim 3, characterized in that According to the impurity scale, bubble scale, and skin-to-core shrinkage ratio of the raw material defects, several crack simulations and hairiness simulations are performed to obtain several first crack distribution scales and several first hairiness distribution scales, including: Limited to the production process, production equipment model and carbon fiber model, collect carbon fiber preparation record data with impurity scale, bubble scale and skin layer to core layer shrinkage ratio as unique variables, extract multiple groups of data, any one of the multiple groups of data includes impurity scale record parameters, bubble scale record parameters, skin layer to core layer shrinkage ratio record parameters, crack distribution scale detection value and hairiness distribution scale detection value; According to the impurity scale recording parameters, bubble scale recording parameters, skin layer and core layer shrinkage ratio recording parameters, and crack distribution scale detection values ​​of the multiple sets of data, supervise the training of the crack simulation generator; According to the impurity scale recording parameters, bubble scale recording parameters, skin layer and core layer shrinkage ratio recording parameters, and hairiness distribution scale detection values ​​of the multiple groups of data, supervised training of the hairiness simulation generator.

5. The method according to claim 1, characterized in that The method of simulating and generating the amount of hairiness according to the crack distribution scale and the hairiness distribution scale to obtain the predicted amount of hairiness includes: The carbon fiber preparation record data is collected based on the production process, production equipment model and carbon fiber model, wherein multiple groups of data are extracted, and any group of data of the multiple groups of data includes a crack distribution scale identification value, a hairiness distribution scale identification value and a hairiness amount detection true value; According to the crack distribution scale identification values, hairiness distribution scale identification values ​​and hairiness amount detection true values ​​of the plurality of groups of data, supervised training of a hairiness amount generation simulator; The crack distribution scale and the hairiness distribution scale are trained according to the hairiness amount generation simulator to obtain the predicted hairiness amount.

6. The method according to claim 1, characterized in that Retrieving carbon fibers of the same type and position produced in a preset time zone that meet the crack distribution scale and the hairiness distribution scale, performing confidence analysis, and obtaining a confidence interval of the hairiness amount, including: Retrieving a primary historical sample set of carbon fibers of the same type and position produced in a preset time zone that meet the crack distribution scale and the hairiness distribution scale from the test historical samples; Extracting the primary crack distribution scale and the primary hairiness distribution scale of each primary historical sample of the primary historical sample set, and retrieving the secondary historical sample set that satisfies the primary crack distribution scale and the primary hairiness distribution scale; A central tendency analysis is performed on the hair quantity detection values ​​of the secondary historical sample set and the primary historical sample set to obtain the hair quantity confidence interval.

7. The method according to claim 2, characterized in that When the first deviation distance is less than the deviation threshold, the predicted hair amount and the hair amount concentration value are averaged to obtain a corrected hair amount, including: Calculating a ratio of the first deviation distance to the deviation threshold, and setting the ratio as a first fitting weight; Subtracting the first fitting weight from 1 to obtain a second fitting weight; The concentrated value of the hair amount is weighted according to the first fitting weight, and the predicted hair amount is weighted according to the second fitting weight to obtain the corrected hair amount.

8. An efficient testing device for the amount of carbon fiber filaments, characterized in that: include: Preparation data module, used to trace back the carbon fiber preparation data of the test sample, extract the raw material defects, process mismatch parameter timing values ​​and equipment misalignment parameter timing values; A crack and hairiness generation simulation module, used to perform several crack simulations and hairiness simulations according to the raw material defects, the process mismatch parameter timing values ​​and the equipment misalignment parameter timing values, to obtain crack distribution scales and hairiness distribution scales; A hair quantity simulation generation module is used to simulate and generate the hair quantity according to the crack distribution scale and the hair distribution scale to obtain a predicted hair quantity; A confidence interval configuration module is used to retrieve carbon fibers of the same type and position produced in a preset time zone that meets the crack distribution scale and the hairiness distribution scale, perform confidence analysis, and obtain a confidence interval of the hairiness amount, wherein the preset time zone is a time zone obtained by pushing a preset time length forward from the current moment; The trusted identification module is used to receive the tested hair amount from the hair amount indirect testing end when the predicted hair amount belongs to the hair amount confidence interval, and if the deviation from the predicted hair amount is less than the deviation threshold, perform a trusted identification and feedback to the hair amount indirect testing end.

Images