Method, system and device for predicting tensile strength of fibrilia

Abstract

The invention provides a jute fiber tensile strength prediction method, system and device, and relates to the technical field of jute fiber mechanical property analys.The method mainly comprises the steps that target chemical components are determined through a data analysis method; on the basis of the characteristic wave band of the target chemical component, near infrared spectrum data of the detected fibrilia are collected; after the near infrared spectrum data are preprocessed, component content data of target chemical components in the detected fibrilia are obtained through component content prediction; and performing tensile strength prediction on the component content data to obtain the tensile strength of the detected fibrilia. According to the scheme, the target chemical component content and the tensile strength of the fibrilia sample can be reliably predicted, the prediction efficiency is high, the prediction conclusion is reliable, the fibrilia sample does not make contact with harmful products such as chemical reagents in prediction application, the human body is not hurt, and errors caused by manual test operation can be effectively reduced.

Description

Technical Field

[0001] The invention relates to the technical field of hemp fiber mechanical property analysis, and in particular to a hemp fiber tensile strength prediction method, system and device. Background Art

[0002] In hemp fiber composites, the properties of the fiber play a decisive role in the performance of the composites, so the analysis of the mechanical properties of hemp fibers is the key to improving the quality of hemp fiber composites.

[0003] The composition of hemp fiber is complex, and determining the content of its effective ingredients is the premise and basis for mechanical property analysis.

[0004] At present, when analyzing the composition of hemp fiber, it is often carried out according to the provisions of the Chinese national standard GB / T5889-1986 "Quantitative Analysis Method of Chemical Composition of Ramie". However, this analysis method has many problems such as long time consumption, complex operation, large operation error, harm to human body and environmental pollution.

[0005] At the same time, when conducting mechanical analysis on hemp fibers, there are also technical problems such as cumbersome sample preparation, complex operation, and low efficiency. Summary of the invention

[0006] The object of the present invention is to provide a method, system and device for predicting the tensile strength of hemp fibers, so as to solve at least one of the above-mentioned technical problems existing in the prior art.

[0007] In a first aspect, in order to solve the above technical problems, the present invention provides a method for predicting the tensile strength of hemp fiber, comprising the following steps:

[0008] Step 1: Based on the chemical component content, near infrared spectrum data and raw data of tensile strength of hemp fiber, target chemical components are determined by data analysis method, so as to determine the chemical components that mainly affect the tensile strength of hemp fiber; the data analysis method includes principal component analysis, cluster analysis and correlation analysis, etc.;

[0009] The principal component analysis is used to preliminarily determine the range of the target chemical components;

[0010] The cluster analysis is used to further narrow the scope of target chemical components;

[0011] The correlation analysis is used to determine the degree of correlation between the target chemical composition and the tensile strength;

[0012] The target chemical components include cellulose and / or hemicellulose and / or lignin;

[0013] Step 2: Based on the characteristic bands of the target chemical components, collect near infrared spectrum data of the tested hemp fiber;

[0014] Step 3, after preprocessing the near infrared spectrum data, the component content data of the target chemical component in the tested hemp fiber is obtained by component content prediction;

[0015] Step 4: The component content data is used to predict the tensile strength to obtain the tensile strength of the tested hemp fiber.

[0016] Through the above steps, the specific spectral characteristics of cellulose, hemicellulose and lignin in hemp fiber can be used to predict the component content, and the component content data of the three can be obtained simply, effectively and safely; then, through tensile strength prediction, the tensile strength of the corresponding hemp fiber can be obtained simply, effectively and safely; this scheme can be used for non-destructive and rapid tensile strength analysis of hemp fiber, without causing problems such as chemical reagent consumption and pollution, and will not cause harm to human health.

[0017] Of course, this solution can also be extended to other plants with complex chemical compositions, and similar technical effects can be achieved.

[0018] In a feasible implementation, the specific method of principal component analysis includes: establishing a principal component regression model; taking all chemical components of a hemp fiber sample as original variables; calculating eigenvalues ​​and eigenvectors through data standardization processing and correlation coefficient matrix; determining principal component factors through variance contribution rate and cumulative variance contribution rate; taking root mean square error and mean absolute error as evaluation indicators to evaluate the regression effect of the predicted value of the principal component regression model on the test value (obtained from wet quantitative analysis); performing principal component regression analysis on the hemp fiber sample, calculating the variance contribution rate and the cumulative variance contribution rate and sorting them, and screening out a number of principal component factors when the cumulative variance contribution rate reaches a preset cumulative variance contribution rate (for example, 98%); the principal component factors are linear combinations of target chemical components.

[0019] Preferably, the specific calculation formula of the root mean square error is:

[0020]

[0021] Where RMSE represents the root mean square error; n represents the number of samples; represents the test value of the i-th sample; y i represents the predicted value of the i-th sample;

[0022] The specific calculation formula of the mean absolute error is:

[0023]

[0024] Here, MAE stands for mean absolute error.

[0025] In a feasible implementation, the specific method of the cluster analysis includes:

[0026] A certain number of samples or indicators are regarded as one category respectively; according to the closeness of the samples or indicators, the two categories with the highest closeness are merged; considering the closeness between the merged category and other categories, merge them again until all samples or indicators are merged into one category.

[0027] In a feasible implementation manner, the specific process of the correlation analysis includes:

[0028] Step a1: Take the tensile strength as the dependent variable series and the target chemical component content as the independent variable series to form the original data matrix W1. The specific expression can be:

[0029] W1=|X1 X2 … X i-1 Y|

[0030] Among them, X i-1 represents the content of the i-1th target chemical component; Y represents the tensile strength;

[0031] The data is normalized to eliminate the large differences between the data and increase comparability, and the normalized matrix W2 is obtained. The specific expression can be:

[0032] W2=(w ki );

[0033] Among them, w ki represents the normalized value of the i-th target chemical component of the k-th hemp fiber sample. The specific expression can be:

[0034]

[0035] Among them, maxx ki represents the maximum eigenvalue of the i-th target chemical component of the k-th hemp fiber sample; minx ki represents the minimum eigenvalue of the i-th target chemical component of the k-th hemp fiber sample;

[0036] Step a2, calculating the correlation coefficient; the correlation coefficient is the correlation degree value between the independent variable sequence and the dependent variable sequence at each moment; the larger the correlation coefficient, the greater the influence of the independent variable at that moment; for example, for different types of hemp fiber samples, there may be multiple independent variable sequences, namely, cellulose content sequence, hemicellulose content sequence, lignin content sequence, etc., and one dependent variable sequence is a tensile strength sequence; the specific calculation formula for the correlation coefficient between each independent variable sequence and the dependent variable sequence is:

[0037]

[0038] Among them, ζi(k) represents the correlation coefficient; ζ represents the resolution coefficient, and ζ∈[0,1], generally taking the value as 0.5; min(Δi(min)) represents the minimum difference between the two levels; max(Δi(max)) represents the maximum difference between the two levels. The specific calculation formula can be:

[0039]

[0040] Step a3: Calculate the correlation between the independent variable series and the dependent variable series; then calculate the average correlation r i , used to indicate the degree of correlation between the independent variable series and the dependent variable series, which is convenient for overall comparison when there are many correlations; the average correlation r i The larger the value is, the greater the influence of the corresponding independent variable is. The specific formula can be:

[0041]

[0042] Wherein, N represents the number of hemp fiber samples.

[0043] In a feasible implementation manner, the characteristic band specifically includes:

[0044] The data points in the 1600-2400 nm band correspond to the prediction of cellulose component content;

[0045] The data points in the 1400-2400 nm band correspond to the predicted content of hemicellulose components;

[0046] The data points in the 1200-2400 nm band correspond to the predicted lignin content.

[0047] In a feasible implementation manner, the method for determining the characteristic band may specifically be: within the full band range of near-infrared spectral data (800-2400nm), a preset band interval length (which may be selected with reference to the common range of peak width of 100-200nm, for example, 200nm) is used as a tolerance, and several candidate characteristic bands are progressively selected from the short-wave end to the long-wave end in an arithmetic progression, and a cyclic test is performed through component content prediction, and the root mean square error (RMSE) and mean absolute error (MAE) of the target chemical component content are calculated, and the candidate characteristic band with the lowest root mean square error is used as the characteristic band of the target chemical component;

[0048] Preferably, the PCR correlation of the component content prediction can also be statistically analyzed to test the slope of the linear fitting equation and the Pearson correlation coefficient between the predicted value and the test value (for example, the mean cellulose content detected by wet chemical analysis) under the alternative characteristic band, so as to further verify the internal cross-validation effect of the spectral data in the alternative characteristic band.

[0049] In a feasible implementation manner, the method for determining the characteristic band may specifically be as follows: within the full band range of near-infrared spectral data, the bands are divided based on the positions where the peaks appear, and several alternative characteristic bands are progressively selected from the short-wave end to the long-wave end. Cyclic testing is performed through component content prediction, and the root mean square error and mean absolute error of the target chemical component content are calculated. The alternative characteristic band with the lowest root mean square error is used as the characteristic band of the target chemical component. In this way, the characteristic band can be obtained more quickly.

[0050] In a feasible implementation manner, the method for determining the characteristic band may also be specifically as follows: within the full band range of the near-infrared spectrum data (800-2400nm), based on the initial starting interval length LC (e.g., 100nm) as a moving window, using the preset scanning accuracy of the scanning device (e.g., 6nm) as a step length, iteratively selecting the candidate characteristic band (e.g., 800-900nm, 806-906nm, ..., 2300-2400nm) by equidistant movement from the short-wave end to the long-wave end, the specific calculation formula may be:

[0051] LT = LC + (T-1) * s;

[0052] Among them, LT represents the length of the starting interval of the Tth generation; T represents the number of iterations;

[0053] Until the start interval length is equal to the preset extreme value (e.g. 2298nm);

[0054] All candidate characteristic bands of each generation are tested through component content prediction, and the root mean square error and mean absolute error of the target chemical component content are calculated. The alternative characteristic band with the lowest root mean square error is used as the characteristic band of the target chemical component; in this way, a more accurate characteristic band can be obtained.

[0055] In a feasible implementation manner, the pretreatment method in step 3 mainly includes:

[0056] Step b1, counting the lowest value of the near infrared spectrum data obtained from each hemp fiber sample, and then subtracting the lowest value of each spectrum data from each near infrared spectrum data to obtain an intermediate value; thereby completing the baseline removal operation;

[0057] Step b2, in the intermediate value, after calculating the slopes of two adjacent points, the preprocessed near-infrared spectral data is obtained; thereby completing the first-order derivative, which can avoid problems such as background noise and light scattering, thereby making the data features more obvious, and greatly eliminating the errors caused by differences in detection position, instrument status, and test environment, which is beneficial to the prediction accuracy and stability of subsequent modeling.

[0058] In a feasible implementation, the component content prediction can be achieved by a pre-trained component content prediction model (e.g., a classification model), and the specific training method includes:

[0059] Step c1, collecting a preset number (e.g., 35) of hemp fiber samples and numbering them; selecting a preset weight (e.g., 1 gram) of hemp fiber from each hemp fiber sample, chopping it into fiber powder of a preset length (e.g., less than 1 mm), drying it to a constant weight at a preset temperature (e.g., 105° C.), and cooling it in a dryer for a preset time (e.g., 30 minutes);

[0060] Step c2: using a spectrophotometer (such as Perkin-Elmer UV / VIS / NIR 750) to collect near infrared spectra of the fiber powder by diffuse reflection in a preset spectral band range;

[0061] Preferably, step c2 specifically comprises: calibrating the spectrophotometer by a standard white plate; crushing the fiber powder to a preset particle size (e.g., 60 mesh), placing it in a near-infrared absorption sample pool of the spectrophotometer, and performing multiple sample loading and diffuse reflection scanning (e.g., repeating the sample loading twice, and scanning twice each time) at a preset scanning interval (e.g., 6 nm), and taking the average spectral value as the spectral curve of each hemp fiber sample (each spectral curve may contain about 267 absorbance data);

[0062] Step c3, preprocessing the spectral curve (refer to the above-mentioned preprocessing method);

[0063] Step c4, sampling the hemp fiber samples according to the numbers, and performing wet quantitative analysis according to GB / T5889-1986 "Quantitative Analysis Method for Chemical Components of Ramie" to obtain test values ​​of the contents of fat wax, water-soluble matter, pectin, hemicellulose, lignin and cellulose in the hemp fiber samples; for each hemp fiber sample, the test value is marked according to the number;

[0064] Step c5, according to the test value, the hemp fiber sample is divided into a calibration set and a prediction set according to a preset ratio (for example, 6:1); the test value range of the calibration set is larger than the test value range of the prediction set; the calibration set is used to train the component content prediction model; the prediction set is used to verify the component content prediction model; this is conducive to limiting the predicted component within the calibration range to avoid interference from the superposition effect of the near-infrared spectrum peak information.

[0065] In a feasible implementation manner, the specific method for predicting tensile strength includes:

[0066] Step d1, based on the standard test method for tensile properties (e.g., the standard test method for tensile properties of single textile fibers, i.e., ASTM D3822 / D3822M-14 Standard Test Method for Tensile Properties of Single Textile Fibers), measuring the tensile strength of the hemp fiber sample; using the content of the target chemical component as input data and the tensile strength as output data to construct a sample set;

[0067] Step d2: normalize the input data and output data. The specific formula can be:

[0068]

[0069] in, represents the normalized value of the i-th target chemical component in the sample with the j-th tensile strength; X ij represents the i-th target chemical composition in the sample with the j-th tensile strength; x imin represents the i-th target chemical composition in the sample with the minimum tensile strength; x imax represents the i-th target chemical composition in the sample with the maximum tensile strength; represents the normalized value of the j-th tensile strength; y j represents the jth tensile strength; y min Indicates the minimum tensile strength; y max Indicates the maximum tensile strength;

[0070] Step d3, constructing a support vector machine: the input layer is the content of the target chemical component; the middle layer is to select the kernel function by learning the sample set; the last layer is to construct a nonlinear prediction function.

[0071] Preferably, the specific steps of step d3 include:

[0072] Step d31: Based on the sample set and any given ε>0, the ε-linear programming problem is transformed into an optimization problem, and then:

[0073]

[0074] Wherein, w represents the normal vector of the hyperplane; N represents the number of hemp fiber samples;

[0075] Step d32: Find a kernel function K(·,·) that satisfies the condition

[0076] in, Represents the mapping from the input space to a feature space;

[0077] Step d33: Introduce slack variable ξ i , and i=1,2,3,...,N, and the optimization model is established by Lagrange multiplier method:

[0078]

[0079] Where C represents the penalty parameter; b represents the optimization model constant;

[0080] According to the nonlinear support vector machine theory, the dual problem of the optimization model is obtained, and the specific formula is:

[0081]

[0082] in, represents the optimal solution for the i-th target chemical composition; α i represents the general solution of the i-th target chemical composition; represents the optimal solution for the jth target chemical composition; α j represents the general solution of the jth target chemical composition;

[0083] Then solve to get the support vector of the i-th target chemical component and the conjugate vector of the support vector of the i-th target chemical component

[0084] Step d34: Calculate the optimization model constant. The specific formula may be:

[0085]

[0086] Where N' represents the number of target chemical components; represents the support vector of the jth target chemical component; express The conjugate vector of ; Support vectors representing constants of the optimization model;

[0087] Step d35: construct a nonlinear prediction function for prediction and simulation. The specific formula may be:

[0088]

[0089] Among them, x i represents the i-th chemical component, and x i ∈R 6 ; b represents the optimization model constant, and b∈R; R represents a real number;

[0090] In this way, a tensile strength prediction model is constructed. Compared with the commonly used neural network regression model, this tensile strength prediction model has a rigorous theoretical and mathematical basis, is based on the principle of structural risk minimization, has better generalization ability, and the algorithm has global optimality, avoiding the problem that neural networks are prone to fall into local optimality.

[0091] In the second aspect, based on the same inventive concept, the present application also provides a hemp fiber tensile strength prediction system, including a data receiving module, a data processing module and a result generating module;

[0092] The data receiving module is used to receive near-infrared spectrum data of characteristic bands of target chemical components in the tested hemp fiber;

[0093] The data processing module includes a data analysis unit, a component content prediction unit and a tensile strength prediction unit;

[0094] The data analysis unit determines the target chemical composition of the hemp fiber through a data analysis method based on the chemical composition content, near infrared spectrum data and raw data of tensile strength of the hemp fiber; the data analysis method includes principal component analysis, cluster analysis and correlation analysis in sequence;

[0095] The principal component analysis is used to preliminarily determine the range of the target chemical components;

[0096] The cluster analysis is used to further narrow the scope of target chemical components;

[0097] The correlation analysis is used to determine the degree of correlation between the target chemical composition and the tensile strength;

[0098] The component content prediction unit is used to obtain component content data of target chemical components in the tested hemp fiber by component content prediction after preprocessing the near infrared spectrum data;

[0099] The tensile strength prediction unit is used to obtain the tensile strength of the tested hemp fiber by using the component content data through tensile strength prediction;

[0100] The result generation module is used to externalize the target chemical composition and tensile strength.

[0101] In the third aspect, based on the same inventive concept, the present application also provides a hemp fiber tensile strength prediction device, comprising a processor, a memory and a bus, wherein the memory stores instructions and data readable by the processor, the processor is used to call the instructions and data in the memory to execute the hemp fiber tensile strength prediction method as described above, and the bus connects the functional components for transmitting information.

[0102] By adopting the above technical solution, the present invention has the following beneficial effects:

[0103] The present invention provides a hemp fiber tensile strength prediction method, system and device, which can perform data analysis on the complex chemical components of hemp fiber, automatically obtain three target chemical components that play a leading role in the tensile strength of hemp fiber, and quickly predict the contents of these three target chemical components based on non-destructive near-infrared spectral data and component content prediction, and then reliably predict the tensile strength of hemp fiber samples through tensile strength prediction. This solution not only has high prediction efficiency and reliable prediction conclusions, but also does not contact hazardous substances such as chemical reagents in the prediction application, does not cause harm to the human body, and can effectively reduce errors caused by human experimental operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0105] Figure 1 A flow chart of a method for predicting tensile strength of hemp fiber provided by an embodiment of the present invention;

[0106] Figure 2 A flowchart of correlation analysis provided by an embodiment of the present invention;

[0107] Figure 3 An example diagram of the PCR correlation between the predicted value and the tested value of cellulose content provided in an embodiment of the present invention;

[0108] Figure 4 An example diagram of the PCR correlation between the predicted value and the tested value of hemicellulose content provided in an embodiment of the present invention;

[0109] Figure 5 An example diagram of the PCR correlation between the predicted value and the tested value of lignin content provided in the embodiment of the present invention;

[0110] Figure 6A comparison diagram of near infrared spectroscopy data before and after pretreatment provided by an embodiment of the present invention: Figure a shows cellulose before pretreatment; Figure b shows cellulose after pretreatment; Figure c shows hemicellulose before pretreatment; Figure d shows hemicellulose after pretreatment; Figure e shows lignin before pretreatment; Figure f shows lignin after pretreatment;

[0111] Figure 7 A flow chart of a method for training a component content prediction model provided by an embodiment of the present invention;

[0112] Figure 8 A flow chart of a method for constructing a tensile strength prediction model provided by an embodiment of the present invention;

[0113] Fig. 9 A diagram of a hemp fiber tensile strength prediction system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0114] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0115] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0116] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0117] The present invention is further explained below in conjunction with specific implementation modes.

[0118] It should also be noted that the following specific embodiments or specific implementations are a series of optimized settings listed in the present invention to further explain the specific content of the invention, and these settings can be used in combination or in association with each other.

[0119] Embodiment 1:

[0120] like Figure 1 As shown, this embodiment provides a method for predicting tensile strength of hemp fiber, comprising the following steps:

[0121] Step 1: Based on the chemical component content, near infrared spectrum data and raw data of tensile strength of 35 kinds of hemp fibers, the target chemical components are determined by data analysis methods to determine the chemical components that mainly affect the tensile strength of hemp fibers; the data analysis methods include principal component analysis, cluster analysis and correlation analysis, etc.;

[0122] The principal component analysis is used to preliminarily determine the range of the target chemical components;

[0123] The cluster analysis is used to further narrow the scope of target chemical components;

[0124] The correlation analysis is used to determine the degree of correlation between the target chemical composition and the tensile strength;

[0125] The target chemical components include cellulose and / or hemicellulose and / or lignin;

[0126] Step 2: Based on the characteristic bands of the target chemical components, collect near infrared spectrum data of the tested hemp fiber;

[0127] Step 3, after preprocessing the near infrared spectrum data, the component content data of the target chemical component in the tested hemp fiber is obtained by component content prediction;

[0128] Step 4: The component content data is used to predict the tensile strength to obtain the tensile strength of the tested hemp fiber.

[0129] Through the above steps, the specific spectral characteristics of cellulose, hemicellulose and lignin in hemp fiber can be used to predict the component content, and the component content data of the three can be obtained simply, effectively and safely; then, through tensile strength prediction, the tensile strength of the corresponding hemp fiber can be obtained simply, effectively and safely; this scheme can be used for non-destructive and rapid tensile strength analysis of hemp fiber, without causing problems such as chemical reagent consumption and pollution, and will not cause harm to human health.

[0130] Furthermore, the specific method of principal component analysis is as follows: a principal component regression model is established through data analysis software (i.e., Matlab); all chemical components of the hemp fiber sample are used as original variables; eigenvalues ​​and eigenvectors are calculated through data standardization processing and correlation coefficient matrix; principal component factors are then determined through variance contribution rate and cumulative variance contribution rate; root mean square error and mean absolute error are used as evaluation indicators to evaluate the regression effect of the predicted value of the principal component regression model on the test value (obtained from wet quantitative analysis); principal component regression analysis is performed on the hemp fiber sample, the variance contribution rate and cumulative variance contribution rate are calculated and sorted, and several principal component factors are screened out when the variance cumulative contribution rate reaches a preset variance cumulative contribution rate (i.e., 98%); the principal component factor is a linear combination of the target chemical components;

[0131] In this embodiment, 35 kinds of hemp fiber samples are collected, and the spectrum curve of each hemp fiber sample contains 267 absorbance data. After principal component analysis, it can be simplified into 17 principal component factors, which can represent 98% of the spectrum data information. These 17 principal component factors are linear combinations of absorbance data variables;

[0132] Preferably, the specific calculation formula of the root mean square error is:

[0133]

[0134] Where RMSE represents the root mean square error; n represents the number of samples; represents the test value of the i-th sample; y i represents the predicted value of the i-th sample;

[0135] The specific calculation formula of the mean absolute error is:

[0136]

[0137] Here, MAE stands for mean absolute error.

[0138] Furthermore, the specific method of the cluster analysis includes:

[0139] A certain number of samples or indicators are considered as one category respectively; according to the affinity of the samples or indicators, the two categories with the highest affinity are merged; considering the affinity between the merged category and other categories, they are merged again until all samples or indicators are merged into one category, which can reduce the six target chemical components to three, namely: cellulose, hemicellulose and lignin;

[0140] In particular, cluster analysis was performed on all the chemical components of hemp fiber and three target chemical components (cellulose, hemicellulose and lignin) as two schemes. The number of clusters was 3, and the cluster analysis results of the two schemes were consistent with a consistency of 92.3%. Therefore, the three target chemical components (cellulose, hemicellulose and lignin) can replace the 17 principal component factors to a certain extent to reflect the properties of hemp fiber.

[0141] Furthermore, if Figure 2 As shown, the specific process of the correlation analysis includes:

[0142] Step a1: Take the tensile strength as the dependent variable series and the three target chemical component contents as the independent variable series to form the original data matrix W1. The specific expression can be:

[0143] W1=|X1 X2 … X i-1 Y|

[0144] Among them, X i-1 represents the content of the i-1th target chemical component; Y represents the tensile strength;

[0145] The data is normalized to eliminate the large differences between the data and increase comparability, and the normalized matrix W2 is obtained. The specific expression can be:

[0146] W2=(w ki );

[0147] Among them, w ki represents the normalized value of the i-th target chemical component of the k-th hemp fiber sample. The specific expression can be:

[0148]

[0149] Among them, maxx ki represents the maximum eigenvalue of the i-th target chemical component of the k-th hemp fiber sample; minx ki represents the minimum eigenvalue of the i-th target chemical component of the k-th hemp fiber sample;

[0150] Step a2, calculating the correlation coefficient; the correlation coefficient is the correlation degree value between the independent variable sequence and the dependent variable sequence at each moment; the larger the correlation coefficient, the greater the influence of the independent variable at that moment; for different types of hemp fiber samples, three independent variable sequences are set, namely, cellulose content sequence, hemicellulose content sequence, lignin content sequence, and one dependent variable sequence is tensile strength sequence; the specific calculation formula for the correlation coefficient between each independent variable sequence and the dependent variable sequence is:

[0151]

[0152] Among them, ξi(k) represents the correlation coefficient; ζ represents the resolution coefficient, and ζ∈[0,1], generally taking the value as 0.5; min(Δi(min)) represents the minimum difference between the two levels; max(Δi(max)) represents the maximum difference between the two levels. The specific calculation formula can be:

[0153]

[0154] Step a3: Calculate the correlation between the independent variable series and the dependent variable series; then calculate the average correlation r i , used to indicate the degree of correlation between the independent variable series and the dependent variable series, which is convenient for overall comparison when there are many correlations; the average correlation r i The larger the value is, the greater the influence of the corresponding independent variable is. The specific formula can be:

[0155]

[0156] Wherein, N represents the number of hemp fiber samples;

[0157] After calculation, the correlation between the three target chemical components of hemp fiber and the average tensile strength is shown in Table 1:

[0158] Table 1

[0159] Chemical composition of hemp fiber Cellulose Hemicellulose Lignin Relevance 0.71 0.57 0.58

[0160] As can be seen from Table 1, cellulose is most closely related to tensile strength, while hemicellulose and lignin also play an important role in affecting tensile strength.

[0161] Furthermore, the characteristic band specifically includes:

[0162] The data points in the 1600-2400 nm band correspond to the prediction of cellulose component content;

[0163] The data points in the 1400-2400 nm band correspond to the predicted content of hemicellulose components;

[0164] The data points in the 1200-2400 nm band correspond to the predicted lignin content.

[0165] Furthermore, the method for determining the characteristic band includes: within the full band range of near-infrared spectral data (800-2400nm), using a preset band interval length (i.e., 200nm) as a tolerance, progressively selecting several alternative characteristic bands from the short-wave end to the long-wave end in an arithmetic progression, performing a cyclic test through component content prediction, calculating the root mean square error (RMSE) and mean absolute error (MAE) of the target chemical component content, and taking the alternative characteristic band when the root mean square error is the lowest as the characteristic band of the target chemical component.

[0166] Specifically, according to the above-mentioned method for determining the characteristic bands, the root mean square error and mean absolute error of the prediction results of the statistical component content prediction model for cellulose in the seven candidate characteristic bands are shown in Table 2:

[0167] Table 2

[0168]

[0169] As can be seen from Table 2, when the wavelength range is in the range of 1600-2400nm (the 4th band), the root mean square error and the mean absolute error both reach the lowest value, that is, the component content prediction model has the best prediction effect for cellulose in the 4th band. Compared with the full-band data (800-2400nm), the collection efficiency of the 4th band spectral data can be increased by 50% (wavelength range reduction rate);

[0170] Preferably, if Figure 3 As shown, the PCR correlation of the component content prediction model can also be statistically analyzed to test the slope (0.92) and Pearson correlation coefficient (0.91) of the linear fitting equation between the predicted value and the test value (i.e., the mean cellulose content detected by wet chemical analysis) under the 4th band spectral data, further indicating that the internal cross-validation effect of the 4th band spectral data is better.

[0171] Similarly, the root mean square error and mean absolute error of the prediction results of the statistical component content prediction model for hemicellulose in the seven candidate characteristic bands are shown in Table 3:

[0172] Table 3

[0173]

[0174] As can be seen from Table 3, when the wavelength range is in the range of 1400-2400nm (the third band), the root mean square error and the mean absolute error both reach the lowest value, that is, the component content prediction model has the best prediction effect for cellulose in the third band. Compared with the full-band data (800-2400nm), the collection efficiency of the third band spectrum data can be improved by about 37.5% (wavelength range reduction rate);

[0175] Preferably, if Figure 4 As shown, the PCR correlation of the component content prediction model can also be statistically analyzed to test the slope (0.99) and Pearson correlation coefficient (0.96) of the linear fitting equation between the predicted value and the test value (i.e., the mean value of the hemicellulose content detected by wet chemical analysis) of the 3rd band spectral data, further indicating that the internal cross-validation effect of the 3rd band spectral data is good.

[0176] Similarly, the root mean square error and mean absolute error of the prediction results of the statistical component content prediction model for lignin in the six candidate characteristic bands are shown in Table 4:

[0177] Table 4

[0178]

[0179] As can be seen from Table 4, when the wavelength range is in the range of 1000-2400nm (band 1), the root mean square error reaches the lowest value, that is, the component content prediction model has the best prediction effect on cellulose in band 1. Compared with the full-band data (800-2400nm), the collection efficiency of the spectral data in band 1 can be improved by about 25% (wavelength range reduction rate);

[0180] Preferably, if Figure 5 As shown in the figure, the PCR correlation of the component content prediction model can also be statistically analyzed to test the slope (0.88) and Pearson correlation coefficient (0.93) of the linear fitting equation between the predicted value and the test value (i.e., the mean lignin content detected by wet chemical analysis) under the spectral data of the first band, which further illustrates that the internal cross-validation effect of the spectral data of the first band is good.

[0181] Furthermore, if the full band range of near-infrared spectral data is used, the three target chemical components (cellulose, hemicellulose and lignin) are regressed through the principal component regression model and the partial least squares regression model, and the statistical comparison of the root mean square error and the mean absolute error is shown in Table 5:

[0182] Table 5

[0183]

[0184] It can be seen from Table 5 that the regression effect of the principal component regression model is better than that of the partial least squares regression model.

[0185] Furthermore, the pretreatment method in step 3 mainly includes:

[0186] Step b1, counting the lowest value of the near-infrared spectrum data obtained from each hemp fiber sample, and then subtracting the lowest value of each spectrum data from each near-infrared spectrum data to obtain an intermediate value; thereby completing the baseline removal operation and eliminating the influence of the baseline on the spectrum data;

[0187] Step b2, in the intermediate value, after calculating the slopes of two adjacent points, the pre-processed near-infrared spectrum data is obtained; thereby completing the first-order derivative, which can avoid problems such as baseline drift, background noise and light scattering, improve the spectral resolution, and enhance the separation degree of overlapping peaks, thereby making the data features more obvious, and greatly eliminating the errors caused by differences in detection position, instrument status, and test environment, which is beneficial to the prediction accuracy and stability of subsequent modeling; specifically, Figure 6 As shown:

[0188] Among them, Figure a represents cellulose before pretreatment; Figure b represents cellulose after pretreatment, and the ordinate is a dimensionless value; conventional hemp fiber chemical treatment methods can be used to obtain hemicellulose and lignin-removed hemp fibers and cellulose alone; from the comparison of these two figures, it can be seen that due to the difference in cellulose content, the spectrum peaks before and after pretreatment have obvious differences, and are mainly concentrated in the area with larger wavelengths;

[0189] Figure c shows the hemicellulose before pretreatment; Figure d shows the hemicellulose after pretreatment, and the ordinate is a dimensionless value; by using conventional hemp fiber chemical treatment methods and controlling the treatment time, hemp fibers with a small amount, a medium amount, and a large amount of hemicellulose removed can be obtained respectively; from the comparison of these two figures, it can be seen that due to the difference in hemicellulose content, the spectrum peaks before and after pretreatment also show obvious differences;

[0190] Figure e represents the state before lignin pretreatment; Figure f represents the state after lignin pretreatment, and the vertical axis is a dimensionless value; by using conventional hemp fiber chemical treatment methods and controlling the treatment time, hemp fibers with a small amount, a medium amount, and a large amount of lignin removed can be obtained respectively; by comparing the two figures, it can be seen that due to the difference in lignin content, the spectral peaks also show obvious differences before and after pretreatment.

[0191] Furthermore, if Figure 7 As shown, the training method of the component content prediction model (classification model) includes:

[0192] Step c1, collecting a preset number (i.e., 35) of hemp fiber samples and numbering them; selecting a preset weight (i.e., 1 gram) of hemp fiber from each hemp fiber sample, cutting it into fiber powder of a preset length (i.e., less than 1 mm), drying it to a constant weight at a preset temperature (i.e., 105° C.), and cooling it in a dryer for a preset time (i.e., 30 minutes);

[0193] Step c2, using a spectrophotometer (i.e., Perkin-Elmer UV / VIS / NIR 750) to collect near-infrared spectra of the fiber powder by diffuse reflection in a preset spectral band range;

[0194] Preferably, step c2 specifically comprises: calibrating the spectrophotometer by two standard white boards; crushing the fiber powder to a preset particle size (i.e., 60 meshes), placing it in a near-infrared absorption sample pool of the spectrophotometer, and performing multiple sample loading and diffuse reflection scanning (i.e., repeating the sample loading twice, and scanning twice each time) at a preset scanning interval (i.e., 6 nm), taking the average spectral value as the spectral curve of each hemp fiber sample (each spectral curve contains 267 absorbance data); thereby obtaining a first data matrix of 267 rows and 35 columns;

[0195] Step c3, preprocessing the spectrum curve;

[0196] Step c4, sampling the hemp fiber samples according to the numbers, and performing wet quantitative analysis according to GB / T5889-1986 "Quantitative Analysis Method for Chemical Composition of Ramie" to obtain test values ​​of the contents of hemicellulose, lignin and cellulose components in the hemp fiber samples; for each hemp fiber sample, marking the test value according to the number; thereby obtaining a second data matrix of 3 rows and 35 columns, and merging the first data matrix with the second data matrix to obtain a third data matrix of 270 rows and 35 columns;

[0197] Step c5, according to the test value, the hemp fiber samples are divided into a calibration set (i.e., 30 hemp fiber samples) and a prediction set (i.e., 5 hemp fiber samples) in a preset ratio (i.e., 6:1); the test value range of the calibration set is larger than the test value range of the prediction set; the calibration set is used to train the component content prediction model; the prediction set is used to verify the component content prediction model; this is conducive to limiting the predicted component within the calibration range to avoid interference from the superposition effect of the near-infrared spectrum peak information.

[0198] Furthermore, if Figure 8 As shown in Figure 2, the construction process of the tensile strength prediction model includes:

[0199] Step d1, based on the standard test method for tensile properties (e.g., the standard test method for tensile properties of single textile fibers, i.e., ASTM D3822 / D3822M-14 Standard Test Method for Tensile Properties of Single Textile Fibers), measuring the tensile strength of the hemp fiber sample; using the content of the target chemical component as input data and the tensile strength as output data to construct a sample set;

[0200] Step d2: normalize the input data and output data. The specific formula can be:

[0201]

[0202] in, represents the normalized value of the i-th target chemical component in the sample with the j-th tensile strength; X ij represents the i-th target chemical composition in the sample with the j-th tensile strength; x imin represents the i-th target chemical composition in the sample with the minimum tensile strength; x imax represents the i-th target chemical composition in the sample with the maximum tensile strength; represents the normalized value of the j-th tensile strength; y j represents the jth tensile strength; y min Indicates the minimum tensile strength; y max Indicates the maximum tensile strength;

[0203] Step d3, constructing a support vector machine: the input layer is the content of the target chemical component; the middle layer is to select the kernel function by learning the sample set; the last layer is to construct a nonlinear prediction function.

[0204] Preferably, the specific steps include:

[0205] Step d31: Based on the sample set and any given ε>0, the ε-linear programming problem is transformed into an optimization problem, and then:

[0206]

[0207] Wherein, w represents the normal vector of the hyperplane; N represents the number of hemp fiber samples;

[0208] Step d32: Find a kernel function K(·,·) that satisfies the condition

[0209] in, Represents the mapping from the input space to a feature space;

[0210] Step d33: Introduce slack variable ξi , and i=1, 2, 3, ..., N, and the optimization model is established through the Lagrange multiplier method:

[0211]

[0212] Where c represents the penalty parameter; b represents the optimization model constant;

[0213] According to the nonlinear support vector machine theory, the dual problem of the optimization model is obtained, and the specific formula is:

[0214]

[0215] in, represents the optimal solution for the i-th target chemical composition; α i represents the general solution of the i-th target chemical composition; represents the optimal solution for the jth target chemical composition; α j represents the optimal solution of the jth target chemical composition;

[0216] Then solve to get the support vector of the i-th target chemical component and the conjugate vector of the support vector of the i-th target chemical component

[0217] Step d34: Calculate the optimization model constant. The specific formula may be:

[0218]

[0219] Where N' represents the number of target chemical components; represents the support vector of the jth target chemical component; express The conjugate vector of ; Support vectors representing constants of the optimization model;

[0220] Step d35: construct a nonlinear prediction function for prediction and simulation. The specific formula may be:

[0221]

[0222] Among them, x i represents the i-th chemical component, and x i ∈R 6 ; b represents the optimization model constant, and b∈R; R represents a real number.

[0223] Embodiment 2:

[0224] like Fig. 9As shown, this embodiment provides a hemp fiber tensile strength prediction system, including a data receiving module, a data processing module and a result generating module;

[0225] The data receiving module is used to receive near-infrared spectrum data of characteristic bands of target chemical components in the tested hemp fiber;

[0226] The data processing module includes a data analysis unit, a component content prediction unit and a tensile strength prediction unit;

[0227] The data analysis unit determines the target chemical composition of the hemp fiber through a data analysis method based on the chemical composition content, near infrared spectrum data and raw data of tensile strength of the hemp fiber; the data analysis method includes principal component analysis, cluster analysis and correlation analysis in sequence;

[0228] The principal component analysis is used to preliminarily determine the range of the target chemical components;

[0229] The cluster analysis is used to further narrow the scope of target chemical components;

[0230] The correlation analysis is used to determine the degree of correlation between the target chemical composition and the tensile strength;

[0231] The component content prediction unit is used to obtain component content data of target chemical components in the tested hemp fiber by component content prediction after preprocessing the near infrared spectrum data;

[0232] The tensile strength prediction unit is used to obtain the tensile strength of the tested hemp fiber by using the component content data through tensile strength prediction;

[0233] The result generation module is used to externalize the target chemical composition and tensile strength.

[0234] Embodiment three:

[0235] The present embodiment provides a hemp fiber tensile strength prediction device, comprising a processor, a memory and a bus, wherein the memory stores instructions and data readable by the processor, the processor is used to call the instructions and data in the memory to execute the hemp fiber tensile strength prediction method as described above, and the bus connects the functional components to transmit information.

[0236] In another embodiment, the present solution can be implemented by an integrated device, which may include corresponding modules for performing each or several steps in the above-mentioned embodiments. The module may be one or more hardware modules specially configured to perform the corresponding steps, or implemented by a processor configured to perform the corresponding steps, or stored in a computer-readable medium for implementation by a processor, or implemented by some combination.

[0237] The processor performs the various methods and processes described above. For example, the method implementation in the present solution can be implemented as a software program, which is tangibly contained in a machine-readable medium, such as a memory. In some embodiments, part or all of the software program can be loaded and / or installed via a memory and / or a communication interface. When the software program is loaded into the memory and executed by the processor, one or more steps in the method described above can be executed. Alternatively, in other embodiments, the processor can be configured to perform one of the above methods by any other appropriate means (for example, by means of firmware).

[0238] The device can be implemented using a bus architecture. The bus architecture can include any number of interconnecting buses and bridges, depending on the specific application and overall design constraints of the hardware. The bus connects various circuits including one or more processors, memories, and / or hardware modules together. The bus can also connect various other circuits such as peripherals, voltage regulators, power management circuits, external antennas, etc.

[0239] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0240] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting tensile strength of hemp fiber, characterized in that: include: Step 1: Based on the chemical component content, near infrared spectrum data and raw data of tensile strength of hemp fiber, the target chemical component is determined by a data analysis method; the data analysis method includes principal component analysis, cluster analysis and correlation analysis in sequence; The principal component analysis is used to preliminarily determine the range of the target chemical components; The cluster analysis is used to further narrow the scope of target chemical components; The correlation analysis is used to determine the degree of correlation between the target chemical composition and the tensile strength; The target chemical components include cellulose and / or hemicellulose and / or lignin; Step 2: Based on the characteristic bands of the target chemical components, collect near infrared spectrum data of the tested hemp fiber; Step 3, after preprocessing the near infrared spectrum data, the component content data of the target chemical component in the tested hemp fiber is obtained by component content prediction; Step 4: The component content data is used to predict the tensile strength to obtain the tensile strength of the tested hemp fiber.

2. The method according to claim 1, characterized in that The specific method of principal component analysis includes: establishing a principal component regression model; taking all chemical components of hemp fiber samples as original variables; calculating eigenvalues ​​and eigenvectors through data standardization processing and correlation coefficient matrix; determining principal component factors through variance contribution rate and cumulative variance contribution rate; taking root mean square error and mean absolute error as evaluation indicators to evaluate the regression effect of the predicted value of the principal component regression model on the test value obtained by wet quantitative analysis; performing principal component regression analysis on hemp fiber samples, calculating variance contribution rate and cumulative variance contribution rate and sorting them, and screening out several principal component factors when the variance cumulative contribution rate reaches a preset variance cumulative contribution rate; the principal component factors are linear combinations of target chemical components.

3. The method according to claim 1, characterized in that: The specific process of the correlation analysis includes: Step a1: Take the tensile strength as the dependent variable series and the target chemical component content as the independent variable series to form the original data matrix W1. The specific expression is: W1=|X1 X2 … X i-1 Y| Among them, X i-1 represents the content of the i-1th target chemical component; Y represents the tensile strength; Normalize the data to get the normalized matrix W2, the specific expression is: W2=(in ki ); Among them, w ki It represents the normalized value of the i-th target chemical component of the k-th hemp fiber sample. The specific expression is: Among them, mxxx ki represents the maximum eigenvalue of the i-th target chemical component of the k-th hemp fiber sample; minx ki represents the minimum eigenvalue of the i-th target chemical component of the k-th hemp fiber sample; Step a2, calculate the correlation coefficient; the correlation coefficient is the correlation degree value between the independent variable sequence and the dependent variable sequence at each moment; the larger the correlation coefficient, the greater the influence of the independent variable at that moment; the specific calculation formula of the correlation coefficient between each independent variable sequence and the dependent variable sequence is: Among them, ζi(k) represents the correlation coefficient; ζ represents the resolution coefficient, and ζ∈[0,1]; min(Δi(min)) represents the minimum difference between the two levels; max(Δi(max)) represents the maximum difference between the two levels. The specific calculation formula is: Step a3: Calculate the correlation between the independent variable series and the dependent variable series; then calculate the average correlation r i , used to indicate the degree of association between the independent variable series and the dependent variable series; the average correlation r i The larger the value is, the greater the influence of the corresponding independent variable is. The specific formula is: Wherein, N represents the number of hemp fiber samples.

4. The method according to claim 1, characterized in that: The characteristic bands specifically include: The data points in the 1600-2400 nm band correspond to the prediction of cellulose component content; The data points in the 1400-2400 nm band correspond to the predicted content of hemicellulose components; The data points in the 1200-2400 nm band correspond to the predicted lignin content.

5. The method according to claim 1, characterized in that The method for determining the characteristic band is specifically as follows: within the full band range of near-infrared spectral data, a preset band interval length is used as a tolerance, and several alternative characteristic bands are selected in an arithmetic progression from the short-wave end to the long-wave end, and a cyclic test is performed through component content prediction, and the root mean square error and mean absolute error of the target chemical component content are calculated, and the alternative characteristic band with the lowest root mean square error is used as the characteristic band of the target chemical component.

6. The method according to claim 1, characterized in that The method for determining the characteristic band is specifically: In the full band range of near-infrared spectral data, based on the initial starting interval length LC as the moving window, the preset scanning accuracy s is used as the step size, and the candidate characteristic bands are iteratively selected from the short-wave end to the long-wave end with equal distance. The specific calculation formula is: LT = LC + (T-1) * s; Among them, LT represents the length of the starting interval of the Tth generation; T represents the number of iterations; Until the starting interval length equals the preset limit; All candidate characteristic bands of each generation are tested through component content prediction, and the root mean square error and mean absolute error of the target chemical component content are calculated. The alternative characteristic band with the lowest root mean square error is used as the characteristic band of the target chemical component.

7. The method according to claim 1, characterized in that The pretreatment method in step 3 comprises: Step b1, counting the lowest value of the near infrared spectrum data obtained from each hemp fiber sample, and then subtracting the lowest value of each spectrum data from each near infrared spectrum data to obtain an intermediate value; Step b2: in the intermediate value, after calculating the slopes of two adjacent points, the pre-processed near infrared spectrum data is obtained.

8. The method according to claim 1, characterized in that The specific method for predicting the tensile strength includes: Step d1, based on the tensile performance standard test method, measuring the tensile strength of the hemp fiber sample; taking the content of the target chemical component as input data and the tensile strength as output data to construct a sample set; Step d2: normalize the input data and output data. The specific formula is: in, represents the normalized value of the i-th target chemical component in the sample with the j-th tensile strength; X ij represents the i-th target chemical composition in the sample with the j-th tensile strength; x imin represents the i-th target chemical composition in the sample with the minimum tensile strength; x imax represents the i-th target chemical composition in the sample with the maximum tensile strength; represents the normalized value of the j-th tensile strength; y j represents the jth tensile strength; y min Indicates the minimum tensile strength; y max Indicates the maximum tensile strength; Step d3, constructing a support vector machine: the input layer is the content of the target chemical component; the middle layer is to select the kernel function by learning the sample set; the last layer is to construct a nonlinear prediction function.

9. A hemp fiber tensile strength prediction system, characterized in that: It includes a data receiving module, a data processing module and a result generating module; The data receiving module is used to receive near-infrared spectrum data of characteristic bands of target chemical components in the tested hemp fiber; The data processing module includes a data analysis unit, a component content prediction unit and a tensile strength prediction unit; The data analysis unit determines the target chemical composition of the hemp fiber through a data analysis method based on the chemical composition content of the hemp fiber, the near infrared spectrum data and the original data of the tensile strength; the data analysis method includes principal component analysis, cluster analysis and correlation analysis in sequence; The principal component analysis is used to preliminarily determine the range of the target chemical components; The cluster analysis is used to further narrow the scope of target chemical components; The correlation analysis is used to determine the degree of correlation between the target chemical composition and the tensile strength; The component content prediction unit is used to obtain component content data of target chemical components in the tested hemp fiber by component content prediction after preprocessing the near infrared spectrum data; The tensile strength prediction unit is used to obtain the tensile strength of the tested hemp fiber by using the component content data through tensile strength prediction; The result generation module is used to externalize the target chemical composition and tensile strength.

10. A hemp fiber tensile strength prediction device, characterized in that: It includes a processor, a memory and a bus, wherein the memory stores instructions and data read by the processor, the processor is used to call the instructions and data in the memory to execute any method as claimed in claim 1-8, and the bus connects the functional components for transmitting information.

Images