Method and system for analyzing components of complex systems in multiple scenarios

Abstract

The application provides a method and system for complex system component analysis under multiple scenes, the method comprising: optimizing linear model coefficients of multivariate analytical chemistry signals involving multiple scenes according to an objective function and corresponding parameter-free constraint conditions, to obtain a model for each analysis scene; and predicting the content of the target composition under multiple scenes according to a new target analysis model. The system executes the method. The application establishes a linear model for multivariate analytical chemistry data collected by an analytical instrument under each scene and system composition, simultaneously establishes a corresponding linear model for each terminal by combining the loss function and the constraint condition mentioned, and realizes component analysis of a new measurement system under different detection scenes based on the optimized target analysis model of each terminal, so that the accuracy of the prediction result is improved compared with a traditional multivariate model.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for analyzing the material composition of complex systems in multiple scenarios. Background Technology

[0002] With the advent of high-throughput, large-scale analytical instruments, modeling multivariate analytical chemistry data has become a focus of attention, especially when the research system is complex or when the spectra of multiple pure components within the system overlap significantly. For example, when using near-infrared spectroscopy to characterize complex aqueous systems, the broad and overlapping peaks of the water molecule spectrum can mask the signals of other components. Multivariate calibration methods can effectively integrate multivariate data for qualitative and quantitative analysis of the components of interest.

[0003] In practical applications, factors such as the detection environment, sample morphology, and instrument status can all cause systematic changes in the spectrum, rendering the established model unusable. A common approach is to use model transfer methods, such as multi-level simultaneous component analysis (MSCA). These methods require acquiring corresponding standard spectra of the same batch of standard samples under different conditions, thus limiting their use to a small range where standard spectra can be acquired under all conditions. Furthermore, these methods often require adjusting the model transfer parameters according to specific circumstances, making the process quite cumbersome.

[0004] Based on this, existing technologies have proposed a parameter-free framework for calibration enhancement (PFCE) based on relevant constraints. PFCE enhances a pre-built model for a specific scenario (the host model) by minimizing the sum of spectral prediction residuals under two conditions, combined with the constraint that the correlation coefficient of the model parameters is not less than a specific threshold, to adapt to a new detection scenario (the optimized model becomes the slave model). This method can achieve good model transfer without complex parameter optimization. When standard spectra corresponding to different scenarios can be collected, unsupervised PFCE (NS-PFCE) can be used; when standard spectra cannot be collected from the host, semi-supervised PFCE (SS-PFCE) can also achieve good results; when standard spectra can be collected and the reference values ​​of standard samples are known, full-supervised PFCE (FS-PFCE) can be used, which not only achieves model transfer but also further enhances the model's prediction performance.

[0005] PFCE solves the problem of cumbersome parameter optimization in model transfer and can select the corresponding scheme for different known conditions (detection scenarios), solving the problem of model prediction effect degradation caused by changes in external detection conditions in practical applications. However, PFCE has the following drawbacks: (1) The correlation constraint in PFCE is weak. When the number of standard spectra is small, the model enhancement effect is not robust; (2) When the submachine involves multiple detection scenarios, it is difficult to use a general model transfer method to model all detection scenarios at the same time. Summary of the Invention

[0006] The present invention provides a method and system for analyzing the material composition of complex systems under multiple scenarios, which solves the problems of existing technologies that can only optimize the model of the slave machine but cannot optimize the model of the host machine. Furthermore, when the slave machine involves multiple detection scenarios, it is difficult to use a general model transfer method to transfer to all detection scenarios, resulting in low computational efficiency and low data utilization.

[0007] This invention provides a method for analyzing the material composition of complex systems under multiple scenarios, comprising:

[0008] A linear model is established based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0009] Based on the loss function and constraints of the linear model, the linear model is optimized to obtain the target analysis model;

[0010] Based on the target analysis model, the new analytical chemistry data collected by the target terminal is analyzed to obtain the predicted component values ​​of the target substance.

[0011] According to the present invention, a method for analyzing the material composition of complex systems under multiple scenarios includes obtaining the loss function and constraints of the linear model in the following ways:

[0012] The loss function is determined based on the number of tasks in the target terminal, the analytical chemical data of the target substance collected by each terminal, the component reference value of the target substance, and the model coefficients of the linear model. The number of tasks is the number of tasks that collect the analytical chemical data of the target substance, and the model coefficients include an intercept term and a coefficient term.

[0013] The constraint conditions are determined based on the normalization constraint of the model coefficients between any two terminals, the number of tasks, and the preset threshold.

[0014] According to a method for analyzing the material composition of complex systems under multiple scenarios provided by the present invention, the method for obtaining the normalization constraint of the model coefficients between any two terminals includes:

[0015] The normalization constraint is determined based on the coefficient terms of the linear model corresponding to any two terminals and the difference between the coefficient terms of the linear model corresponding to any two terminals; or

[0016] The normalization constraint is determined based on the absolute value of the difference and the absolute value of the coefficients of the linear model corresponding to any two terminals.

[0017] According to the present invention, a method for analyzing the material composition of complex systems under multiple scenarios includes optimizing the linear model based on the loss function and constraints of the linear model to obtain a target analysis model, comprising:

[0018] Based on the loss function of the linear model and the constraints, the intercept term and coefficient term of the linear model are optimized to obtain the target intercept term and target coefficient term corresponding to the linear model.

[0019] The target analysis model is obtained by replacing the intercept term with the target intercept term and replacing the coefficient term with the target coefficient term.

[0020] According to the present invention, a method for analyzing the composition of a complex system under multiple scenarios includes analyzing new analytical chemical data collected by the target terminal based on the target analysis model to obtain the predicted composition value of the target substance, comprising:

[0021] The newly collected analytical chemical data of the target substance are input into the target analysis model to obtain the predicted values ​​of the target substance's components.

[0022] According to the present invention, a method for analyzing the composition of complex systems under multiple scenarios is provided, wherein the analytical chemical data includes any one of the following:

[0023] The infrared absorption spectrum, mass spectrum, nuclear magnetic resonance spectrum and ultraviolet absorption spectrum of the target substance.

[0024] This invention also provides a system for analyzing the material composition of complex systems in multiple scenarios, including: a model building module, a model optimization module, and a composition prediction module;

[0025] The model building module is used to build a linear model based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0026] The model optimization module is used to optimize the linear model according to the loss function and constraints of the linear model to obtain the target analysis model;

[0027] The component prediction module is used to analyze the new analytical chemistry data collected by the target terminal according to the target analysis model, and obtain the component prediction value of the target substance.

[0028] The present invention also provides an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the method for analyzing the material composition of complex systems in multiple scenarios as described above.

[0029] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for analyzing the material composition of complex systems in multiple scenarios as described above.

[0030] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for analyzing the material composition of complex systems under multiple scenarios as described above.

[0031] The present invention provides a method and system for analyzing the composition of complex systems under multiple scenarios. It establishes a linear model by combining analytical chemistry data collected by target terminals such as analytical instruments in each scenario with reference values ​​of the composition of the complex system to be tested. The linear model established by each terminal in the target terminal is optimized by combining loss function and constraint conditions. Based on the optimized target analysis model corresponding to each terminal, the composition analysis of the complex system under different detection scenarios is realized. This method improves the accuracy of prediction results compared with traditional multivariate models. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is one of the flowcharts illustrating the method for analyzing the material composition of complex systems in multiple scenarios provided by this invention.

[0034] Figure 2 This is the second flowchart of the method for analyzing the material composition of complex systems in multiple scenarios provided by the present invention;

[0035] Figure 3 This is a schematic diagram of the prediction results of the PFCE model established by combining different constraint thresholds and different constraint conditions provided by the present invention;

[0036] Figure 4 This is a schematic diagram of the prediction results of the PFCE model established using different numbers of standard samples provided by the present invention;

[0037] Figure 5 This is a schematic diagram of the prediction results of the MT-PFCE model with different constraints provided by the present invention;

[0038] Figure 6 This is a schematic diagram of the structure of the system for analyzing the material composition of complex systems in multiple scenarios provided by the present invention;

[0039] Figure 7 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0041] Figure 1 This is one of the flowcharts illustrating the method for analyzing the material composition of complex systems in multiple scenarios provided by this invention, such as... Figure 1 As shown, the method includes:

[0042] Step 110: Based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal, establish a linear model. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0043] Step 120: Optimize the linear model based on the loss function and constraints of the linear model to obtain the target analysis model;

[0044] Step 130: Analyze the new analytical chemistry data collected by the target terminal according to the target analysis model to obtain the predicted component values ​​of the target substance.

[0045] It should be noted that the above method can be implemented by computer equipment.

[0046] Optionally, the method for analyzing the material composition of complex systems under multiple scenarios provided by the present invention is an improvement and innovation based on model transfer using PFCE.

[0047] The target terminal may specifically include multiple terminals, which may be analytical devices or instruments capable of collecting analytical chemical data of the target substance. Each terminal may be able to collect analytical chemical data of the target substance under a detection environment. The target substance may be organic and inorganic substances in a complex system to be detected.

[0048] Furthermore, in one embodiment, the analytical chemistry data includes any of the following:

[0049] The infrared absorption spectrum, mass spectrum, nuclear magnetic resonance spectrum and ultraviolet absorption spectrum of the target substance.

[0050] Optionally, the collected analytical chemistry data of the target substance can specifically be any one of the infrared absorption spectrum, mass spectrometry, nuclear magnetic resonance spectrum and ultraviolet absorption spectrum of the target substance.

[0051] The analytical chemistry data of the target substance can be obtained by collecting any one of the infrared absorption spectrum, mass spectrometry, nuclear magnetic resonance spectrum and ultraviolet absorption spectrum of the sample of the target substance.

[0052] Based on the analytical chemistry data of the target substance collected by each terminal and the reference values ​​of the target substance's components, a correlation relationship is established between the reference values ​​of the target substance's components and the analytical chemistry data of the target substance to obtain a linear model corresponding to each terminal. The reference values ​​of the target substance's components can specifically refer to the actual content of the constituent components in the target substance.

[0053] A uniform loss function and constraints are used to simultaneously optimize the linear model corresponding to each terminal. Based on the optimized linear model, the target analysis model corresponding to each terminal is obtained. It should be noted that optimizing the linear model means optimizing the model coefficients of the linear model. Substituting the optimized model coefficients into the linear model yields the target analysis model.

[0054] By utilizing the target analysis model obtained for each terminal, it is possible to predict the composition of the target substance in the complex system under different detection scenarios, obtain the predicted value of the composition of the target substance, and perform content analysis of the composition of the target material based on the predicted value.

[0055] For example, in some embodiments, a linear model is established by using analytical instruments, such as near-infrared spectrometers, to collect infrared absorption spectra of agricultural products such as apples and reference values ​​of the content of components (e.g., sugar, moisture, etc.) in apples.

[0056] By sampling a uniform loss function and constraints, the linear model in the near-infrared spectrometer is optimized to obtain the target analysis model of the near-infrared spectrometer.

[0057] By using the target analysis model of a near-infrared spectrometer, reference values ​​of components such as sugar and water in apples, and the collected infrared absorption spectra of apples, the predicted values ​​of the content of components such as sugar and water in apples can be obtained.

[0058] It should be noted that the process of model transfer using PFCE is as follows:

[0059] Before model transfer, it is necessary to first model the multivariate analytical chemistry data. Linear models, such as multiple linear models, are a common choice. The relationship between the analytical chemistry data X and the component reference values ​​y of batches of samples of a given target substance can be represented by a multiple linear model. In connection with this, the data can be specifically identified as a detection signal:

[0060]

[0061] Where b0 and b are the intercept and coefficient terms in the linear model coefficients, respectively. e is the prediction residual. The component prediction value is the component reference value y of the target substance.

[0062] Model transfer involves a variety of situations. This article will use the transfer of linear models between different terminals, such as near-infrared spectrometers, as an example. However, without loss of generality, these methods can also be directly applied to model degradation caused by changes in samples, near-infrared spectrometer status, and detection environment. Typically, the near-infrared spectrometer with the established linear model is called the master, denoted by the subscript m; the near-infrared spectrometer to be transferred is called the slave, denoted by the subscript s. Therefore, for model transfer, the intercept term b of the linear model in the master near-infrared spectrometer... 0,m and coefficient term b m It is known that the host's linear model can be established by correlating the collected analytical chemistry data of the target substance with the component reference values ​​through partial least squares regression.

[0063] Unsupervised PFCE (NS-PFCE) is used when both the host and slave devices are available. A batch of representative standard samples corresponding to the target substance are used to acquire analytical chemical data, such as standard spectra, for the standard samples on both the host and slave devices. Let's assume the analytical chemical data of the target substance acquired by the host is X. m The analytical chemical data of the target substance collected by the ionizer is X. s Based on the target loss function shown in Equation (2) The target constraint condition constr(b) shown in formula (3) s ,b m )>r th This is obtained by optimizing the linear model of the slave machine, which is derived by transferring the linear model of the host machine using a model transfer method.

[0064]

[0065] st constr(b s ,b m ) > r th (3)

[0066] Among them, b 0,s To obtain the intercept term of the linear model obtained by transferring the linear model from the host machine to the slave machine, b s To transfer the linear model from the host machine to the slave machine, the coefficients of the resulting linear model are r. th For the preset threshold, the constraint condition of NS-PFCE (as shown in Equation 4) is the correlation coefficient Corr(b) between the host model and the slave model. s ,b m () greater than a specific preset threshold r th(Typically 0.98).

[0067]

[0068] Semi-supervised PFCE (SS-PFCE) is suitable when the host computer is unavailable. It involves acquiring analytical chemical data (such as standard spectra X) of the target substance, such as a batch of representative samples with reference values ​​y, on a slave device. s The target loss function in formula (5) and the target constraint condition constr(b) in formula (6) s ,b m )>r th Optimization was performed to obtain the model of the transferred slave machine:

[0069]

[0070] st constr(b s ,b m ) > r th (6)

[0071] The target loss function is optimized using a global optimization algorithm. Optimization will yield the coefficient term b of the model in the transferred sub-machine. s and intercept term b 0,s .

[0072] Full-supervised PFCE (FS-PFCE) is a model augmentation algorithm that combines NS-PFCE and SS-PFCE. Specifically, it involves acquiring data of the target object, such as standard spectra (X-rays), from both the host and slave machines. m and X s When the corresponding component reference value y is given, the model of the submachine is solved by formula (7) and formula (8).

[0073]

[0074] st constr(b s ,b m ) > r th (8)

[0075] The correlation coefficient Corr(b) s ,b m Substitute NS-PFCE, SS-PFCE, and the corresponding constr(b) into the input. s ,b m Then the corresponding target constraints can be obtained.

[0076] The present invention provides a method for analyzing the composition of complex systems under multiple scenarios. It establishes a linear model by combining analytical chemistry data collected by target terminals such as analytical instruments in each scenario with reference values ​​of the composition of the complex system to be tested. The linear model established by each terminal in the target terminal is optimized by combining loss function and constraint conditions. Based on the optimized target analysis model corresponding to each terminal, the composition analysis of the complex system under different detection scenarios is realized. This method improves the accuracy of prediction results compared with traditional multivariate models.

[0077] Furthermore, in one embodiment, the method for obtaining the loss function and constraints of the linear model may specifically include:

[0078] The loss function is determined based on the number of tasks in the target terminal, the analytical chemical data of the target substance collected by each terminal, the component reference value of the target substance, and the model coefficients of the linear model. The number of tasks is the number of tasks that collect the analytical chemical data of the target substance, and the model coefficients include an intercept term and a coefficient term.

[0079] The constraint conditions are determined based on the normalization constraint of the model coefficients between any two terminals, the number of tasks, and the preset threshold.

[0080] Furthermore, in one embodiment, the method for obtaining the normalization constraint of the model coefficients between any two terminals may specifically include:

[0081] The normalization constraint is determined based on the coefficient terms of the linear model corresponding to any two terminals and the difference between the coefficient terms of the linear model corresponding to any two terminals; or

[0082] The normalization constraint is determined based on the absolute value of the difference and the absolute value of the coefficients of the linear model corresponding to any two terminals.

[0083] Optionally, PFCE employs relevant constraints (constr(b) s ,b m )>r th PFCE exhibits stable performance across different datasets, thus allowing for a unified preset threshold to limit its performance, making it a method that does not require hyperparameter optimization. This is significant for practical applications and large-scale deployments. However, related constraints can lead to the model on the terminal (as a slave machine) not being subject to scaling limitations, resulting in optimization failures on the slave machine's terminal model. To address these issues, this invention proposes normalization constraints.

[0084] The number of tasks in the target terminal can be specifically obtained by counting the number of tasks in each terminal. These tasks can be specifically the task of collecting analytical chemical data of the target substance.

[0085] The normalization constraints between the linear model of the host and the linear model of the slave can specifically include L2 and L1 constraints. The L2 constraint can be calculated as shown in formula (9), where L2(b s ,b m The value range is 0 to 1. It should be noted that the closer the linear model of the host and the linear model of the slave are, the better L2(b) is considered to be. s ,b m The smaller ) is. To maintain consistency with relevant constraints, let 1-L2(b) be. s ,b m )>r th As a new constraint condition for PFCE.

[0086]

[0087] The L1 constraint can be calculated using the method shown in formula (10), where L1(b s ,b m The value range is 0 to 1. The closer the linear model of the host and the linear model of the slave are, the better L1(b) is considered to be. s ,b m The smaller the value, the better. 1-L1(b) s ,b m )>r th As a new constraint condition for PFCE.

[0088]

[0089] Here, abs represents the absolute value.

[0090] These two normalization constraints retain the advantage of PFCE in that the relevant constraints do not depend on specific samples to adjust the threshold, while overcoming the problem that they cannot constrain model scaling.

[0091] Based on this, under the new constraints of PFCE, this invention proposes a multitask PFCE (MT-PFCE) modeling method. This method links the analytical chemistry data and component reference values ​​of the target substance collected at each terminal using a linear model, and simultaneously optimizes the linear models of multiple terminals. When optimizing the linear models of each terminal simultaneously, the normalization constraint of the model coefficients between any two terminals can be used, combined with a unified loss function, to optimize the linear models of each terminal, as detailed below:

[0092] Based on the number of tasks in the target terminal, the reference values ​​of the target substance's composition, the intercept and coefficient terms of the linear model, and using formula (11), a unified loss function is obtained:

[0093]

[0094] Where N represents the number of tasks in the target terminal, y i X represents the reference value of the terminal corresponding to the i-th task. i b represents the analytical chemistry data of the target substance collected by the terminal corresponding to the i-th task. 0,i b represents the intercept term of the linear model in the terminal corresponding to the i-th task. i This represents the coefficient term of the linear model in the terminal corresponding to the i-th task.

[0095] Based on the normalized constraint value between any two terminals, the number of tasks, and the preset threshold, and based on formula (12), a unified constraint condition is obtained, namely the normalized constraint condition:

[0096]

[0097] Among them, constr(b i ,b j ) represents the normalized constraints for the terminal corresponding to the i-th task and the terminal corresponding to the j-th task, including the L2 and L1 constraints for the terminal corresponding to the i-th task and the terminal corresponding to the j-th task (the calculation method is the same as the L2 and L1 constraints between the host and the slave).

[0098] The method for analyzing the composition of complex systems in multiple scenarios provided by this invention uses two normalization constraints to optimize the linear model of each terminal. It retains the advantage of PFCE in that the relevant constraints do not depend on specific samples to adjust the threshold, while overcoming the problem that it cannot constrain model scaling, thereby improving the accuracy of predicting the composition of the target substance.

[0099] Further, in one embodiment, optimizing the linear model based on its loss function and constraints to obtain the target analysis model includes:

[0100] Based on the loss function of the linear model and the constraints, the intercept term and coefficient term of the linear model are optimized to obtain the target intercept term and target coefficient term corresponding to the linear model.

[0101] The target analysis model is obtained by replacing the intercept term with the target intercept term and replacing the coefficient term with the target coefficient term.

[0102] Optionally, the unified loss function and constraints obtained above can be used to optimize the linear models of each terminal simultaneously. Specifically, the intercept term and coefficient term of the linear model of each terminal can be adjusted to obtain the intercept term and coefficient term of each terminal when the loss function reaches its minimum value as shown in formula (12), and these can be used as the target intercept term and target coefficient term.

[0103] Substituting the target intercept term and target coefficient term of each terminal into their corresponding linear model yields the target analysis model for each terminal.

[0104] Furthermore, in one embodiment, the step of analyzing the new analytical chemistry data collected by the target terminal according to the target analysis model to obtain the predicted component value of the target substance may specifically include:

[0105] The newly collected analytical chemical data of the target substance are input into the target analysis model to obtain the predicted values ​​of the target substance's components.

[0106] Optionally, the collected analytical chemistry data of the target substance can be input into the target analysis model, and the predicted values ​​of the components in the target substance can be obtained by combining the target intercept term and the target coefficient term in the target analysis model.

[0107] For example, Figure 2 This is the second flowchart illustrating the method for analyzing the material composition of complex systems in multiple scenarios provided by this invention. Figure 2 As shown, standard spectra (e.g., infrared absorption spectra X1 to X2) of the sample sets (sample set 1 to sample set n) corresponding to the target substance are collected using various terminals. n ) and the corresponding reference values ​​(y1~y n The linear model corresponding to each terminal is determined, and the linear model corresponding to each terminal is optimized simultaneously based on the MT-PFCE method to obtain the target analysis model (model 1 to model n) corresponding to each terminal. Taking the model transfer of tablets reported in the previous report as an example, the performance of normalized constraints in PFCE is evaluated.

[0108] Figure 3 This is a schematic diagram illustrating the prediction results of the PFCE model established under different constraint thresholds and different constraint conditions, as provided by the present invention. Figure 3 As shown, the root mean square error of prediction (RMSEP) is used to evaluate the model established by PFCE. The results show that under the optimal preset threshold (i.e., Threshold) (r th =0.98), the three constraints perform similarly. However, overall, the normalized constraints determined by L2 constraints and L1 constraints are more robust to parameter changes than the related constraints, especially when rth The difference is even more pronounced at >0.98.

[0109] Figure 4 This is a schematic diagram illustrating the prediction results of the PFCE model established using different numbers of standard samples provided by this invention, as shown in the figure. Figure 4 As shown, 30 standard samples were randomly sampled, and 5, 10, 15, 20 and 25 standard samples were selected for PFCE modeling. The results show that when the sample size is sufficient, the three constraints perform similarly. However, when the sample size is small, the model established by normalization constraint has smaller prediction error and higher prediction accuracy.

[0110] Figure 5 This is a schematic diagram illustrating the prediction results of the MT-PFCE model with different constraints provided by this invention, as shown below. Figure 5 As shown, the performance of MT-PFCE was tested using tablet data. The results show that before correction, the Slave's RMSEP was 5.6699, significantly lower than the Master's RMSEP of 3.3530. After using MT-PFCE, the Slave's RMSEP improved significantly, especially the L1-constrained MT-PFCE model, which achieved an RMSEP of 3.1550. Furthermore, the Master model's prediction performance was also significantly enhanced, with the RMSEP decreasing from 3.3530 to 2.3612. Traditional PFCE and model transfer cannot further improve the prediction performance of the master model.

[0111] The method for analyzing the composition of complex systems under multiple scenarios provided by this invention correlates the data and reference values ​​of target objects collected by each terminal under different detection environments with a linear model. This enables the optimization of the linear model corresponding to each terminal, improving computational efficiency and data utilization. Based on the optimized target analysis model corresponding to each terminal, the method can analyze the composition of target substances under different detection scenarios, thereby improving the accuracy of predicting the composition of target substances.

[0112] The system for analyzing the composition of complex systems under multiple scenarios provided by this invention is described below. The system for analyzing the composition of complex systems under multiple scenarios described below can be referred to in correspondence with the method for analyzing the composition of complex systems under multiple scenarios described above.

[0113] Figure 6 This is a schematic diagram of the structure of the system for analyzing the material composition of complex systems in multiple scenarios provided by the present invention, as shown below. Figure 6 As shown, it includes:

[0114] Model building module 610, model optimization module 611, and component prediction module 612;

[0115] The model building module 610 is used to build a linear model based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0116] The model optimization module 611 is used to optimize the linear model according to the loss function and constraints of the linear model to obtain the target analysis model;

[0117] The component prediction module 612 is used to analyze the new analytical chemistry data collected by the target terminal according to the target analysis model to obtain the component prediction value of the target substance.

[0118] The system for analyzing the composition of complex systems under multiple scenarios provided by this invention establishes a linear model by combining analytical chemistry data collected by target terminals such as analytical instruments in each scenario with reference values ​​of the composition of the complex system to be tested. The linear model established by each terminal in the target terminal is optimized by combining loss function and constraint conditions. Based on the optimized target analysis model corresponding to each terminal, the system can perform composition analysis of the complex system under different detection scenarios. This method improves the accuracy of prediction results compared with traditional multivariate models.

[0119] Figure 7 This is a schematic diagram of the physical structure of an electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include a processor 710, a communication interface 711, a memory 712, and a bus 713. The processor 710, communication interface 711, and memory 712 communicate with each other via the bus 713. The processor 710 can call logical instructions from the memory 712 to execute the following methods:

[0120] A linear model is established based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0121] Based on the loss function and constraints of the linear model, the linear model is optimized to obtain the target analysis model;

[0122] Based on the target analysis model, the new analytical chemistry data collected by the target terminal is analyzed to obtain the predicted component values ​​of the target substance.

[0123] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer power supply (which may be a personal computer, server, or network power supply, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] Furthermore, this invention discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when these instructions are executed by a computer, the computer can perform the methods for analyzing the composition of complex systems in multiple scenarios provided in the above-described method embodiments, for example including:

[0125] A linear model is established based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0126] Based on the loss function and constraints of the linear model, the linear model is optimized to obtain the target analysis model;

[0127] Based on the target analysis model, the new analytical chemistry data collected by the target terminal is analyzed to obtain the predicted component values ​​of the target substance.

[0128] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for analyzing the material composition of complex systems in multiple scenarios provided in the above embodiments, for example including:

[0129] A linear model is established based on the analytical chemical data and component reference values ​​of the target substance in the complex system to be tested collected by the target terminal. The target terminal includes multiple terminals, and each terminal is in a different detection environment when collecting the analytical chemical data of the target substance. The linear model is used to characterize the correlation between the component content of the target substance in the complex system to be tested and the analytical chemical data collected by the target terminal. The component reference value is the actual content of the components of the target substance in the complex system to be tested.

[0130] Based on the loss function and constraints of the linear model, the linear model is optimized to obtain the target analysis model;

[0131] Based on the target analysis model, the new analytical chemistry data collected by the target terminal is analyzed to obtain the predicted component values ​​of the target substance.

[0132] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0133] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer power supply (which may be a personal computer, server, or network power supply, etc.) to execute the methods described in various embodiments or some parts of the embodiments.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing components of a complex system in multiple scenarios, characterized in that, The application relates to a method for establishing a target analysis model for a target terminal, and the method comprises the following steps: collecting analysis chemical data of a target substance in a complex system to be measured by a target terminal, wherein the target terminal comprises a plurality of terminals, and each terminal is in a different detection environment when collecting the analysis chemical data of the target substance; establishing a linear model according to the analysis chemical data of the target substance collected by the target terminal and reference values of components of the target substance, wherein the linear model is used to represent a correlation between component contents of the target substance in the complex system to be measured and the analysis chemical data collected by the target terminal, and the reference values of the components are actual contents of constituent components of the target substance in the complex system to be measured; optimizing the linear model according to a loss function and a constraint condition of the linear model to obtain a target analysis model; and analyzing new analysis chemical data collected by the target terminal according to the target analysis model to obtain a component prediction value of the target substance. The method comprises the following steps: According to the number of tasks of the target terminal, the analysis chemical data of the target substance collected by each terminal, the reference values of the components of the target substance, and the model coefficients of the linear model, the loss function is determined, the number of tasks is the number of tasks of collecting the analysis chemical data of the target substance, and the model coefficients comprise an intercept term and a coefficient term. According to the normalization constraint of the model coefficients between any two terminals, the number of tasks, and a preset threshold, the constraint condition is determined. The method for obtaining the normalization constraint of the model coefficients between any two terminals comprises: According to the coefficient terms of the linear models corresponding to any two terminals and the difference between the coefficient terms of the linear models corresponding to any two terminals, the normalization constraint is determined; or According to the absolute values of the difference between the coefficient terms of the linear models corresponding to any two terminals and the absolute values of the coefficient terms of the linear models corresponding to any two terminals, the normalization constraint is determined.

2. The method for multi-scene complex system material component analysis according to claim 1, characterized in that, The method for optimizing the linear model according to the loss function and the constraint condition of the linear model to obtain a target analysis model comprises: According to the loss function and the constraint condition of the linear model, the intercept term and the coefficient term of the linear model are optimized to obtain a target intercept term and a target coefficient term corresponding to the linear model; The target intercept term is substituted for the intercept term, and the target coefficient term is substituted for the coefficient term to obtain the target analysis model.

3. The method for multi-scene complex system material component analysis according to claim 1, characterized in that, The method for analyzing new analysis chemical data collected by the target terminal according to the target analysis model to obtain a component prediction value of the target substance comprises: The new analysis chemical data of the target substance collected is input into the target analysis model to obtain the component prediction value of the target substance. The analysis chemical data comprises any one of the following:

4. The method for multi-scene complex system material component analysis according to claim 1, characterized in that, The target substance infrared absorption spectrum, mass spectrum, nuclear magnetic resonance spectrum, and ultraviolet absorption spectrum. The application relates to a method for establishing a target analysis model for a target terminal, and the method comprises the following steps: collecting analysis chemical data of a target substance in a complex system to be measured by a target terminal, wherein the target terminal comprises a plurality of terminals, and each terminal is in a different detection environment when collecting the analysis chemical data of the target substance; establishing a linear model according to the analysis chemical data of the target substance collected by the target terminal and reference values of components of the target substance, wherein the linear model is used to represent a correlation between component contents of the target substance in the complex system to be measured and the analysis chemical data collected by the target terminal, and the reference values of the components are actual contents of constituent components of the target substance in the complex system to be measured; optimizing the linear model according to a loss function and a constraint condition of the linear model to obtain a target analysis model; and analyzing new analysis chemical data collected by the target terminal according to the target analysis model to obtain a component prediction value of the target substance.

5. The method for analyzing the composition of a complex system of substances in multiple scenarios according to any one of claims 1-4, characterized in that, The method comprises the following steps: According to the number of tasks of the target terminal, the analysis chemical data of the target substance collected by each terminal, the reference values of the components of the target substance, and the model coefficients of the linear model, the loss function is determined, the number of tasks is the number of tasks of collecting the analysis chemical data of the target substance, and the model coefficients comprise an intercept term and a coefficient term.

6. A system for multi-scenario complex system material component analysis, characterized in that, According to the normalization constraint of the model coefficients between any two terminals, the number of tasks, and a preset threshold, the constraint condition is determined. The method for obtaining the normalization constraint of the model coefficients between any two terminals comprises: According to the coefficient terms of the linear models corresponding to any two terminals and the difference between the coefficient terms of the linear models corresponding to any two terminals, the normalization constraint is determined; or According to the absolute values of the difference between the coefficient terms of the linear models corresponding to any two terminals and the absolute values of the coefficient terms of the linear models corresponding to any two terminals, the normalization constraint is determined. The method for optimizing the linear model according to the loss function and the constraint condition of the linear model to obtain a target analysis model comprises: According to the loss function and the constraint condition of the linear model, the intercept term and the coefficient term of the linear model are optimized to obtain a target intercept term and a target coefficient term corresponding to the linear model; The target intercept term is substituted for the intercept term, and the target coefficient term is substituted for the coefficient term to obtain the target analysis model. The method for analyzing new analysis chemical data collected by the target terminal according to the target analysis model to obtain a component prediction value of the target substance comprises: The new analysis chemical data of the target substance collected is input into the target analysis model to obtain the component prediction value of the target substance. The analysis chemical data comprises any one of the following: The target substance infrared absorption spectrum, mass spectrum, nuclear magnetic resonance spectrum, and ultraviolet absorption spectrum. The application relates to a method for establishing a target analysis model for a target terminal, and the method comprises the following steps: collecting analysis chemical data of a target substance in a complex system to be measured by a target terminal, wherein the target terminal comprises a plurality of terminals, and each terminal is in a different detection environment when collecting the analysis chemical data of the target substance; establishing a linear model according to the analysis chemical data of the target substance collected by the target terminal and reference values of components of the target substance, wherein the linear model is used to represent a correlation between component contents of the target substance in the complex system to be measured and the analysis chemical data collected by the target terminal, and the reference values of the components are actual contents of constituent components of the target substance in the complex system to be measured; optimizing the linear model according to a loss function and a constraint condition of the linear model to obtain a target analysis model; and analyzing new analysis chemical data collected by the target terminal according to the target analysis model to obtain a component prediction value of the target substance. The method comprises the following steps: According to the number of tasks of the target terminal, the analysis chemical data of the target substance collected by each terminal, the reference values of the components of the target substance, and the model coefficients of the linear model, the loss function is determined, the number of tasks is the number of tasks of collecting the analysis chemical data of the target substance, and the model coefficients comprise an intercept term and a coefficient term. According to the normalization constraint of the model coefficients between any two terminals, the number of tasks, and a preset threshold, the constraint condition is determined. The method for obtaining the normalization constraint of the model coefficients between any two terminals comprises: According to the coefficient terms of the linear models corresponding to any two terminals and the difference between the coefficient terms of the linear models corresponding to any two terminals, the normalization constraint is determined; or According to the absolute values of the difference between the coefficient terms of the linear models corresponding to any two terminals and the absolute values of the coefficient terms of the linear models corresponding to any two terminals, the normalization constraint is determined. The method for optimizing the linear model according to the loss function and the constraint condition of the linear model to obtain a target analysis model comprises: According to the loss function and the constraint condition of the linear model, the intercept term and the coefficient term of the linear model are optimized to obtain a target intercept term and a target coefficient term corresponding to the linear model; The target intercept term is substituted for the intercept term, and the target coefficient term is substituted for the coefficient term to obtain the target analysis model. The method for analyzing new analysis chemical data collected by the target terminal according to the target analysis model to obtain a component prediction value of the target substance comprises: The new analysis chemical data of the target substance collected is input into the target analysis model to obtain the component prediction value of the target substance. The analysis chemical data comprises any one of the following: The target substance infrared absorption spectrum, mass spectrum, nuclear magnetic resonance spectrum, and ultraviolet absorption spectrum. The application relates to a method for establishing a target analysis model for a target terminal, and the method comprises the following steps: collecting analysis chemical data of a target substance in a complex system to be measured by a target terminal, wherein the target terminal comprises a plurality of terminals, and each terminal is in a different detection environment when collecting the analysis chemical data of the target substance; establishing a linear model according to the analysis chemical data of the target substance collected by the target terminal and reference values of components of the target substance, wherein the linear model is used to represent a correlation between component contents of the target substance in the complex system to be measured and the analysis chemical data collected by the target terminal, and the reference values of the components are actual contents of constituent components of the target substance in the complex system to The model establishing module is configured to establish a linear model according to analysis chemistry data of a target substance in a complex system to be measured and component reference values, the target terminals including a plurality of terminals, each terminal being in a different detection environment when collecting the analysis chemistry data of the target substance, the linear model being used to represent a correlation between component content of the target substance in the complex system to be measured and the analysis chemistry data collected by the target terminals, the component reference values being actual content of constituent components of the target substance in the complex system to be measured; The model optimizing module is configured to optimize the linear model according to a loss function and a constraint condition of the linear model, and obtain a target analysis model; The component predicting module is configured to analyze new analysis chemistry data collected by the target terminals according to the target analysis model, and obtain a component prediction value of the target substance. The loss function and the constraint condition of the linear model are obtained in the following manner: The loss function is determined according to a number of tasks, analysis chemistry data of the target substance collected by each terminal, component reference values of the target substance, and model coefficients of the linear model, the number of tasks being a number of tasks of collecting the analysis chemistry data of the target substance, the model coefficients including an intercept term and a coefficient term; The constraint condition is determined according to a normalization constraint of model coefficients between any two terminals, the number of tasks, and a preset threshold.

7. An electronic device comprising a processor and a memory having a computer program stored therein, characterized in that The processor executes the computer program to implement the method for analyzing components of a complex system substance in multiple scenarios according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method for analyzing components of a complex system substance in multiple scenarios according to any one of claims 1 to 5.

9. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method for analyzing components of a complex system substance in multiple scenarios according to any one of claims 1 to 5.

Images