A rapid detection system for aqueous solution and water content of latex explosives

Abstract

The present invention relates to the field of explosives detection technology, and more specifically to a system for rapidly detecting aqueous solutions and water content of latex explosives. The system comprises a refractive index detection module, a near-infrared spectroscopy detection module, a production line monitoring module, a data processing module, and a data fusion module. The refractive index detection module is used to collect the refractive index of the aqueous solution, the near-infrared spectroscopy detection module is used to collect the water content of the aqueous solution, and the production line monitoring module is used to collect process parameters. By calculating comprehensive indicators of the aqueous solution, the present invention effectively resolves the interference that may arise between water content, solute concentration, and process parameter changes, making parameter control in the production process more precise. By comparing the comprehensive indicators with indicator thresholds, it is possible to determine whether the production process of the latex explosives deviates from the target range, helping the system to promptly detect production anomalies and prevent accidents or quality problems caused by process fluctuations.

Description

Technical Field

[0001] The invention relates to the technical field of explosive detection, in particular to a system for quickly detecting a latex explosive aqueous solution and its water content. Background Art

[0002] Latex explosives are industrial explosives widely used in mining, civil engineering, and other blasting applications. Due to their excellent mechanical properties, high explosive power, and stable storage, latex explosives have experienced rapid development in recent years. The aqueous solution is a crucial component of latex explosive production. Its primary ingredients include oxidants such as ammonium nitrate and sodium nitrate, as well as auxiliary ingredients such as emulsifiers and stabilizers. The ratio of these components directly impacts the explosive's explosive performance, storage safety, and production stability. The formulation and water content of the aqueous solution are two key parameters in the preparation of latex explosives. The accuracy of the formulation directly determines the emulsification properties and explosive power of the latex explosive. The water content significantly impacts the emulsification process and the stability of the emulsified explosive. Excessively high water content can reduce the viscosity and explosive power of the latex explosive, or even lead to emulsification failure. Excessively low water content can impair emulsification efficiency and increase the risk of crystallization during production. Therefore, accurate detection of the composition and water content of the aqueous solution is of decisive significance for the high-quality production of latex explosives.

[0003] Currently, in the production process of latex explosives, the detection of aqueous solution components and water content is mainly divided into two methods: laboratory analysis and online detection technology. Online detection technology can use sensors to monitor the aqueous solution on the production line in real time. For example, the solution composition and water content can be inferred through parameters such as infrared spectroscopy, near-infrared spectroscopy or conductivity. These technologies have the advantages of fast detection speed and non-destructiveness to samples.

[0004] After searching, Chinese patent number CN202010160788.8 discloses an automatic detection system for the aqueous phase properties of on-site mixed emulsion explosives, including: an aqueous phase solution storage tank, an aqueous phase solution manual valve, an aqueous phase solution solenoid valve, a pH value measuring tank, a pH value measuring sensor and a first liquid level sensor, an aqueous phase solution circulation pump, a crystallization point measurement tank manual valve, a crystallization point measurement tank solenoid valve, a crystallization point measurement tank, a second liquid level sensor, a turbidity sensor, a temperature sensor and an external discharge solenoid valve. The above scheme automatically calculates the mass of water, ammonium nitrate or acid that needs to be supplemented to the aqueous phase solution, thereby providing conditions for realizing fully automatic and unattended production of on-site mixed emulsion explosives aqueous phase solution; at the same time, it can store the pH value and crystallization point data of the aqueous phase solution during the production process for a long time, providing data support for subsequent production management.

[0005] However, during the operation of the existing online detection system for the aqueous phase solution and water content of latex explosives, industrial production is a dynamic process, and the process conditions such as temperature, pressure, flow rate, etc. of the production line change frequently. In particular, the water content monitoring is susceptible to phase changes in the production fluid. For example, the actual solution in production may be significantly different from the laboratory standard sample due to impurities or uneven distribution of emulsion particles. This difference causes errors in the chemometrics-based model, which cannot accurately reflect the composition and water content of the actual solution, making it difficult for the detection sensor and optical components to adapt to the instantaneous working conditions. In addition, during production, it is impossible to judge the comprehensive status of the production line based on the aqueous phase solution and water content of the latex explosives, and thus it is impossible to adapt and regulate the production parameters.

[0006] Therefore, a latex explosive aqueous solution and a water content rapid detection system are proposed to solve the above-mentioned problems. Summary of the Invention

[0007] Technical problems solved

[0008] In view of the above-mentioned shortcomings of the prior art, the present invention provides a system for rapid detection of aqueous solutions and moisture content of latex explosives, which can effectively solve the problem of detection deviation caused by the inability of detection sensors and optical elements in the prior art to adapt to the dynamic changes of process conditions in the production process.

[0009] Technical Solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0011] The present invention provides a system for quickly detecting aqueous solutions and water content of latex explosives, comprising a refractive index detection module, a near-infrared spectrum detection module, a production line monitoring module, a data processing module, and a data fusion module. The refractive index detection module is used to collect the detected refractive index RI of the aqueous solution, the near-infrared spectrum detection module is used to collect the detected water content MC of the aqueous solution, and the production line monitoring module is used to collect process parameters. The system is characterized in that the data processing module calculates a parameter deviation ΔB based on the process parameters and the reference parameters, and constructs a refractive index correction model based on the detected refractive index RI and the parameter deviation ΔB to obtain a corrected concentration c correct and refractive index correction term ΔD; the corrected concentration c correct The refractive index correction term ΔD is returned to the refractive index detection module to update the corrected refractive index RI correct ; Based on the corrected concentration c correct The moisture content correction model is constructed with the deviation parameter ΔB to obtain the moisture content correction term ΔW; the moisture content correction term ΔW is returned to the near-infrared spectrum detection module to update the corrected moisture content MC correct ; The corrected water content MC of the aqueous solution correct , corrected concentration ccorrect The parameter deviation ΔB is input into the pre-trained fusion model in the data fusion module to obtain the comprehensive index Q of the aqueous solution, which is used to evaluate the comprehensive status of the emulsion explosive production line.

[0012] Furthermore, the process parameters include the current temperature T of the production line. actual 、Current pressure P actual and the current flow rate V actual .

[0013] Furthermore, the parameter deviation ΔB includes temperature deviation ΔT, pressure deviation ΔP and flow rate deviation ΔV; the calculation formulas are:

[0014] ΔT=T actual -T ref Where, T ref is the reference temperature;

[0015] ΔP=P actual -P ref Where, P ref is the base pressure;

[0016] ΔV=V actual -V ref Where V ref is the base flow rate.

[0017] Furthermore, the method of constructing the refractive index correction model includes:

[0018] Calculate the corrected concentration c based on the parameter deviation ΔB correct , the calculation formula is:

[0019] c correct =-β T ΔT+β P ΔP+β V ΔV;

[0020] Where -β T ·ΔT is the temperature correction term, β T is the temperature correction coefficient; β P ΔP is the pressure correction term, β P is the pressure correction coefficient; β V ·ΔV is the velocity correction term, β V is the flow rate correction factor;

[0021] The refractive index correction term ΔD is calculated based on the deviation parameter ΔB. The calculation formula is:

[0022] ΔD=-ε T ·ΔT+ε P ΔP+ε V ·ΔV; where εT , ε P and ε V

[0023] are the correction coefficients of temperature deviation ΔT, pressure deviation ΔP and flow velocity deviation ΔV for refractive index, respectively.

[0024] Furthermore, the corrected refractive index RI correct The calculation method is:

[0025] The basic calculation formula for the refractive index detection module is defined as: RI = RI0 + α·c; where RI is the detected refractive index of the aqueous solution; RI0 is the refractive index of the pure solvent; α is the proportional constant; c is the solute concentration; the concentration threshold of the aqueous solution is defined, including the standard concentration threshold TH normal and the limit concentration threshold TH max ;

[0026] Based on the corrected concentration c correct And the refractive index correction term ΔD updates the basic calculation formula:

[0027]

[0028] Where α1 is the linear proportional coefficient, α2 is the quadratic proportional coefficient, e is a natural constant, and λ is the growth rate.

[0029] Furthermore, the method of constructing the moisture content correction model includes:

[0030] Calculate the moisture content correction term ΔW using the following formula:

[0031] ΔW=θ T ·ΔT+θ P ΔP+θ V ΔV+θ C c correct Where θ T ,θ P ,θ V ,θ C are the temperature, pressure, flow rate and concentration influence coefficients respectively.

[0032] Furthermore, the calculated corrected moisture content MC correct The methods include:

[0033] The basic calculation formula for defining the near-infrared spectrum detection module is:

[0034] MC=τ1·A 1450 +τ2·A 1920 Where A 1450 and A 1920are the absorption intensities of the near-infrared spectrum at 1450nm and 1920nm wavelengths, respectively; τ1 and τ2 are the absorption coefficients of the aqueous solution at 1450nm and 1920nm wavelengths, respectively;

[0035] The basic calculation formula is updated based on the moisture content correction term ΔW:

[0036] MC correct =τ1·(A 1450 +ΔW)+τ2·(A 1920 +ΔW).

[0037] Furthermore, the calculation formula of the comprehensive index Q is:

[0038] Q=η(t)·MC correct +δ(t)·c correct +κ(t)·ΔB;

[0039] Where t is the time scale; η(t)·MC correct is the moisture content influencing factor; η(t) is the dynamic weight coefficient of moisture content; δ(t)·c correct is the concentration influencing factor; δ(t) is the concentration dynamic weight coefficient, which indicates the importance of the concentration of the aqueous solution in the current process; κ(t)·ΔB is the process parameter influencing factor; κ(t) is the process parameter dynamic weight coefficient;

[0040] The training method of the fusion model includes:

[0041] Collect the data of the currently produced latex explosives, including: comprehensive index Q, dynamic weight coefficient of moisture content η(t), dynamic weight coefficient of concentration δ(t), dynamic weight coefficient of process parameters κ(t), and corrected moisture content MC correct , corrected concentration c correct and parameter deviation ΔB; clean and normalize the above historical data, and combine the above data to construct a fusion feature vector X d ={(Q,η(t),MC correct ,δ(t),c correct ,κ(t),ΔB) d}; where d is the vector time step;

[0042] Define recursive neural network as the basic structure of the fusion model, which includes input layer, hidden layer and output layer;

[0043] Collect historical data of emulsion explosives of different specifications in the past, including: comprehensive index Q, dynamic weight coefficient of moisture content η(t), dynamic weight coefficient of concentration δ(t), dynamic weight coefficient of process parameters κ(t), and corrected moisture content MC correct , corrected concentration ccorrect and parameter deviation ΔB; clean and normalize the above historical data, and combine the above historical data to construct a historical fusion feature vector s is the number of vector time steps d; the historical fusion feature vector is expanded according to the time step d to form time series data as the training input of the fusion model, and the time series data is labeled according to the vector time step d; the label is the historical comprehensive index of latex explosives

[0044] Initialize the network parameters of the fusion model; the network parameters include the weight matrix W from the input layer to the hidden layer x , the cyclic weight matrix W of the hidden layer h , the bias vector b of the hidden layer h , the weight matrix W from the hidden layer to the output layer o and the bias vector b of the output layer o ;

[0045] Define the loss function of the ensemble model

[0046] Where S is the length of the sequence data; Y true,d The label of the d-th vector time step, that is, the historical comprehensive index of the historical production of latex explosives Y pred,d is the prediction result of the dth vector time step, that is, the comprehensive index Q output by the fusion model;

[0047] The time series data is passed to the input layer, and then passes through the hidden layer and the output layer in sequence to output the predicted comprehensive index Q, and calculate the value of the loss function of the corresponding comprehensive index Q, and backpropagate the error gradient along the output layer to the input layer; according to the error gradient, use the optimization algorithm (such as gradient descent, Adam, etc.) to update the network parameters to reduce the value of the loss function; repeatedly train the time series data until the fusion model converges or reaches the preset number of iterations, which means that the fusion model training is completed.

[0048] Furthermore, the input layer is used to receive the fused feature vector X d As input; the hidden layer consists of y state vectors, and the calculation formula for each state vector is: d =f(W h ·h d-1 +W x ·X d +b h );where h d is the hidden state of the current vector time step d, f(...) is the activation function, W x is the weight matrix from the input layer to the hidden layer, W his the cyclic weight matrix of the hidden layer, b h is the bias vector of the hidden layer;

[0049] The output layer is based on the output of the hidden layer, i.e., the hidden state h of the current vector time step d d , calculate the output result of the fusion model, that is, the comprehensive index Q, and its calculation formula is: Q = η (W O ·h d +b o ); where η(...) is the activation function of the output layer, W o is the weight matrix from the hidden layer to the output layer, b o is the bias vector of the output layer.

[0050] Furthermore, the method of evaluating the comprehensive status of the latex explosive production line includes:

[0051] Setting the index threshold Q based on the production specifications of latex explosives TH , the comprehensive index Q and index threshold Q output by the fusion model TH For comparison, if the comprehensive index Q is less than the index threshold Q TH When the comprehensive status of the production line is normal, if the comprehensive index Q is greater than or equal to the index threshold Q TH When , the overall status of the production line is abnormal.

[0052] Beneficial effects

[0053] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0054] The refractive index correction model of the present invention is calculated by calculating the corrected concentration c correct , and then use the corrected concentration c correct Compare with the concentration threshold and select the corrected refractive index RI correct and the corrected concentration c correct The different linear relationships between the refractive index correction term ΔD are calculated so that the refractive index correction model can be used to calculate the corrected refractive index RI. correct Perform more accurate calculations, increase the scope of application and flexibility of the model, and avoid errors that may be introduced when a single model is forcibly fitted across the entire range;

[0055] In this scheme, the concentration c correctDynamic correction calculation of the detected water content MC based on the deviation parameters can correct the obvious interference of different parameters on the spectral absorption bandwidth and intensity, eliminate noise, and control the deviation to the minimum range under complex conditions, thereby improving the accuracy of the water content detection of aqueous solutions; and enable the test results to adapt to the fluctuations of production parameters, ensuring the stability of the production process, avoiding the problems of waste liquid and repeated testing in traditional detection methods, promoting the development of water content detection technology, and providing an efficient solution for online quality monitoring of solution systems;

[0056] The present invention is also based on the corrected water content MC of the aqueous solution correct , corrected concentration c correct The fusion model is constructed based on the parameter deviation ΔB, and the comprehensive index Q of the aqueous solution is calculated to represent the comprehensive status of the latex explosive production line, the real-time and transparent production status, and effectively solve the possible interference between the water content, solute concentration and process parameter changes, making the parameter control of the production link more accurate, and through the comprehensive index Q and the index threshold Q TH By comparing the data, it can be determined whether the production process of latex explosives deviates from the target range, helping the system to detect production anomalies in a timely manner and prevent accidents or quality problems caused by process fluctuations. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0058] Figure 1 Schematic diagram of the system structure in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. 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 making any creative efforts shall fall within the scope of protection of the present invention.

[0060] The present invention will be further described below with reference to the embodiments.

[0061] Example 1:

[0062] See attached Figure 1This case proposes a rapid detection system for aqueous solutions and water content of latex explosives, including a refractive index detection module, a near-infrared spectrum detection module, a production line monitoring module, a data processing module and a data fusion module; the aqueous solution pipeline of the latex explosive production line is connected to the flow cell, and the aqueous solution will enter the flow cell during the flow process, and a connected detection window is provided on the surface of the flow cell; the refractive index detection module includes a light source and a photodetector installed on the flow cell, the light source is used to emit a light beam of stable intensity to transmit to the flow cell, the photoelectric sensor is used to collect the projected or refracted light beam, detect the change in the refraction angle of the light beam, and convert the optical signal into an electrical signal, calculate the refractive index of the aqueous solution, and compare it with a pre-calibrated solution concentration-refractive index curve to infer the concentration of nitrate, impurity content or other target components in the aqueous solution; the near-infrared spectrum detection module includes a light source and a detector installed on the detection window, the light source emits a light beam that passes through the aqueous solution in the detection window, the transmitted spectrum is received by the detector, and the absorbance of the transmitted light is analyzed to quantitatively reflect the water content in the aqueous solution.

[0063] Specifically, the production line monitoring module is used to collect the process parameters of the latex explosives production line, including the current temperature, pressure, and flow rate of the production line. During the detection process of the aqueous solution, temperature changes will affect the spectral absorption characteristics and refractive index of the aqueous solution. Pressure changes have a slight effect on the density and refractive index of the solution. Flow rate fluctuations may cause changes in the uniformity of material flow in the pipeline, thereby affecting the propagation of light or the stability of detection. Therefore, dynamic correction of refractive index detection and near-infrared spectroscopy detection based on process parameters can improve detection accuracy. Based on abnormal monitoring of process parameters (such as excessively high temperature or excessively low flow rate), the production line can automatically alarm and trigger adjustments to ensure safe and efficient production of the production line. By inputting the process parameters collected by the production line monitoring module into the data processing module, the parameter deviation ΔB between the process parameters and the baseline parameters can be calculated, including:

[0064] ΔT=T actual -T ref ; Where ΔT is the temperature deviation; T actual Current temperature, T ref is the reference temperature;

[0065] ΔP=P actual -P ref ; Where ΔP is the pressure deviation; P actual Current pressure, P ref is the base pressure;

[0066] ΔV=V actual -V ref ; Where ΔV is the flow rate deviation; V actual Current flow rate, V ref is the base flow rate.

[0067] Furthermore, in this solution, the basic calculation formula of the refractive index detection module is:

[0068] RI = RI0 + α·c; where RI is the detected refractive index of the aqueous solution, a unitless value, usually in the range of 1.0 to 2.0; RI0 is the refractive index of the pure solvent, which indicates the refractive index of the aqueous solution without any solute added. It is the benchmark value of the solution's refractive index and is usually obtained through actual experimental measurement. It is a unitless value, usually in the range of 1.0 to 1.6, and is affected by temperature and pressure. For example, the refractive index of water at 20°C is 1.333; α is the proportionality constant, which indicates the effect of solute concentration on the refractive index of the aqueous solution. Its value is related to multiple factors such as solute type, solvent composition, and temperature, and is usually obtained through experimental calibration; c is the solute concentration, which indicates the concentration of nitrate, impurity content, or other target components in the aqueous solution. It is usually calibrated by preparing a standard solution of known concentration from the aqueous solution.

[0069] The data processing module constructs a refractive index correction model based on the detected refractive index RI of the aqueous solution collected by the refractive index detection module and the parameter deviation ΔB, and calculates the refractive index correction term ΔD. The construction method of the refractive index correction model includes:

[0070] Calculate the corrected concentration c based on the parameter deviation ΔB correct , the calculation formula is:

[0071] c correct =-β T ΔT+β P ΔP+β V ΔV;

[0072] Where -β T ·ΔT is the temperature correction term, which represents the effect of temperature change on the aqueous solution, β T β is the temperature correction coefficient, which is obtained by experimentally calibrating the concentration change of the solution at different temperatures. For example, the aqueous solution is heated and controlled to be stable at a series of temperature points (such as 20℃, 25℃, 30℃, etc.), the concentration at different temperatures is measured respectively, and the slope of the concentration change with temperature is fitted to obtain the temperature correction coefficient. The temperature correction term is negative, indicating that the concentration generally decreases when the temperature increases; β P ·ΔP is the pressure correction term, which represents the effect of pressure change on the aqueous solution, β P is the pressure correction coefficient, which is also obtained based on experimental calibration; β V ΔV is the flow rate correction term, which represents the effect of flow rate change on the aqueous solution, β V is the flow rate correction coefficient, which is also obtained based on experimental calibration;

[0073] The concentration of the aqueous solution is corrected and calculated by the parameter deviation ΔB, and the corrected concentration c of the aqueous solution under the current process parameters can be obtained. correct , which can effectively counteract the interference of multi-parameter perturbations on the calculation of the refractive index of the aqueous solution, thereby improving the calculation accuracy of the refractive index of the aqueous solution;

[0074] The refractive index correction term ΔD is calculated based on the deviation parameter ΔB. The calculation formula is:

[0075] ΔD=-ε T ·ΔT+ε P ΔP+ε V ·ΔV; where ε T , ε P and ε V are the correction coefficients of temperature deviation ΔT, pressure deviation ΔP and flow rate deviation ΔV for refractive index, which are obtained based on experimental calibration;

[0076] The corrected concentration c correct The refractive index correction term ΔD is returned to the basic calculation formula of the refractive index detection module, and the basic calculation formula is updated to obtain the corrected refractive index RI of the aqueous solution. correct ; Define the concentration threshold of the aqueous solution, including the standard concentration threshold TH normal and the limit concentration threshold TH max , the updated basic calculation formula is:

[0077]

[0078] Where α1 is the linear proportional coefficient, which represents the corrected concentration c correct Corrected refractive index RI correct The linear effect of α2 is the quadratic proportional coefficient, which means the concentration c after correction correct Corrected refractive index RI correct The quadratic nonlinear effect of ; e is a natural constant, λ is the growth rate;

[0079] Corrected refractive index RI correct and the corrected concentration c correct The corresponding relationship is divided into three stages:

[0080] When the corrected concentration c correct Below the standard concentration threshold TH normal When the corrected refractive index RI correct and the corrected concentration c correct , and the refractive index correction term ΔD are linearly related. In this case, the properties of the aqueous solution are relatively simple, and the nonlinear effect of intermolecular interactions on the refractive index is relatively small.

[0081] When the corrected concentration ccorrect Above the standard concentration threshold TH normal And below the limit concentration threshold TH max When the corrected refractive index RI correct and the corrected concentration c correct , and the refractive index correction term ΔD form a quadratic nonlinear relationship. At this time, the interaction between solute molecules in the aqueous solution (such as molecular aggregation, close arrangement between solute-solvent molecules, etc.) makes the relationship between refractive index and concentration show a quadratic nonlinearity.

[0082] When the corrected concentration c correct Above the limit concentration threshold TH max When , the corrected refractive index and the corrected concentration and refractive index correction term form an exponential nonlinear relationship; at this time, the interaction between the solute molecules in the aqueous solution gradually tends to equilibrium (such as the saturation effect, that is, the effect of increasing concentration on the refractive index change decreases);

[0083] The refractive index correction model calculates the corrected concentration c correct , and then use the corrected concentration c correct Compare with the concentration threshold and select the corrected refractive index RI correct and the corrected concentration c correct The different linear relationships between the refractive index correction term ΔD are calculated so that the refractive index correction model can be used to calculate the corrected refractive index RI. correct Perform more accurate calculations, increase the applicability and flexibility of the model, and avoid errors that may be introduced when a single model is forcibly fitted across the entire range.

[0084] Furthermore, in this solution, the basic calculation formula of the near-infrared spectrum detection module is:

[0085] MC=τ1·A 1450 +τ2·A 1920 ; Wherein, MC is the detected water content of the aqueous solution; A 1450 and A 1920 are the absorption intensities of the near-infrared spectrum at 1450nm and 1920nm, respectively; τ1 and τ2 are the absorption coefficients of the aqueous solution at 1450nm and 1920nm, respectively, indicating the linear response between absorption intensity and water content, usually obtained based on experimental calibration;

[0086] Based on the corrected concentration c correct The moisture content correction model is constructed by combining the deviation parameter ΔB and the moisture content correction term ΔW. The calculation formula is:

[0087] ΔW=θ T ·ΔT+θ P ΔP+θ V ΔV+θC c correct Where θ T ,θ P ,θ V ,θ C They are the influence coefficients of temperature, pressure, flow rate and concentration, obtained by experimental fitting, which indicate the degree of influence of these variables on the absorption intensity;

[0088] Return the moisture content correction term ΔW to the basic calculation formula of the near-infrared spectrum detection module and update it to obtain:

[0089] MC correct =τ1·(A 1450 +ΔW)+τ2·(A 1920 +ΔW);

[0090] Where MC correct is the corrected water content of the aqueous solution;

[0091] After correction, the concentration c correct The dynamic correction calculation of the detected water content MC by using the deviation parameters can correct the obvious interference of different parameters on the spectral absorption bandwidth and intensity, eliminate noise, and control the deviation to the minimum range under complex conditions, thereby improving the accuracy of the water content detection of the aqueous solution; and enable the test results to adapt to the fluctuation of production parameters, ensure the stability of the production process, avoid the problems of waste liquid and repeated detection in traditional detection methods, promote the development of water content detection technology, and provide an efficient solution for online quality monitoring of solution systems.

[0092] It is worth noting that in this case, the data fusion module is based on the corrected water content MC of the aqueous solution. correct , corrected concentration c correct A fusion model is constructed with the parameter deviation ΔB to calculate the comprehensive index Q of the aqueous solution, which is used to represent the comprehensive status of the emulsion explosive production line, the real-time and transparent production status, and effectively solve the possible interference between water content, solute concentration and process parameter changes, making the parameter control of the production link more accurate. The calculation formula of the comprehensive index Q is:

[0093] Q=η(t)·MC correct +δ(t)·c correct +κ(t)·ΔB;

[0094] Where t is the time scale; η(t)·MC correctis the moisture content influencing factor, reflecting the impact of moisture content on the safety and detonation performance of emulsion explosives; η(t) is the dynamic weight coefficient of moisture content, indicating the importance of moisture content in current production. It changes with time scale t and is adjusted in real time according to production demand. For example, if the current batch of products is more sensitive to moisture content, η(t) increases to emphasize the contribution of moisture content. If other factors (such as solute concentration) are more critical, their weight is reduced; δ(t)·c correct is the concentration influencing factor, reflecting the impact of the concentration of the aqueous solution on the production status; δ(t) is the concentration dynamic weight coefficient, indicating the importance of the concentration of the aqueous solution in the current process, which changes with the time scale t and is adjusted in real time according to production needs; κ(t)·ΔB is the process parameter influencing factor, reflecting the degree of system stability fluctuation during the production process; κ(t) is the process parameter dynamic weight coefficient, indicating the importance of the current process parameter fluctuation to the comprehensive index Q, protecting the key production parameters from excessive fluctuations.

[0095] The dynamic weight coefficient of moisture content η(t), the dynamic weight coefficient of concentration δ(t) and the dynamic weight coefficient of process parameters κ(t) are all coefficients that are dynamically adjusted with the time scale t and changes in production conditions to ensure that the focus of the comprehensive indicator Q is consistent with the real-time process requirements.

[0096] The methods for training the constructed fusion model include:

[0097] Collect the data of the currently produced latex explosives, including: comprehensive index Q, dynamic weight coefficient of moisture content η(t), dynamic weight coefficient of concentration δ(t), dynamic weight coefficient of process parameters κ(t), and corrected moisture content MC correct , corrected concentration c correct and parameter deviation ΔB; clean and normalize the above historical data, and combine the above data to construct a fusion feature vector X d ={(Q,η(t),MC correct ,δ(t),c correct ,κ(t),ΔB) d}; where d is the vector time step;

[0098] Define recursive neural network as the basic structure of the fusion model, which includes input layer, hidden layer and output layer;

[0099] The input layer is used to receive the fused feature vector X d As input; the hidden layer consists of y state vectors, and the calculation formula for each state vector is: d =f(W h ·h d-1 +W x ·X d +b h); where h d is the hidden state of the current vector time step d, f(...) is the activation function, W x is the weight matrix from the input layer to the hidden layer, W h is the cyclic weight matrix of the hidden layer, b h is the bias vector of the hidden layer;

[0100] The output layer is based on the output of the hidden layer, i.e. the hidden state h at the current vector time step d. d , calculate the output result of the fusion model, that is, the comprehensive index Q, and its calculation formula is: Q = η (W O ·h d +b o ); where η(...) is the activation function of the output layer, W o is the weight matrix from the hidden layer to the output layer, b o is the bias vector of the output layer;

[0101] Collect historical data of emulsion explosives of different specifications in the past, including: comprehensive index Q, dynamic weight coefficient of moisture content η(t), dynamic weight coefficient of concentration δ(t), dynamic weight coefficient of process parameters κ(t), and corrected moisture content MC correct , corrected concentration c correct and parameter deviation ΔB; clean and normalize the above historical data, and combine the above historical data to construct a historical fusion feature vector s is the number of vector time steps d, that is, the number of samples collected; the historical fusion feature vector is expanded according to the time step d to form time series data as the training input of the fusion model, and the time series data is labeled according to the vector time step d; the label is the historical comprehensive index of latex explosives Including the water content, concentration and process parameters of the aqueous solution of latex explosives;

[0102] Initialize the network parameters of the fusion model; the network parameters include the weight matrix W from the input layer to the hidden layer x , the cyclic weight matrix W of the hidden layer h , the bias vector b of the hidden layer h , the weight matrix W from the hidden layer to the output layer o and the bias vector b of the output layer o ;

[0103] Define the loss function of the ensemble model

[0104] Where S is the length of the sequence data; Y true,d The label of the d-th vector time step, that is, the historical comprehensive index of the historical production of latex explosives Ypred,d is the prediction result of the dth vector time step, that is, the comprehensive index Q output by the fusion model;

[0105] The time series data is passed to the input layer, and then passes through the hidden layer and the output layer in sequence to output the predicted comprehensive index Q, and calculate the value of the loss function of the corresponding comprehensive index Q, and backpropagate the error gradient along the output layer to the input layer; according to the error gradient, use the optimization algorithm (such as gradient descent, Adam, etc.) to update the network parameters to reduce the value of the loss function; repeatedly train the time series data until the fusion model converges (that is, the value of the loss function no longer changes) or reaches the preset number of iterations, which means that the fusion model training is completed.

[0106] Through training of the fusion model, the optimized dynamic weight coefficient of moisture content η(t), the dynamic weight coefficient of concentration δ(t), and the dynamic weight coefficient of process parameters κ(t) can be obtained, so that the comprehensive index Q is more in line with the current production target and is convenient for quickly judging the production status of latex explosives; based on the production specifications of latex explosives, an index threshold Q is set. TH When the fusion model predicts and calculates the comprehensive index Q, the comprehensive index Q and the index threshold Q are combined. TH For comparison, when the comprehensive index Q is less than the index threshold Q TH When the comprehensive index Q is greater than or equal to the index threshold Q TH When , the comprehensive state of the production line is abnormal; based on the comprehensive index Q and the index threshold Q TH It can determine whether the production process of latex explosives deviates from the target range, helping the system to detect production anomalies in a timely manner and prevent accidents or quality problems caused by process fluctuations.

[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A system for rapid detection of aqueous solutions and water content of latex explosives, comprising a refractive index detection module, a near-infrared spectrum detection module, a production line monitoring module, a data processing module, and a data fusion module, wherein the refractive index detection module is used to collect the refractive index (RI) of the aqueous solution, the near-infrared spectrum detection module is used to collect the water content (MC) of the aqueous solution, and the production line monitoring module is used to collect process parameters, characterized in that: The data processing module calculates the parameter deviation ΔB based on the process parameters and the reference parameters, and constructs a refractive index correction model based on the detected refractive index RI and the parameter deviation ΔB to obtain the corrected concentration c correct and refractive index correction term ΔD; the corrected concentration c correct The refractive index correction term ΔD is returned to the refractive index detection module to update the corrected refractive index RI correct ; Based on the corrected concentration c correct The moisture content correction model is constructed with the parameter deviation ΔB to obtain the moisture content correction term ΔW; the moisture content correction term ΔW is returned to the near-infrared spectrum detection module to update the corrected moisture content MC correct ; The corrected water content MC of the aqueous solution correct , corrected concentration c correct The parameter deviation ΔB is input into the pre-trained fusion model in the data fusion module to obtain the comprehensive index Q of the aqueous solution, which is used to evaluate the comprehensive status of the latex explosive production line; The process parameters include the current temperature T of the production line actual 、Current pressure P actual and the current flow rate V actual ; The parameter deviation ΔB includes temperature deviation ΔT, pressure deviation ΔP and flow rate deviation ΔV; the calculation formulas are: ΔT=T actual -T ref Where, T ref is the reference temperature; ΔP=P actual -P ref Where, P ref is the base pressure; ΔV=V actual -V ref Where V ref is the reference flow rate; The method of constructing the refractive index correction model includes: Calculate the corrected concentration c based on the parameter deviation ΔB correct , the calculation formula is: c correct =-β T ·ΔT+β P ·ΔP+β V ·ΔV; Where -β T ·ΔT is the temperature correction term, β T is the temperature correction coefficient; β P ΔP is the pressure correction term, β P is the pressure correction coefficient; β V ·ΔV is the velocity correction term, β V is the flow rate correction factor; The refractive index correction term ΔD is calculated based on the parameter deviation ΔB. The calculation formula is: ΔD=-ε T ·ΔT+ε P ΔP+ε V ·ΔV; where ε T , ε P and ε V are the correction coefficients of temperature deviation ΔT, pressure deviation ΔP and flow rate deviation ΔV to refractive index respectively; The corrected refractive index RI correct The calculation method is: The basic calculation formula for the refractive index detection module is defined as: RI = RI0 + α·c; where RI is the detected refractive index of the aqueous solution; RI0 is the refractive index of the pure solvent; α is the proportional constant; c is the solute concentration; the concentration threshold of the aqueous solution is defined, including the standard concentration threshold TH normal and the limit concentration threshold TH max ; Based on the corrected concentration c correct And the refractive index correction term ΔD updates the basic calculation formula: Where α1 is the linear proportional coefficient, α2 is the quadratic proportional coefficient; e is the natural constant, and λ is the growth rate; The method of constructing the moisture content correction model includes: Calculate the moisture content correction term ΔW using the following formula: ΔW=θ T ·ΔT+θ P ·ΔP+θ V ΔV+θ C c correct Where θ T ,θ P ,θ V ,θ C are the temperature, pressure, flow rate, and concentration influence coefficients respectively; Calculate the corrected moisture content MC correct The methods include: The basic calculation formula for defining the near-infrared spectrum detection module is: MC=τ1·A 1450 +τ2·A 1920 Where A 1450 and A 1920 are the absorption intensities of the near-infrared spectrum at 1450nm and 1920nm wavelengths, respectively; τ1 and τ2 are the absorption coefficients of the aqueous solution at 1450nm and 1920nm wavelengths, respectively; The basic calculation formula is updated based on the moisture content correction term ΔW: MC correct =τ1·(A 1450 +ΔW)+τ2·(A 1920 +ΔW)。 2. A system for rapidly detecting a latex explosive aqueous solution and water content according to claim 1, characterized in that: The calculation formula of the comprehensive index Q is: Q=η(t)·MC correct +δ(t)·c correct +κ(t)·ΔB; Where t is the time scale; η(t)·MC correct is the moisture content influencing factor; η(t) is the dynamic weight coefficient of moisture content; δ(t)·c correct is the concentration influencing factor; δ(t) is the concentration dynamic weight coefficient, which indicates the importance of the concentration of the aqueous solution in the current process; κ(t)·ΔB is the process parameter influencing factor; κ(t) is the dynamic weight coefficient of process parameters; The training method of the fusion model includes: Collect historical data of the currently produced latex explosives, including: comprehensive index Q, dynamic weight coefficient of moisture content η(t), dynamic weight coefficient of concentration δ(t), dynamic weight coefficient of process parameters κ(t), and corrected moisture content MC correct , corrected concentration c correct and parameter deviation ΔB; clean and normalize the above historical data, and combine the above data to construct a fusion feature vector X d ={(Q,η(t),MC correct ,δ(t),c correct ,κ(t),ΔB) d }; where d is the vector time step; Define recursive neural network as the basic structure of the fusion model, which includes input layer, hidden layer and output layer; Collect historical data of emulsion explosives of different specifications in the past, including: comprehensive index Q, dynamic weight coefficient of moisture content η(t), dynamic weight coefficient of concentration δ(t), dynamic weight coefficient of process parameters κ(t), and corrected moisture content MC correct , corrected concentration c correct and parameter deviation ΔB; clean and normalize the above historical data, and combine the above data to construct a historical fusion feature vector s is the number of vector time steps d; the historical fusion feature vector is expanded according to the time step d to form time series data as the training input of the fusion model, and the time series data is labeled according to the vector time step d; the label is the historical comprehensive index of latex explosives Initialize the network parameters of the fusion model; the network parameters include the weight matrix W from the input layer to the hidden layer x , the cyclic weight matrix W of the hidden layer h , the bias vector b of the hidden layer h , the weight matrix W from the hidden layer to the output layer o and the bias vector b of the output layer o ; Define the loss function of the ensemble model Where S is the length of the sequence data; Y true,d The label of the d-th vector time step, that is, the historical comprehensive index of the historical production of latex explosives Y pred,d is the prediction result of the dth vector time step, that is, the comprehensive index Q output by the fusion model; The time series data is passed to the input layer, and then passes through the hidden layer and the output layer in sequence to output the predicted comprehensive index Q, and calculate the value of the loss function of the corresponding comprehensive index Q, and backpropagate the error gradient along the output layer to the input layer; according to the error gradient, the network parameters are updated using the optimization algorithm; the time series data is repeatedly trained until the fusion model converges or reaches the preset number of iterations, which means that the fusion model training is completed.

3. A system for rapidly detecting a latex explosive aqueous solution and water content according to claim 2, characterized in that: The input layer is used to receive the fused feature vector X d As input; the hidden layer consists of y state vectors, and the calculation formula for each state vector is: d =f(W h ·h d-1 +W x ·X d +b h );where h d is the hidden state of the current vector time step d, f(...) is the activation function, W x is the weight matrix from the input layer to the hidden layer, W h is the cyclic weight matrix of the hidden layer, b h is the bias vector of the hidden layer; The output layer is based on the output of the hidden layer, i.e., the hidden state h of the current vector time step d d , calculate the output result of the fusion model, that is, the comprehensive index Q, and its calculation formula is: Q = η (W O ·h d +b o ); where η(...) is the activation function of the output layer, W o is the weight matrix from the hidden layer to the output layer, b o is the bias vector of the output layer.

4. A system for rapidly detecting a latex explosive aqueous solution and water content according to claim 2, characterized in that: The method for evaluating the comprehensive status of the latex explosive production line includes: Setting the index threshold Q based on the production specifications of latex explosives TH , the comprehensive index Q and index threshold Q output by the fusion model TH For comparison, if the comprehensive index Q is less than the index threshold Q TH When the comprehensive status of the production line is normal, if the comprehensive index Q is greater than or equal to the index threshold Q TH When , the overall status of the production line is abnormal.

Images