Automatic quality monitoring system based on nickel hydrazine nitrate manufacturing technology

Abstract

The invention discloses an automatic quality monitoring system based on a nickel hydrazine nitrate manufacturing technology, and relates to the technical field of nickel hydrazine nitrate manufacturing. The system has the functions of integrating a multi-wavelength filter wheel module, a process stability monitoring module and the like, and full-parameter real-time monitoring of the reaction process is realized. A Savitzky-Golay filtering algorithm is adopted to calculate a reaction rate, an XGBoost model is combined to predict a quality grade, and a hierarchical control strategy is designed to cope with different quality risks. And the risk assessment module simulates a crystal growth process by using a reaction kinetic model, identifies key risk factors, constructs a risk matrix to divide production areas, and generates an assessment report. Spectral data are corrected through a three-dimensional compensation matrix, the type of a quality problem is judged by adopting a double-layer fusion algorithm, and closed-loop control is realized in combination with a PID controller and an emergency emission program. A monitoring interface displays key parameters and early warning information in real time, dynamic adjustment of process parameters is supported, the production process is optimized, and the product quality and the production efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of nickel hydrazine nitrate manufacturing, and particularly relates to an automated quality monitoring system based on nickel hydrazine nitrate manufacturing technology. Background Art

[0002] In the industrial production of nickel hydrazine nitrate, the crystal quality directly determines the performance of the final product (such as energy density, thermal stability), and parameter fluctuations during the production process (such as temperature gradient, stirring speed deviation, supersaturation out of control) are extremely likely to cause problems such as a decrease in crystal purity, uneven particle size distribution, or excessive by-products. Traditional quality control methods mainly rely on the manual sampling-offline detection mode, which has the following significant defects:

[0003] Lack of real-time performance: Offline detection requires interrupting the production process, which takes up to several hours and cannot capture transient anomalies (such as local supersaturation or sudden temperature rise) in a timely manner, resulting in the accumulation of quality out-of-control risks.

[0004] Weak multi-parameter coupling analysis ability: Existing systems usually use a single sensor (such as a pH meter or a conductivity meter) or fixed threshold rules, and it is difficult to cope with complex quality fluctuations caused by the interaction of multiple factors such as temperature, stirring speed, and solvent compensation.

[0005] Separation of risk assessment and control: Most process monitoring systems only implement abnormal alarms and do not establish a direct mapping relationship between risk levels and control strategies.

[0006] Limited model generalization ability: Existing quality prediction methods based on empirical formulas (such as the Arrhenius equation) have large prediction errors in non-linear and high-dimensional parameter spaces, especially the modeling accuracy for dynamic processes such as crystal growth rate and by-product generation is insufficient.

[0007] In recent years, some studies have tried to introduce online monitoring technologies (such as near-infrared spectroscopy) or machine learning models, but there are still the following problems:

[0008] Data noise interference: Environmental noises such as intense stirring and bubble generation in the reaction kettle will contaminate the spectral signal, and traditional filtering algorithms (such as moving average) cannot effectively separate the effective information.

[0009] Lack of compensation mechanism: The dynamic effects of temperature and solvent conditions on sensor data (such as non-linear drift of conductivity caused by temperature increase) are not considered, and directly using the original data will introduce systematic errors.

[0010] Rigid control strategy: Most existing systems adopt a "one-size-fits-all" control logic. For example, regardless of the risk level, the parameters are adjusted at a fixed frequency, resulting in waste of system resources in high-load scenarios or missed detection of key anomalies.

[0011] In view of the above problems, the present invention proposes an automated quality monitoring system for nickel hydrazine nitrate manufacturing based on multi-source data fusion and dynamic risk grading, which improves the stability and safety of nickel hydrazine nitrate manufacturing through multi-wavelength spectroscopy analysis technology, dynamic calculation of process stability index, double-layer fusion quality determination algorithm, hierarchical control strategy, and reaction kinetics-driven risk assessment, providing a reliable intelligent solution for the industrial production of high-energy materials. Summary of the Invention

[0012] In order to overcome the shortcomings and deficiencies of the above-mentioned prior art, the present invention adopts the following technical solutions:

[0013] An automated quality monitoring system based on nickel hydrazine nitrate manufacturing technology, comprising

[0014] On-line reaction monitoring device: including an infrared light source, a reaction liquid detection cell, a multi-wavelength filter wheel and a spectroanalyzer; the multi-wavelength filter wheel includes a hydrazine nitrate filter, a nickel ion filter and a by-product filter, and the spectroanalyzer calculates the concentration of each component through the change in the transmitted light intensity;

[0015] Solid-liquid separation module: arranged at the outlet of the reaction kettle, used to separate solid precipitates and extract liquid phase samples to the on-line reaction monitoring device;

[0016] Process stability monitoring module: integrated with a stirring speed sensor and a temperature gradient sensor, used to collect the stirring speed and the temperature data of the upper, middle and lower layers of the reaction kettle in real time, dynamically adjust the weight coefficient in combination with the data of the viscosity sensor, and output the process stability index;

[0017] Reaction state evaluation module: perform Savitzky-Golay filtering on the concentration sequences of hydrazine nitrate and nickel ions and calculate the second derivative as the reaction rate, and trigger the high-frequency sampling mode according to the process stability index;

[0018] Quality prediction module: based on the XGBoost model, input the real-time concentration ratio, reaction rate, process stability index and quality problem type, output the quality grades of A / B / C and execute the corresponding control strategies;

[0019] Risk assessment module: establish a reaction kinetics model, simulate the crystal growth process under different parameter combinations, identify key risk factors, construct a quality risk matrix, divide risk levels, divide production areas according to risk levels, and generate a quality risk assessment report.

[0020] Preferably, the inner wall of the reaction liquid detection cell is integrated with a pH sensor, a conductivity sensor and a viscosity sensor, and a solvent compensation coefficient monitoring strategy is implemented, specifically including:

[0021] A three-dimensional compensation matrix of temperature-pH-conductivity is established, and the solvent compensation coefficient is obtained by looking up a table to dynamically correct the spectral analysis data; the experimental determination range of the three-dimensional compensation matrix is temperature 20-80 °C and pH 3-9.

[0022] Preferably, the calculation formula of the process stability index is:

[0023]

[0024] Where N 平均 is the moving average value; N 额定 is the moving rated value; ΔT is the temperature difference; T 允许最大值 is the maximum allowable temperature difference; ω1 and ω2 are weight coefficients; ω1 and ω2 are automatically adjusted according to the real-time viscosity data:

[0025] ω1 = 1 - ω2;

[0026] Where η0 is the reference viscosity; Δη is the real-time viscosity change.

[0027] Preferably, a double-layer fusion algorithm is used for the judgment of the quality problem type. Specifically:

[0028] The first layer is the multi-parameter threshold judgment, based on the dynamic threshold rules of particle size, purity, impurity content, temperature and rotation speed;

[0029] The second layer is the improved random forest model, which inputs basic features, derived features and time trend features;

[0030] The specific algorithm of the first layer is as follows:

[0031] The data of particle size D, purity P, impurity content I, reaction temperature T, and stirring speed n are obtained in real time, and dynamic thresholds are set:

[0032]

[0033] Where μD and σD are the mean and standard deviation of the historical particle size data; D min is the lower limit threshold of particle size; D max is the upper limit threshold of particle size; P min is the lowest allowable value of purity; I max is the highest allowable value of impurity content; T safe is the temperature safety range value; n safe is the rotation speed safety range value;

[0034] The judgment rule of the quality problem type:

[0035] If D < D min and P < P min and I > I max, it is determined that the raw material purity is abnormal;

[0036] If or , it is determined that the process parameters are abnormal;

[0037] If D > D max and P > 99.5%, it is determined that over-reaction crystallization occurs;

[0038] If I > 0.05%, it is determined that the by-product exceeds the standard;

[0039] If other situations occur, it is determined to be normal;

[0040] The second-layer algorithm is as follows:

[0041] Feature engineering:

[0042] Basic features: D, P, I, T, n;

[0043] Derived features: Among them, P is the concentration of the current solution, T is the real-time temperature of the reaction system, Psat(T) is the saturation concentration at temperature T, C is the concentration of the target component, and t is the time;

[0044] Time feature: Trend change rate in the recent 30 minutes;

[0045] Train the model and output the quality type; The result fusion strategy fuses the two-layer judgment results. Specifically:

[0046] If the two-layer results are consistent, directly output;

[0047] If there are differences, start the expert system.

[0048] Preferably, the control strategy of the quality prediction module includes:

[0049] The output labels are as follows:

[0050] Grade A: Prime product (comprehensive score ≥ m2)

[0051] Grade B: Qualified product (m1 ≤ score < m2)

[0052] Grade C: Rework product (score < m1)

[0053] Among them, m1 and m2 are the set score thresholds;

[0054] When the quality prediction output is Grade A, the system determines that the current process parameters are in the optimal state, locks the process parameters and switches the PID controller to the manual mode;

[0055] When the quality prediction output is Grade B, the system determines that the current process parameters are in the qualified state, dynamically balances quality and efficiency, and pre-emptively controls risks;

[0056] When the quality prediction output is at level C, the system determines that the current process parameters are in a dangerous state and triggers an emergency discharge procedure, which includes:

[0057] The first stage: Close the feed valve and inject liquid nitrogen for cooling in stages, and empty it into the explosion-proof isolation tank;

[0058] The second stage: Inject nitric acid, deionized water and sodium hydroxide for cleaning in sequence;

[0059] The third stage: Generate a root cause report and lock the system until manual confirmation.

[0060] Preferably, the staged injection of liquid nitrogen for cooling is specifically as follows: from 0 to 30 seconds, the injection rate of liquid nitrogen is 5 °C / min, and from 30 to 60 seconds, the injection rate of liquid nitrogen is increased to 10 °C / min; a 5% urea solution is preset in the explosion-proof isolation tank.

[0061] Preferably, the system also includes a monitoring interface for real-time display of:

[0062] The nitric hydrazide / nickel ion concentration curve, the process stability index, the three-dimensional temperature distribution heat map and the stress warning area;

[0063] The alarm management classification is as follows: yellow warning when nitrite ≥ 0.05%; orange warning when supersaturation > 1.3 or < 0.7; red warning when temperature > 80 °C or rotation speed < 180 rpm.

[0064] Preferably, the system also includes a final concentration calculation formula, specifically:

[0065]

[0066] where α is the solvent compensation coefficient; Δk is the change in conductivity; β is the process stability attenuation factor, with a value of 0.1 - 0.3; Var(Δt) is the temperature gradient variance.

[0067] Preferably, the expert system is as follows: When the first layer judges as "normal" and the second layer alarms, check the trend data in the recent 2 hours; when the first layer judges as "by-product exceeding the standard" and the second layer does not alarm, trigger the manual verification process.

[0068] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention are:

[0069] 1. The present invention constructs a PSI index through the non - linear fusion of stirring speed, three - layer temperature gradient, and viscosity data. The weight coefficient is adjusted in real - time according to viscosity, accurately reflecting the hydrodynamic state in the reaction kettle. When the PSI exceeds the threshold, the sampling interval is automatically shortened from 10 seconds to 1 second. The second - order derivative is extracted as the reaction rate by combining Savitzky - Golay filtering (5 - order polynomial, 15 - point window) to capture instantaneous process fluctuations.

[0070] 2. The present invention adopts a hybrid diagnosis architecture: in the first layer, dynamic threshold logic is used (e.g., the threshold of particle size D is adaptively adjusted according to historical data), and in the second layer, an improved random forest model is used to mine derived features (e.g., supersaturation ratio). Combining with an expert system to solve diagnosis conflicts and reduce the quality misjudgment rate; a quality prediction model driven by XGBoost: the input parameters cover real - time concentration ratio, reaction rate, PSI, and the fused quality problem types, and the output is the quality grade of A / B / C level, and it is linked with a PID controller or an emergency discharge program (such as liquid nitrogen grading cooling) to shorten the troubleshooting time.

[0071] 3. The risk assessment module of the present invention simulates the crystal growth process under different parameter combinations by establishing a reaction kinetics model, identifies the key risk factors affecting product quality, and constructs a quality risk matrix to divide the risk levels. Based on this, the system can finely divide the production area and implement differentiated monitoring and control strategies for areas with different risk levels. For example, measures such as increasing the number of sensors and shortening the sampling interval are taken in high - risk areas, while only routine monitoring is maintained in low - risk areas. This method effectively improves the resource utilization efficiency, ensures the safety and controllability of the entire production process, and greatly reduces the risk of product defects or safety accidents caused by unforeseen factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0073] Figure 1 FIG. shows the module diagram of the automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology of the present invention;

[0074] Figure 2 FIG. shows the working flow chart of the system of the present invention;

[0075] Figure 3 FIG. shows the emergency discharge program flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0076] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0077] In addition, the described features, structures, or characteristics can be combined in any suitable manner in one or more example embodiments. In the following description, many specific details are provided to give a thorough understanding of the example embodiments of the present disclosure. However, those skilled in the art will realize that the technical solutions of the present disclosure can be practiced without one or more of the specific details, or other methods, components, steps, etc. can be adopted. In other cases, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0078] Embodiment 1:

[0079] Refer to Figure 1 As shown, the automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology in this embodiment includes:

[0080] Online reaction monitoring device: used to detect the state of the nickel hydrazine nitrate synthesis reaction in real time; the online reaction monitoring device includes an infrared light source, a reaction liquid detection cell, a multi-wavelength filter wheel, and a spectral analyzer. The infrared light source is used to emit infrared light of a specific wavelength; the reaction liquid detection cell provides a closed light path environment and receives the infrared light; the multi-wavelength filter wheel includes a hydrazine nitrate filter, a nickel ion filter, and a by-product filter, and is used to transmit light waves matching the characteristic absorption peaks of the target components; the spectral analyzer calculates the concentrations of various components in the reaction liquid by detecting the change in the intensity of the transmitted light. A pH sensor, a conductivity sensor, and a viscosity sensor are installed on the inner wall of the reaction liquid detection cell to implement a solvent compensation coefficient monitoring strategy, and the detection data is dynamically corrected through the solvent compensation coefficient. The spectral analyzer integrates a reaction state evaluation strategy, adjusts the concentration calculation model in combination with process stability parameters and the solvent compensation coefficient, and constructs a quality prediction model for nickel hydrazine nitrate based on the multi-component concentrations.

[0081] Solid-liquid separation module: configured at the outlet of the reaction kettle, used to separate the solid precipitate in the reaction product and extract the liquid phase sample to the online reaction monitoring device; a stirring speed sensor and a temperature gradient sensor are integrated in the reaction kettle to implement a process stability monitoring strategy, and the rotation speed of the stirrer and the temperature difference in different regions of the reaction kettle are collected in real time;

[0082] Process Stability Monitoring Module: Integrates a stirring speed sensor and a temperature gradient sensor to collect real-time stirring speed and temperature data of the upper, middle, and lower layers of the reactor, dynamically adjusts the weight coefficient in combination with the viscosity sensor data, and outputs a process stability index;

[0083] Reaction State Evaluation Module: Performs Savitzky-Golay filtering on the concentration sequences of hydrazine nitrate and nickel ions and calculates the second derivative as the reaction rate, and triggers a high-frequency sampling mode according to the process stability index;

[0084] Quality Prediction Module: Based on the XGBoost model, inputs real-time concentration ratio, reaction rate, process stability index, and quality problem type, outputs A / B / C quality grades, and executes corresponding control strategies.

[0085] Figure 2 The working flow chart of this embodiment is shown as follows.

[0086] The implementation method of the process stability monitoring strategy includes: obtaining the average rotation speed within time T through the stirring speed sensor; obtaining the temperature data of the upper, middle, and lower layers of the reactor through the temperature gradient sensor, and calculating the maximum temperature difference value; obtaining real-time viscosity data through the viscosity sensor, dynamically adjusting the weight coefficient, and outputting the process stability index:

[0087] In this embodiment, it is collected once per second, and the 10-minute sliding average value is taken; the temperature of each layer is collected once per second, and the temperature difference is calculated.

[0088]

[0089] where N 平均 is the sliding average value; N 额定 is the sliding rated value; ΔT is the temperature difference; T 允许最大值 is the allowable maximum temperature difference; ω1 and ω2 are weight coefficients, and ω2 = 1 - ω1.

[0090] Parameter setting: N 额定 = 200 rpm, T 允许最大值 = 5 °C (configurable), ω1 and ω2 are automatically adjusted according to the real-time viscosity data:

[0091] ω1 = 1 - ω2;

[0092] where η0 is the reference viscosity; Δη is the real-time viscosity change.

[0093] In this embodiment,

[0094]

[0095] Calculation logic of the solvent compensation coefficient monitoring strategy: Obtain the ionic strength of the liquid-phase sample through a conductivity sensor, and establish a three-dimensional compensation matrix of temperature-pH-conductivity; combine the data of the pH sensor and the temperature sensor, and look up the table to obtain the solvent compensation coefficient. Experimentally determine the compensation coefficients under different temperature (20 - 80 °C) and pH (3 - 9) conditions.

[0096] Optimization process of the reaction state evaluation strategy: Perform Savitzky-Golay filtering (5th-order polynomial, 15-point window) on the concentration sequences of hydrazine nitrate and nickel ions, and calculate the second derivative of the filtered data as the reaction rate; when the process stability index exceeds the threshold, trigger the high-frequency sampling mode (the sampling interval is adjusted from 10 seconds to 1 second):

[0097] Trigger condition: S 稳定指数 > 1.5;

[0098] Final concentration calculation formula:

[0099]

[0100] Where α is the solvent compensation coefficient, Δk is the change in conductivity, β is the process stability attenuation factor (taking values from 0.1 to 0.3), and Var(Δt) is the temperature gradient variance.

[0101] Judgment of quality problem types: Includes a first-layer algorithm (multi-parameter threshold judgment) and a second-layer algorithm (improved random forest model).

[0102] The first-layer algorithm (multi-parameter threshold judgment) is as follows:

[0103] Obtain the data of particle size (D), purity (P), impurity content (I), reaction temperature (T), and stirring speed (n) in real time, and set dynamic thresholds:

[0104]

[0105] Where μD and σD are the mean and standard deviation of historical particle size data; D min is the lower limit threshold of particle size; D max is the upper limit threshold of particle size; P min is the lowest allowable value of purity; I max is the highest allowable value of impurity content; T safe is the temperature safety range value; n safe is the rotational speed safety range value.

[0106] Quality problem type judgment rules:

[0107] If D < D min and P < P min and I > I max, it is determined that the raw material purity is abnormal;

[0108] If or , it is determined that the process parameters are abnormal;

[0109] If D > D max and P > 99.5%, it is determined that over-reaction crystallization occurs;

[0110] If I > 0.05%, it is determined that the by-product exceeds the standard;

[0111] If other situations occur, it is determined to be normal.

[0112] The two-layer algorithm (improved random forest model) is as follows:

[0113] Feature engineering:

[0114] Basic features: D, P, I, T, n

[0115] Derived features: Among them, P is the concentration of the current solution, T is the real-time temperature of the reaction system, Psat(T) is the saturation concentration at temperature T, C is the concentration of the target component, and t is the time.

[0116] Time feature: Trend change rate in the recent 30 minutes

[0117] Train the model and output the quality type;

[0118] The result fusion strategy fuses the two-layer judgment results. Specifically:

[0119] If the two-layer results are consistent → directly output;

[0120] If there are disagreements → Start the expert system: When the first layer judges as "normal" and the second layer alarms, check the trend data in the recent 2 hours; when the first layer judges as "by-product exceeding the standard" and the second layer does not alarm, trigger the manual verification process.

[0121] Construction of the quality prediction model:

[0122] Establish a nickel hydrazine nitrate quality database, including:

[0123] Main product: Hydrazine nitrate ≥ 98.5%, nickel ion conversion rate ≥ 99.2%

[0124] By-products: Nitrite < 0.1%, unreacted hydrazine < 0.05%

[0125] Process parameters: Temperature 75 ± 2 °C, stirring speed 200 ± 10 rpm, supersaturation 0.8 - 1.2 Train the multi-parameter correlation model through XGBoost:

[0126] Input parameters: real-time concentration ratio (hydrazine nitrate / nickel ion), reaction rate (after Savitzky-Golay filtering), process stability index, supersaturation, and quality problem type (fusion result)

[0127] Output labels:

[0128] Grade A: Prime products (comprehensive score ≥ 95)

[0129] Grade B: Qualified products (85 ≤ score < 95)

[0130] Grade C: Rework products (score < 85)

[0131] When the quality prediction output is Grade A, the system determines that the current process parameters (temperature, stirring speed, feeding rate, etc.) are in the optimal state, locks the current values, and the PID controller switches to the manual mode.

[0132] When the quality prediction output is Grade B, the system determines that the current process parameters (temperature, stirring speed, feeding rate, etc.) are in the qualified state, dynamically balances quality and efficiency, and pre-emptively controls risks.

[0133] When the quality prediction output is Grade C, the system determines that the current process parameters (temperature, stirring speed, feeding rate, etc.) are in a dangerous state, immediately terminates the risk operation to prevent defective products from flowing downstream. Automatically closes the reaction kettle feed valve and starts the emergency discharge procedure.

[0134] Refer to Figure 3 as shown, where the emergency discharge procedure includes the first stage, the second stage, and the third stage.

[0135] The first stage (0 - 1 minute):

[0136] Close the feed valve and start staged cooling: from 0 to 30 seconds, the liquid nitrogen injection rate is 5℃ / min; from 30 to 60 seconds, the liquid nitrogen injection rate is increased to 10℃ / min; the separation module is emptied into the explosion-proof isolation tank (containing 5% urea solution).

[0137] The second stage (1 - 5 minutes):

[0138] Cleaning procedure: inject 5% nitric acid solution (3 minutes), deionized water (2 minutes), and 5% sodium hydroxide solution (3 minutes) in sequence

[0139] Record abnormal data: save all sensor data and quality prediction logs for the first 2 hours.

[0140] The third stage (after 5 minutes):

[0141] The system is locked, and a root cause report containing feature importance analysis is pushed;

[0142] Unlock condition: The engineer confirms and executes the equipment inspection checklist.

[0143] Monitoring screen: Real-time display of the hydrazine nitrate / nickel ion concentration curve, process stability index, quality grade, three-dimensional temperature distribution heat map (reactor cross-section), and stress warning area.

[0144] Alarm management: Hierarchical alarm (yellow warning: nitrite ≥ 0.05%; orange warning: supersaturation > 1.3 or < 0.7; red warning: temperature > 80°C or stirring speed < 180 rpm).

[0145] Risk assessment module: Establish a reaction kinetics model, simulate the crystal growth process under different parameter combinations, identify key risk factors (supersaturation, stirring speed, etc.), construct a quality risk matrix, divide the risk level (low / medium / high), divide the production area according to the risk level, and generate a quality risk assessment report.

[0146] The working process of the risk assessment module is as follows:

[0147] Step 1: Establish a reaction kinetics model and simulate the crystal growth process under different parameter combinations.

[0148] Data collection: Collect relevant data during the crystal growth process of nickel hydrazine nitrate, including crystal growth rate, particle size distribution, etc. under different conditions such as temperature, stirring speed, reactant concentration, supersaturation, etc.

[0149] Model selection and establishment: Select a reaction kinetics model to describe the crystal growth process of nickel hydrazine nitrate. In this embodiment, the Avrami equation is used to describe the nucleation and growth process of crystals:

[0150] X t =1-exp(-kt n );

[0151] where X t is the crystal growth fraction at time t, k is the rate constant, and n is the Avrami exponent. The values of k and n are determined by fitting the experimental data.

[0152] Simulate different parameter combinations: Use the established reaction kinetics model to simulate the crystal growth process under different parameter combinations. For example, set different temperature ranges (such as 50 - 80°C), stirring speed ranges (such as 150 - 300 rpm), and supersaturation ranges (such as 1.1 - 1.5), and calculate indicators such as crystal growth rate and final particle size under these parameter combinations through computer simulation software (such as MATLAB).

[0153] Specifically, S11: Define the parameter space.

[0154] In this embodiment, the three-dimensional parameter space is composed of temperature T, stirring speed S, and supersaturation C, and their ranges are respectively:

[0155] Temperature T ∈ [Tmin, Tmax];

[0156] Stirring speed S ∈ [Smin, Smax];

[0157] Supersaturation C ∈ [Cmin, Cmax].

[0158] S12. Generate all-factor combinations.

[0159] Generate all possible parameter combinations through the Cartesian product:

[0160] Number of combinations = nT × nS × nC

[0161] Where nT, nS, and nC are the number of discrete points of the parameters temperature T, stirring speed S, and supersaturation C respectively. For example, if each parameter takes 3 discrete values, the total number of combinations is 3 × 3 × 3 = 27.

[0162] S13. Establish a reaction kinetics model.

[0163] Assume that the mathematical models of crystal growth rate G and final particle size D are as follows:

[0164] Growth rate model: G = k1·T a ·S b ·C c ; where k1 is the rate constant; a, b, and c are the reaction orders of each parameter respectively.

[0165] Final particle size model: D = k2·G d ; where k2 is the particle size coefficient; d is the generation index.

[0166] Substitute the parameter combinations for calculation:

[0167] For each parameter combination (T i , S i , C i ), calculate:

[0168] G i,j,k = k1·T a i ·S b j ·C c k ; G i,j,k is the growth rate of the parameter combination (T i , S i , C i );

[0169] Di,j,k = k2·G d i,j,k ; D i,j,k is the final particle size model for the parameter combination (T i , S i , C i ).

[0170] Step 2: Identify key risk factors.

[0171] Data analysis: Analyze the crystal growth data obtained from simulations under different parameter combinations to identify the parameters that have a greater impact on crystal quality. Use correlation analysis and sensitivity analysis methods.

[0172] For parameter X and quality index Y, the Pearson correlation coefficient ρ xy is calculated as follows:

[0173]

[0174] where x i and y i are the parameter value and quality index value of the i-th data point respectively, and are the mean values of parameter X and quality index Y respectively.

[0175] Sensitivity analysis: Fix the temperature at T and observe the effects of stirring speed and supersaturation on crystal purity and particle size distribution. When the stirring speed increases from v1 rpm to v2 rpm (ΔX 搅拌速度 = v2 - v1) and the supersaturation is b2, the crystal purity increases from c0% to c2% (ΔY 晶体纯度 = c2% - c0%), and the particle size distribution increases from j1 μm to j2 μm (ΔY 粒径分布 = j2 - j1).

[0176] Sensitivity coefficient of crystal purity to stirring speed:[[]]

[0177]

[0178] Sensitivity coefficient of particle size distribution to stirring speed:[[]]

[0179]

[0180] When the supersaturation increases from b1 to b2 (ΔX 过饱和度 = b2 - b1) and the stirring speed is v2, the crystal purity increases from c1% to c2% (ΔY 晶体纯度 = c2% - c1%), and the particle size distribution increases from j1 μm to j2 μm (ΔY 粒径分布 = j2 - j1).

[0181] Sensitivity coefficient of crystal purity to supersaturation:

[0182]

[0183] Sensitivity coefficient of particle size distribution to supersaturation:

[0184]

[0185] Determine key risk factors: According to the results of correlation analysis and sensitivity analysis, mark the parameters with both the absolute value of the correlation coefficient and the sensitivity coefficient greater than the set threshold as key risk factors. For example, through correlation analysis and sensitivity analysis, it is found that supersaturation has the greatest impact on crystal particle size distribution, and stirring speed has a greater impact on crystal purity. Then supersaturation and stirring speed are key risk factors.

[0186] Step 3. Construct a quality risk matrix and divide risk levels (low / medium / high).

[0187] Determine risk assessment indicators: Select appropriate quality indicators as the basis for risk assessment, such as crystal purity, particle size distribution, impurity content, etc.

[0188] Set risk level thresholds: According to historical data and actual production experience, set thresholds for different risk levels for each quality indicator.

[0189] For example, for crystal purity:

[0190] Low risk: Purity ≥ 99%

[0191] Medium risk: 98% ≤ Purity < 99%

[0192] High risk: Purity < 98%

[0193] Construct a risk matrix: Combine key risk factors and risk assessment indicators into a matrix. For example, use supersaturation and stirring speed as rows and columns, and the risk level of crystal purity as matrix elements. According to the simulation results and thresholds, determine the risk level corresponding to each parameter combination.

[0194] Step 4. Divide production areas according to risk levels and generate a quality risk assessment report.

[0195] Division of production areas: According to the quality risk matrix, divide the production area into different risk level areas. For example, in the reaction kettle, according to the parameter combinations and corresponding risk levels at different positions obtained by simulation, divide the reaction kettle into low-risk area, medium-risk area and high-risk area.

[0196] Low-risk area: Maintain existing monitoring and control measures and regularly check the operation of equipment

[0197] Medium-risk area: Increase the number of sensors to monitor key parameters in real time, and adjust the control strategy in a timely manner according to parameter changes.

[0198] High-risk area: Shorten the sampling interval, strengthen the detection of crystal quality, and adjust the stirring speed and feeding rate when necessary.

[0199] The beneficial effects of this embodiment are as follows: Through multi-parameter real-time online monitoring and dynamic compensation mechanism, the quality stability of nickel hydrazine nitrate synthesis reaction is significantly improved; By integrating process stability index and solvent compensation strategy, the generation of by-products is effectively reduced; The double-layer intelligent diagnosis algorithm combined with the expert system improves the accuracy of abnormal judgment; The quality prediction model based on XGBoost realizes closed-loop control, automatically optimizes process parameters, and improves the A-grade product rate; The emergency discharge procedure classifies and disposes of risks to prevent defective products from flowing into the downstream and ensure production safety; The three-dimensional visual monitoring interface reduces the need for manual intervention and improves decision-making efficiency.

[0200] The weight coefficients of the present invention are used to measure the influence degree of different factors or variables on a certain result or decision. The definition of the weight coefficient refers to the value assigned to each factor when comparing and evaluating multiple factors to reflect its importance or priority. These weight coefficients can be determined according to specific situations and requirements, and are usually formulated and confirmed by professionals or relevant stakeholders. By reasonably setting the weight coefficients, it can help the program or system make more accurate decisions or predictions.

[0201] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on the computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0202] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0203] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0204] In several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only for some logical function divisions. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be electrical, mechanical, or other forms.

[0205] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0206] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. An automated quality monitoring system based on the manufacturing technology of nickel hydrazine nitrate, characterized in that, The system includes: Online reaction monitoring device: comprising an infrared light source, a reaction solution detection cell, a multi-wavelength filter wheel and a spectral analyzer; the multi-wavelength filter wheel includes a hydrazine nitrate filter, a nickel ion filter and a by-product filter, and the spectral analyzer calculates the concentration of each component based on the change in transmitted light intensity; Solid-liquid separation module: configured at the outlet of the reaction kettle, for separating solid precipitates and extracting a liquid phase sample to the online reaction monitoring device; Process stability monitoring module: integrating a stirring speed sensor and a temperature gradient sensor, for real-time collecting the stirring speed and the temperature data of the upper, middle and lower layers of the reaction kettle, dynamically adjusting the weight coefficient in combination with the data of the viscosity sensor, and outputting a process stability index; Reaction state evaluation module: performing Savitzky-Golay filtering on the concentration sequences of hydrazine nitrate and nickel ions and calculating the second derivative as the reaction rate, and triggering a high-frequency sampling mode according to the process stability index; Quality prediction module: based on the XGBoost model, inputting the real-time concentration ratio, reaction rate, process stability index and quality problem type, outputting the quality grades of A / B / C levels and executing corresponding control strategies; Risk assessment module: establishing a reaction kinetics model, simulating the crystal growth process under different parameter combinations, identifying key risk factors, constructing a quality risk matrix, dividing risk levels, dividing the production area according to the risk level, and generating a quality risk assessment report.

2. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 1, characterized in that, The inner wall of the reaction solution detection cell integrates a pH sensor, a conductivity sensor and a viscosity sensor, and implements a solvent compensation coefficient monitoring strategy, specifically including: Establishing a temperature-pH-conductivity three-dimensional compensation matrix, obtaining the solvent compensation coefficient by looking up the table, and dynamically correcting the spectral analysis data; the experimental determination range of the three-dimensional compensation matrix is temperature 20-80°C, pH 3-9.

3. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 2, wherein, The calculation formula of the process stability index is: Among them, N 平均 is the sliding average value; N 额定 is the sliding rated value; ΔT is the temperature difference; T 允许最大值 is the maximum allowable temperature difference; ω1 and ω2 are weighting coefficients; ω1 and ω2 are automatically adjusted according to the real-time viscosity data: ω1 = 1 - ω2; Wherein, η0 is the reference viscosity; Δη is the real-time viscosity change amount.

4. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 2, characterized in that, The judgment of the quality problem type adopts a double-layer fusion algorithm, specifically: The first layer is multi-parameter threshold judgment, based on the dynamic threshold rules of particle size, purity, impurity content, temperature and rotation speed; The second layer is an improved random forest model, inputting basic features, derivative features and time trend features; The specific algorithm of the first layer is as follows: Real-time obtaining the data of particle size D, purity P, impurity content I, reaction temperature T, and stirring speed n, and setting dynamic thresholds: where μD and σD are the mean and standard deviation of historical particle size data; D min is the lower particle size threshold; D max is the upper particle size threshold; P min is the minimum allowable purity value; I max is the maximum allowable impurity content value; T safe is the temperature safety range value; n safe is the rotational speed safety range value; Quality problem type judgment rules: If D < D min and P < P min and I > I max , it is determined that the raw material purity is abnormal; If or then it is determined that the process parameters are abnormal; If D > D max and P > 99.5%, then overreaction crystallization is determined; If I>0.05%, it is determined that the by-product exceeds the standard; If other situations occur, it is determined to be normal; The specific algorithm of the second layer is as follows: Feature engineering: Basic features: D, P, I, T, n; Derived feature: where P is the concentration of the current solution, T is the real-time temperature of the reaction system, Psat(T) is the saturation concentration at temperature T, C is the concentration of the target component, and t is time; Time feature: the trend change rate in the recent 30 minutes; Training the model and outputting the quality type; the result fusion strategy fuses the judgment results of the two layers, specifically: If the results of the two layers are consistent, directly output; If there is a disagreement, start the expert system: when the first layer judges as "normal" and the second layer alarms, check the trend data in the recent 2 hours; when the first layer judges as "by-product exceeding the standard" and the second layer does not alarm, trigger the manual verification process.

5. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 2, wherein The control strategy of the quality prediction module includes: The output labels are as follows: Grade A: premium product, comprehensive score ≥ m2; Grade B: Qualified products, m1 ≤ score < m2; Grade C: Rework products, score < m1; where m1 and m2 are set score thresholds; When the quality prediction output is Grade A, the system determines that the current process parameters are in the optimal state, locks the process parameters and switches the PID controller to the manual mode; When the quality prediction output is Grade B, the system determines that the current process parameters are in the qualified state, dynamically balances quality and efficiency, and pre - ventively controls risks; When the quality prediction output is Grade C, the system determines that the current process parameters are in a dangerous state and triggers an emergency discharge procedure, and the emergency discharge procedure includes: The first stage: Close the feed valve and inject liquid nitrogen for cooling in stages, and empty it into the explosion - proof isolation tank; The second stage: Inject nitric acid, deionized water and sodium hydroxide for cleaning in sequence; The third stage: Generate a root cause report and lock the system until manual confirmation.

6. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 5, characterized in that, The graded injection of liquid nitrogen for cooling is specifically as follows: from 0 to 30 seconds, the liquid nitrogen injection rate is 5℃ / min, and from 30 to 60 seconds, the liquid nitrogen injection rate is increased to 10℃ / min; A 5% urea solution is preset in the explosion - proof isolation tank.

7. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 1, characterized in that, The system also includes a monitoring interface for real - time display of: The nitric hydrazide / nickel ion concentration curve, the process stability index, the three - dimensional temperature distribution thermal map and the stress warning area; The alarm management grading is as follows: yellow warning when nitrite ≥ 0.05%; orange warning when supersaturation > 1.3 or < 0.7; red warning when temperature > 80℃ or rotation speed < 180 rpm.

8. The automated quality monitoring system based on nickel hydrazine nitrate manufacturing technology according to claim 2, characterized in that, The system also includes a final concentration calculation formula, specifically: where α is the solvent compensation coefficient; Δk is the change in conductivity; β is the process stability attenuation factor, with a value range of 0.1 - 0.3; Var(Δt) is the temperature gradient variance.

9. The automated quality monitoring system based on the nickel hydrazine nitrate manufacturing technology according to claim 1, characterized in that, The working process of the risk assessment module is as follows: Step 1: Establish a reaction kinetic model to simulate the crystal growth process under different parameter combinations; Collect relevant data during the crystal growth process of nickel nitrate hydrazide, including crystal growth rate and particle size distribution data under different temperatures, stirring speeds, reactant concentrations, and supersaturation conditions; The Avrami equation is used to describe the crystal nucleation and growth process: X t = 1 - exp(-kt n ); where X t is the crystal growth fraction at time t, k is the rate constant, n is the Avrami exponent, and the values of k and n are determined by fitting the experimental data; Use the established reaction kinetic model to simulate the crystal growth process under different parameter combinations. Specifically, S11: Define the parameter space. The three - dimensional parameter space is composed of temperature T, stirring speed S, and supersaturation C, and their ranges are respectively: temperature T ∈ [Tmin, Tmax]; stirring speed S ∈ [Smin, Smax]; supersaturation C ∈ [Cmin, Cmax]; S12: Generate all - factor combinations, and generate all possible parameter combinations through the Cartesian product: number of combinations = nT × nS × nC; where nT, nS, and nC are the discrete points of the parameters temperature T, stirring speed S, and supersaturation C respectively; S13: Establish a reaction kinetic model. The mathematical models of crystal growth rate G and final particle size D are as follows: Generation rate model: G = k1·T a ·S b ·C c ; where k1 is the rate constant; a, b, and c are the reaction orders of the respective parameters Final particle size model: D = k2·G d ; where k2 is the particle size coefficient; d is the generation exponent; Substitute the parameter combinations for calculation: For each parameter combination (T i , S i , C i ), calculate: G i,j,k = k1·T a i ·S b j ·C c k ; G i,j,k is the growth rate of the parameter combination (T i , S i , C i ); D i,j,k = k2·G d i,j,k ; D i,j,k is the final particle size model of the parameter combination (T i , S i , C i ). Step 2: Identify key risk factors; Analyze the crystal growth data under different parameter combinations obtained by simulation, identify the parameters that have a greater impact on crystal quality, and use correlation analysis and sensitivity analysis methods. For parameter X and quality index Y, the Pearson correlation coefficient ρ xy is calculated as follows: where x i and y i are the parameter value and quality index value of the i-th data point respectively, and are the mean values of parameter X and quality index Y respectively; The fixed temperature is T. When the stirring speed increases from v1 rpm to v2 rpm (ΔX 搅拌速度 = v2 - v1), and the supersaturation is b2, the crystal purity increases from c0% to c2% (ΔY 晶体纯度 = c2% - c0%), and the particle size distribution increases from j1 μm to j2 μm (ΔY 粒径分布 = j2 - j1); Sensitivity coefficient of crystal purity to stirring speed: Sensitivity coefficient of particle size distribution to stirring speed: When the supersaturation increases from b1 to b2 (ΔX 过饱和度 = b2 - b1), and the stirring speed is v2, the crystal purity increases from c1% to c2% (ΔY 晶体纯度 = c2% - c1%), and the particle size distribution increases from j1 μm to j2 μm (ΔY 粒径分布 = j2 - j1); Sensitivity coefficient of crystal purity to supersaturation: Sensitivity coefficient of particle size distribution to supersaturation: Mark the parameters whose absolute value of the correlation coefficient and sensitivity coefficient are both greater than the set threshold as key risk factors; Step 3: Construct a quality risk matrix and divide the risk levels, including low, medium, and high; Determine the risk assessment indicators, set the risk level thresholds, combine the key risk factors and risk assessment indicators into a matrix, and determine the risk level corresponding to each parameter combination according to the simulation results and thresholds; Step 4: Divide the production areas according to the risk levels and generate a quality risk assessment report; Production area division: According to the quality risk matrix, divide the production areas into different risk level areas. Inside the reactor, according to the parameter combinations and corresponding risk levels obtained by simulation at different positions, divide the reactor into a low-risk area, a medium-risk area, and a high-risk area; Low-risk area: Maintain the existing monitoring and control measures and regularly check the equipment operation status; Medium-risk area: Increase the number of sensors, monitor the key parameters in real time, and adjust the control strategy in a timely manner according to the parameter changes; High-risk area: Shorten the sampling interval, strengthen the detection of crystal quality, and adjust the stirring speed and feeding rate if necessary.

Images