A distributed control method for preparing waste nitric acid from rdx

Abstract

The application provides a distributed control method for preparing waste nitric acid from hexogen, comprising the following steps: sending an early warning signal to each monitoring point, each monitoring point receiving the early warning signal as an independent control node, and adjusting the collection pool capacity and the delivery pump frequency respectively; sending adjustment instructions about the adjusted collection pool capacity and the delivery pump frequency to corresponding execution devices, and dynamically adjusting the collection pool capacity and the delivery pump frequency according to the adjustment instructions; collecting the adjusted collection pool capacity and the delivery pump frequency data in real time, comparing the collected data with a preset target value, and calculating the deviation between the adjusted collection pool capacity and the delivery pump frequency and the target value; if the deviation between the adjusted collection pool capacity and the delivery pump frequency and the target value is less than a preset threshold value, it is determined that the distributed control effect reaches the expectation in advance, and the adjusted collection pool capacity and the delivery pump frequency data are stored in a historical database.

Description

Technical Field

[0001] The present invention relates to the field of information technology, in particular to a distributed control method for preparing waste nitric acid using RDX. Background Art

[0002] The treatment of waste nitric acid is a critical step in the RDX production process. Traditional methods for treating waste nitric acid are typically passive, with appropriate treatment measures only implemented when abnormalities in the waste nitric acid's properties are detected, such as the presence of unreacted lead nitrate particles. This lag not only increases the risk of environmental pollution but also potentially impacts the efficiency of subsequent treatment processes. Furthermore, parameter adjustments during the treatment process, such as setting the collection tank capacity and pump frequency, often rely on operator experience, lack scientific data support, and hinder precise control. Therefore, avoiding the environmental risks and reduced treatment efficiency caused by residual unreacted lead nitrate particles in the waste nitric acid is a key issue that must be addressed. Whether the traditional method of adjusting treatment parameters based on manual experience can meet the requirements of precise control and achieve optimal control of the collection tank capacity and pump frequency is a topic that requires further research. Whether centralized control systems can maintain stability and responsiveness in the face of complex and changing operating conditions presents a significant technical challenge. Leveraging long-term accumulated historical data, unlocking its potential, and providing a reliable basis for optimizing and improving process parameters is a crucial research direction. How to achieve collaborative work among multiple processing links and build an efficient, stable and reliable distributed control system is an engineering bottleneck that urgently needs to be solved. Summary of the Invention

[0003] The present invention provides a distributed control method for preparing waste nitric acid from RDX, which mainly includes:

[0004] Reaction parameters and process parameters in the process of preparing waste nitric acid from RDX are obtained in real time, and are associated with pre-stored historical reaction parameters, process parameters, and historical waste nitric acid characteristic data to form an associated data set. The associated data set is then input into a pre-built waste acid characteristic prediction model for training to obtain a trained waste acid characteristic prediction model.

[0005] The trained waste acid characteristic prediction model is used to predict upstream reaction parameters and process parameters, thereby obtaining the probability of carrying incompletely reacted lead nitrate particles. The predicted probability is compared with a preset threshold. If the predicted probability is greater than the threshold, it is determined that there is a risk of the waste nitric acid carrying incompletely reacted lead nitrate particles, and an early warning signal is triggered.

[0006] The early warning signal is sent to each monitoring point. Each monitoring point receives the early warning signal as an independent control node and adjusts the collection tank capacity and delivery pump frequency respectively;

[0007] Sending adjustment instructions for adjusting the collection tank capacity and the delivery pump frequency to the corresponding execution device, and dynamically adjusting the collection tank capacity and the delivery pump frequency according to the adjustment instructions;

[0008] Collect the adjusted collection tank capacity and delivery pump frequency data in real time, compare the collected data with the pre-set target value, and calculate the deviation between the adjusted collection tank capacity and delivery pump frequency and the target value;

[0009] If the deviation between the adjusted collection tank capacity and delivery pump frequency and the target value is less than the preset threshold, it is determined that the advance distributed control effect has achieved the expected result, and the adjusted collection tank capacity and delivery pump frequency data are stored in the historical database.

[0010] Furthermore, the reaction parameters and process parameters of the waste nitric acid preparation process using RDX are acquired in real time, and are associated with pre-stored historical reaction parameters, process parameters, and historical waste nitric acid characteristic data to form an associated data set. This data set is then input into a pre-built waste acid characteristic prediction model for training, thereby obtaining a trained waste acid characteristic prediction model. This includes: collecting real-time monitoring data such as waste acid concentration, pH, temperature, reactant conversion rate, impurity content, and treatment process parameter values ​​in the reaction tank during the waste nitric acid preparation process through multi-point sensing equipment, annotating the data according to the sampling time point, and obtaining a time-stamped original waste acid monitoring data set. Anomaly detection is performed on the original waste acid monitoring data set, and data is truncated by setting the valid range of waste acid concentration, pH, and temperature values. Missing data points are interpolated and supplemented using the sliding average method. Pearson correlation analysis is performed on the waste acid concentration, pH, temperature, and reactant conversion rate values ​​in the pre-processed standard waste acid data set. Feature parameters with correlation coefficients greater than a preset threshold are selected to obtain a key feature set of waste acid characteristics. An initial prediction model for waste acid characteristics was constructed using a random forest algorithm. The key feature set for waste acid characteristics was divided into a training set and a validation set. The model was trained using cross-validation to obtain an initial prediction model with high accuracy. To address the discrepancies between the initial prediction model output and the actual monitored values, a backpropagation algorithm was used to adjust the feature weights for waste acid concentration, pH, and temperature in the model. The waste acid characteristics prediction model was then developed through iterative optimization. The waste acid characteristics prediction model was used to predict the characteristics of a standardized waste acid dataset collected in real time. The predicted values ​​for waste acid concentration, pH, and temperature were output to form a waste acid characteristics prediction dataset.

[0011] Furthermore, the trained waste acid characteristic prediction model is used to predict upstream reaction parameters and process parameters, thereby obtaining the probability of carrying incompletely reacted lead nitrate particles. The predicted probability is compared with a preset threshold value. If the predicted probability is greater than the threshold value, it is determined that there is a risk of the waste nitric acid carrying incompletely reacted lead nitrate particles, and an early warning signal is triggered. This includes: using the gradient boosting algorithm to construct an upstream process predictor for the waste acid concentration, pH value, and temperature value output by the waste acid characteristic prediction model, and obtaining upstream process prediction parameters such as the reactor pressure value, stirring rate value, feed rate value, and reactant ratio value through reverse deduction. The waste acid viscosity coefficient is obtained by an online viscometer installed in the waste acid pipeline, and the flow resistance value of the waste acid is calculated according to the fluid mechanics formula based on the flow rate data measured by the pipeline flow meter to form a waste acid flow characteristic parameter set. Based on the reactor pressure value and stirring rate value in the upstream process prediction parameters, combined with the waste acid flow characteristic parameter set, a fluid dynamics sedimentation model is used to calculate the theoretical sedimentation rate value and spatial distribution density value of lead nitrate particles in the waste acid. Photoelectric turbidity sensors installed at the feed and discharge ports of the waste acid storage tank collect transmittance data from the waste acid solution. Ion-selective electrodes are used to detect the lead ion concentration in the waste acid. The predicted probability of incompletely reacted lead nitrate particles is calculated and compared with a preset threshold for lead nitrate particle settling. If the predicted probability exceeds the threshold, a digital warning signal is sent to the distributed controller. A data logger stores the upstream process prediction parameters, the waste acid flow characteristic parameter set, and the predicted probability of lead nitrate particles at the time of the warning trigger, forming a warning data package.

[0012] Furthermore, waste acid characteristics, reaction parameters, and process parameters are obtained from historical data, and a waste acid characteristics prediction model is established. The currently obtained waste acid characteristics and corresponding process parameters are input into the waste acid characteristics prediction model to predict the reaction parameters. The carryover probability of incompletely reacted lead nitrate particles is calculated using the predicted reaction parameters. If the carryover probability exceeds a preset threshold, the upstream reaction parameters or process parameters are adjusted, and the prediction and calculation are performed again, including: reading the real-time parameters such as waste acid concentration, pH value, temperature value, pressure value, stirring speed value collected by the sensor in the waste acid storage tank through a data collector, using a fixed interval cutoff method to remove outliers that exceed the limit, and combining historical waste acid parameter records to form a waste acid characteristics data set. For the waste acid characteristics data set, the maximum and minimum value normalization method is used to standardize the data. The process parameters such as the material ratio, feed rate value, and waste acid flow rate value in the reactor are collected through an online detector to form a standardized process parameter set. Based on the standardized process parameter set, the random forest algorithm is used to establish a mapping relationship between waste acid characteristics and reaction parameters. The mapping model is trained using a five-fold cross-validation method to obtain a waste acid characteristics prediction model. A waste acid characteristic prediction model is used to calculate the current process parameters and output predicted values ​​for reactor temperature, pressure, and agitation speed. Combined with the waste acid fluid mechanics calculation method, the predicted probability of lead nitrate particle carryover is calculated. A numerical comparator compares the predicted probability of lead nitrate particle carryover with a preset threshold. If the predicted probability exceeds the threshold, a parameter adjustment command is sent to the reactor controller. Based on the parameter adjustment command, the closed-loop controller dynamically adjusts the reactor temperature, pressure, and agitation speed, and the predicted probability of lead nitrate particle carryover is recalculated using the adjusted parameters.

[0013] Furthermore, the early warning signal is sent to each monitoring point. Each monitoring point receives the early warning signal as an independent control node and adjusts the collection tank capacity and the delivery pump frequency respectively, including: using a signal distributor to receive the upstream early warning signal, sending an independent digital early warning data packet to each monitoring point through the optical fiber communication network, and carrying the monitoring point number, early warning signal value, timestamp and other identification information in the data packet to obtain the monitoring point early warning data. The current liquid level value is collected by the liquid level sensor set in the collection tank, and the collection tank capacity control parameter is generated by the adaptive controller according to the early warning signal value in the monitoring point early warning data, and the opening control instruction is sent to the liquid level regulating valve. The waste acid flow rate value is collected by the flow meter at the collection tank outlet, and the proportional integral algorithm is used to generate the delivery pump frequency control parameter in combination with the collection tank liquid level value, and the speed control instruction is sent to the frequency converter. The delivery pressure value is collected by the pressure sensor installed at the delivery pump outlet, and the actual operating parameters of the delivery pump are calculated in combination with the waste acid flow rate value, and the operating status data is sent to the monitoring point controller. Based on the collection tank level and the pump operating parameters, a closed-loop controller calculates the adjustment deviation. Based on the deviation correction, a compensation control instruction is generated and sent to the level control valve and frequency converter. A data logger stores process parameters such as the collection tank level, waste acid flow rate, and delivery pressure, creating a monitoring point adjustment record and providing feedback on the adjustment execution status to the upstream controller.

[0014] Furthermore, an early warning signal is sent to each monitoring point, and each monitoring point receives the early warning signal as an independent control node. Adjusting the collection pool capacity and the delivery pump frequency respectively may include: extracting key information and evaluating the early warning level and impact range after analyzing the early warning signal of each monitoring point, judging whether to adjust the collection pool capacity according to the early warning situation, and if adjustment is required, calculating the target capacity value based on the waste nitric acid production rate, early warning duration and safety margin, and judging whether to adjust the delivery pump frequency, and if adjustment is required, calculating the target frequency value based on the target capacity value, early warning duration, delivery pump characteristics and pipeline resistance; in one embodiment, it may specifically include: using a signal analyzer to digitize the early warning signal of each monitoring point, obtaining the monitoring point number, early warning time, and signal strength value through a data segmentation extractor, and using early warning classification rules to judge the early warning level and impact range. Obtain the current capacity value through the collection pool capacity sensor, measure the real-time production rate using a waste acid production rate detector, and calculate the capacity value to be adjusted according to the capacity calculation rule in combination with the duration recorded by the early warning timer. The required safety capacity value is calculated using a capacity safety margin calculator. Combined with the capacity value to be adjusted, the target capacity value of the collection tank is calculated using capacity optimization rules, and a capacity adjustment instruction is generated. The current frequency value is collected using a delivery pump speed sensor, and the pipeline resistance value is obtained using a pipeline pressure monitor. Combined with the delivery pump performance curve data, the target frequency value of the delivery pump is calculated using frequency calculation rules. A control instruction generator is used to construct a capacity adjustment instruction package and a frequency adjustment instruction package, and adjustment instructions are sent to the collection tank liquid level control valve and the delivery pump frequency converter, respectively. Adjustment process parameters are collected using the collection tank liquid level sensor and the delivery pump pressure sensor, and the adjustment process data is recorded in a data storage device to form an adjustment execution record. In another embodiment, the received warning signal is parsed, the monitoring point number and warning value are extracted, and the warning level is evaluated using a classification algorithm. The warning level and impact range are determined based on the impact range. Based on the warning level and impact range, it is determined whether the collection tank capacity should be adjusted. If adjustment is required, the waste nitric acid production rate is obtained, and the target capacity value is calculated using a regression algorithm based on the warning duration and a preset safety margin. Based on the target capacity, determine whether to adjust the pump frequency. If so, obtain the pump characteristic parameters and pipeline resistance data. Combined with the warning duration, an optimization algorithm is used to calculate the target frequency. An adaptive controller is used to adjust the collection tank capacity based on the target capacity. The adjusted collection tank liquid level is obtained, and the collection tank capacity control parameters are calculated based on the level. Based on the target frequency, a proportional-integral algorithm is used to adjust the pump frequency. The adjusted pump operating parameters are obtained, and the pump frequency control parameters are calculated based on the operating parameters. A closed-loop controller is used to calculate the adjustment deviation value based on the collection tank liquid level value and the pump operating parameters. Compensation control instructions are generated based on the adjustment deviation value.

[0015] Furthermore, adjustment instructions for adjusting the collection tank capacity and delivery pump frequency are sent to the corresponding execution device. Based on the adjustment instructions, the collection tank capacity and delivery pump frequency are dynamically adjusted, including: using an instruction distributor to receive the collection tank capacity adjustment instruction and the delivery pump frequency adjustment instruction, and sending the adjustment instructions to the collection tank control cabinet and the delivery pump frequency conversion cabinet respectively via the fieldbus network to form an execution adjustment data packet. Based on the capacity adjustment instruction received by the collection tank control cabinet, a liquid level comparator is used to calculate the difference between the current liquid level and the target liquid level, and a liquid level valve controller is used to generate a linkage opening instruction for the inlet valve and the outlet valve. The frequency adjustment instruction is parsed by the delivery pump frequency conversion cabinet, and a frequency calculator is used to generate speed control curve parameters. Speed ​​control data is sent to the frequency converter, and the real-time speed value is collected in combination with the speed sensor. Based on the real-time liquid level value collected by the liquid level sensor, a liquid level monitor is used to calculate the deviation from the target liquid level, and a valve opening correction value is generated by the valve compensator to achieve precise adjustment of the collection tank liquid level. Based on the real-time speed of the delivery pump, a pressure sensor is used to collect pipeline pressure. The deviation from the target pressure is calculated using a pressure monitor, and a speed correction value is sent to the frequency converter. An execution status recorder is used to synchronously collect the liquid level valve opening value, delivery pump speed value, and pipeline pressure value to generate adjustment process data, and a completion signal is fed back to the host computer.

[0016] Furthermore, the system collects real-time data on the adjusted collection tank capacity and delivery pump frequency, compares the collected data with pre-set target values, and calculates the deviation between the adjusted collection tank capacity and delivery pump frequency and the target values. This includes: collecting real-time liquid level data from the collection tank via a liquid level sensor and real-time speed data from the delivery pump via a speed sensor; recording pipeline flow and pressure data using a data collector; removing data noise using a signal filter; and eliminating out-of-limit data points using a data validity determiner to obtain filtered data. The filtered data is converted from the liquid level data using a capacity calculator to real-time capacity values, and from the speed data using a frequency converter to real-time frequency values, obtaining parameter conversion data. The real-time capacity and frequency values ​​in the parameter conversion data are compared with pre-set target values ​​using a numerical comparator, and capacity deviation and frequency deviation data are obtained using a deviation calculator. The capacity deviation and frequency deviation data are then smoothed using a Kalman filter algorithm, a deviation trend curve is generated using a data fitter, and periodic sampling is performed using a data counter. The average deviation value is calculated based on the sampled data to generate a report on the equipment adjustment deviation.

[0017] Furthermore, if the deviation between the adjusted collection tank capacity and delivery pump frequency and the target value is less than a preset threshold, the early distributed control effect is determined to have met expectations, and the adjusted collection tank capacity and delivery pump frequency data are stored in a historical database. This includes: using a deviation calculator to numerically compare the collection tank capacity deviation and delivery pump frequency deviation values, calculating the mean squared deviation value using the root mean square method, and determining whether the mean squared deviation value is less than a preset threshold using a threshold comparator. For deviations less than the preset threshold, a stability determiner continuously collects and records deviation data points, and a parameter fluctuation calculator determines whether the fluctuation amplitude of the deviation value meets stability requirements. Based on the deviation data that meets stability requirements, a parameter collector records the collection tank capacity and delivery pump frequency values ​​at the corresponding time, and combines them with the operating time data to generate an operating parameter record. A data verifier performs an integrity check on the operating parameter record, and a data encoder performs time encoding on the records that pass the check, generating a timestamped historical data record. For the timestamped historical data records, a data classifier creates a data index table based on parameter type and time sequence, and the data writer stores the data in the historical database. The data verifier is used to verify the writing results, and the data storage report is generated through the data statistics device to record the data storage time, storage capacity, index structure and other information.

[0018] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0019] The present invention discloses a distributed control method for preparing waste nitric acid from RDX. The method obtains reaction parameters and process parameters in real time, combines historical data to train a waste acid characteristic prediction model, and predicts the probability of carrying unreacted lead nitrate particles in the waste nitric acid. When the predicted probability exceeds a threshold, an early warning signal is triggered and the collection tank capacity and the delivery pump frequency are adjusted in a distributed manner. By using the prediction model to make advance judgments on the characteristics of the waste nitric acid, the risk of environmental pollution caused by the residual unreacted lead nitrate particles can be effectively reduced. The present invention monitors the adjustment effect in real time, and when the expected goal is achieved, the data is stored in a historical database. This method can effectively prevent potential risks in the waste nitric acid treatment process, realize intelligent early warning and precise control, improve the safety and efficiency of the RDX preparation process, and continuously optimize the prediction model through data accumulation, thereby achieving continuous improvement in the waste acid treatment. The distributed control architecture enables each monitoring node to respond independently, shortens the control time, and improves the stability and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The present invention provides a flow chart of a distributed control method for preparing waste nitric acid using RDX.

[0021] Figure 2 The figure is a schematic diagram of a distributed control method for preparing waste nitric acid from RDX according to the present invention. DETAILED DESCRIPTION

[0022] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0023] like Figure 1-2 In this embodiment, a distributed control method for preparing waste nitric acid using RDX may specifically include:

[0024] S101. During the RDX preparation process, the reaction parameters and process parameters of waste nitric acid are acquired in real time, associated with pre-stored historical data to form a data set, and input into a pre-built waste acid characteristic prediction model for training to obtain an optimized prediction model, thereby providing a basis for subsequent waste nitric acid characteristic prediction.

[0025] S1011. Real-time data such as the concentration, pH, temperature, and reactant conversion rate of waste nitric acid is collected within the reactor using multi-point sensing equipment. Timestamps are added based on the sampling time to form a time-stamped raw waste acid monitoring data set. Taking the reactor as an example, concentration sensors installed at the top, middle, and bottom of the reactor collect data every 10 seconds, generating six data points per minute and constructing a multi-dimensional monitoring data stream. The collected parameters are affected by multiple factors, such as temperature and pressure, and may fluctuate or be abnormal, requiring further processing to ensure data reliability.

[0026] S1012. Preprocess the original waste acid monitoring dataset by setting valid ranges for waste acid concentration, pH, and temperature, performing data truncation, and interpolating missing data points using a sliding average method to obtain a normalized waste acid dataset. In this embodiment of the present invention, the valid range for waste acid concentration is set to 15% to 35%, and data outside this range is considered abnormal and eliminated. Missing data is supplemented using a sliding average method based on a 5-minute time window to ensure data continuity and integrity.

[0027] S1013. Perform Pearson correlation analysis on the concentration values, pH values, temperature values, and reactant conversion values ​​in the standardized waste acid data set, screen out characteristic parameters with correlation coefficients greater than the preset threshold, form a key feature set of waste acid characteristics, and train an initial prediction model based on the random forest algorithm. In an embodiment of the present invention, the analysis found that the correlation coefficient between temperature and concentration was above 0.6, and the concentration decreased by about 0.8 percentage points for every 5°C increase in temperature. Using historical data of approximately 770,000 data points over the past three months, the training set and validation set were divided into a 7:3 ratio, and a 5-fold cross-validation optimization model was used. The initial model accuracy rate reached 87.5%.

[0028] S1014. Optimize the initial prediction model, use the back-propagation algorithm to adjust the feature weights of the concentration value, pH value, and temperature value, obtain the optimal prediction model for waste acid characteristics after iterative training, and output the waste acid characteristics prediction data set through the model. In an embodiment of the present invention, in view of the low prediction accuracy in the initial stage of the reaction and the raw material replenishment stage, the training weights of these working condition data are increased, and the average accuracy of the final model on the validation set is increased to 94.3%, especially during the period of drastic parameter fluctuations, the accuracy rate increases by 8 percentage points. It is understandable that the specific layout of the sensing equipment and the threshold setting of the data preprocessing can be flexibly adjusted by technical personnel according to the actual production scenario. Through multi-scenario verification, for example, in the temperature range of 35°C to 45°C, the concentration values ​​predicted by the model are highly consistent with the actual values, proving its applicability under different working conditions.

[0029] S102. The upstream reaction parameters and process parameters are predicted through the optimized waste acid characteristic prediction model. The probability of carrying incompletely reacted lead nitrate particles in the waste nitric acid is calculated and compared with the preset threshold. If the threshold is exceeded, an early warning signal is triggered. At the same time, the model is optimized based on historical data and the parameters are dynamically adjusted to reduce risks.

[0030] During the RDX preparation process, the characteristics of waste nitric acid are affected by multiple factors and need to be analyzed through a combination of real-time data and prediction models. Using the trained waste acid characteristic prediction model, the real-time collected concentration, pH, and temperature values ​​are input, and the gradient boosting algorithm is used to derive the pressure value, stirring rate value, feed rate value, and reactant ratio value in the reactor to form an upstream process prediction parameter set. The gradient boosting algorithm iteratively optimizes the loss function and gradually fits the nonlinear relationship between the parameters. For example, the correlation weight between the waste acid concentration value and the pressure value is increased to 0.32 to ensure that the prediction results are close to the actual working conditions. The prediction parameter set reflects the dynamic state of the reaction system and lays the foundation for subsequent calculations.

[0031] Based on the upstream process prediction parameter set, the embodiment of the present invention combines the waste acid viscosity coefficient measured by the online viscometer and the flow rate data obtained by the pipeline flow meter to calculate the flow resistance value of the waste acid and form a waste acid flow characteristic parameter set. The online viscometer obtains the viscosity coefficient by detecting the shear stress of the waste acid in the pipeline, which usually fluctuates within the range of 1.5 to 2.0 milliPascal seconds, while the flow rate data is monitored by the flow meter in the range of 0.8 to 1.2 meters per second. The calculation of the flow resistance value is based on the Darcy-Weisbach formula, which comprehensively considers the pipeline diameter and fluid density to ensure that the results accurately reflect the physical properties of the waste acid. These parameters provide key inputs for the sedimentation analysis of lead nitrate particles.

[0032] S1021. Using the upstream process prediction parameter set and the waste acid flow characteristic parameter set, a hydrodynamic sedimentation model is used to calculate the theoretical sedimentation rate and spatial distribution density of lead nitrate particles. The predicted probability of incompletely reacted particles is determined by fusing data from a photoelectric turbidity sensor and an ion-selective electrode. In this embodiment of the present invention, the hydrodynamic sedimentation model, based on Stokes' law, combines a reactor pressure of 0.3 to 0.4 MPa and an agitation rate of 180 to 220 rpm to calculate a particle sedimentation rate between 0.5 and 1.0 mm / s. The spatial distribution density indicates that the concentration at the bottom is approximately 3 to 4 times that at the top. Photoelectric turbidity sensors are installed at the inlet and outlet of the waste acid storage tank to monitor changes in light transmittance. The normal value is 85% to 95%, dropping below 70% in abnormal conditions. The ion-selective electrode detects lead ion concentration. The standard value is 0.5 grams per liter, rising to 0.8 grams per liter or above in abnormal conditions. The two data types are fused using a weighted average method to derive a predicted probability. If the value exceeds 0.75, a digital warning signal is sent to the decentralized controller.

[0033] S1022. Extract waste acid characteristics and related parameters from historical data, construct and optimize a waste acid characteristic prediction model, input real-time waste acid characteristic data into the model to predict reaction parameters, calculate the lead nitrate particle carryover probability, and adjust upstream parameters and re-predict when thresholds are exceeded. This embodiment of the present invention uses a sensor array to collect concentration, pH, temperature, pressure, and stirring speed values ​​within the waste acid storage tank. A fixed interval cutoff is used to eliminate outliers, for example, limiting concentration values ​​to 25% to 30%. A maximum-minimum normalization method is then used to normalize the parameters to the range of 0 to 1. For example, a temperature value of 42 degrees Celsius is normalized to 0.467 within the range of 35 to 50 degrees Celsius. The standardized process parameter set includes a material ratio of 0.8 to 1.2, a feed rate of 150 to 200 liters per hour, and a waste acid flow rate of 0.8 to 1.2 meters per second. Based on this, a prediction model is established using a random forest algorithm. Through five-fold cross-validation training, the predicted values ​​for reactor temperature, pressure, and stirring speed are output. The random forest algorithm enhances model robustness through a voting mechanism involving multiple decision trees. The training data covers approximately 2,000 samples from the past three months, achieving an accuracy rate of 92%. If the predicted probability exceeds 0.75, an adjustment command is sent to the reactor controller. The closed-loop controller adjusts the temperature in 2-degree Celsius steps, the pressure in 0.05 MPa steps, and the stirring speed in 20 rpm steps. After optimization, the probability value is recalculated.

[0034] In an embodiment of the present invention, the prediction of waste acid characteristics involves the synergistic effect of multiple parameters. For example, when the concentration value is stable at 25% to 30% and the temperature value is 40 to 45 degrees Celsius, the reaction system is relatively stable and the lead ion concentration fluctuates less. The gradient boosting algorithm reversely derives the ideal range of process parameters by analyzing historical steady-state data, such as a pressure value of 0.35 MPa and a stirring rate value of 200 revolutions per minute. The flow characteristic parameters reveal the behavior of waste acid in the pipeline. For example, when the viscosity coefficient increases, the sedimentation rate decreases by about 0.2 mm per second. The coordinated monitoring of the photoelectric turbidity sensor and the ion selective electrode ensures the reliability of the predicted probability value. The synchronization of the decrease in transmittance and the increase in lead ion concentration under abnormal operating conditions is particularly obvious.

[0035] The predicted probability value is recorded at a frequency of 100 Hz by a data acquisition card and compared with a threshold of 0.75, ensuring real-time performance. After each warning is triggered, the system stores 30 minutes of process parameter data as a data package, including pressure values, stirring rate values, and flow characteristic parameters, to facilitate subsequent traceability and optimization. Historical data analysis shows that when the temperature value deviates from the range of 40 to 45 degrees Celsius or the pressure value deviates from 0.35 MPa, the risk of particle carryover increases significantly. After adjusting the parameters, the risk probability can be reduced to below 0.6.

[0036] The dynamic adjustment mechanism effectively reduces the risk of carryover of unreacted lead nitrate particles. For example, when the feed rate rises above 200 liters per hour, the predicted probability may approach 0.8. By reducing the feed rate to 180 liters per hour and increasing the stirring rate to 220 revolutions per minute, the probability quickly falls back to a safe range. Algorithm parameters and sensor configurations can be adjusted based on actual operating conditions to meet the needs of different production scenarios.

[0037] S103. Distribute the warning signal to each monitoring point. Each monitoring point receives and analyzes the signal as an independent control node, and adjusts the collection pool capacity and delivery pump frequency according to the warning level and real-time data to ensure that the system responds efficiently to abnormal working conditions.

[0038] In an embodiment of the present invention, the early warning signal is transmitted to each monitoring point through an optical fiber communication network, and a signal distributor is used to convert the digital early warning signal generated upstream into an independent data packet, which contains information such as the monitoring point number, the early warning signal value and the timestamp. The optical fiber network is designed based on a star topology, and the transmission delay is controlled within 100 milliseconds to ensure that the signal arrives in real time. The early warning signal value is divided into four levels according to its intensity: Level 1 is a slight abnormality, Level 2 is a moderate deviation, Level 3 is a significant risk, and Level 4 is a serious abnormality. Different levels trigger different control strategies, aiming to quickly respond to changes in the characteristics of waste nitric acid.

[0039] After receiving the warning data packet, each monitoring point uses a signal analyzer to digitize the signal, extracting the monitoring point number, warning time, and signal strength value. The system then assesses the warning level based on pre-set grading rules. For example, a signal strength value of 1 to 3 is considered a minor warning, limited to a single point; a value of 7 to 8 is considered a severe warning, potentially affecting multiple downstream nodes. This parsing process is accomplished using segmented extraction technology. A typical data packet contains a 16-bit number, a 32-bit timestamp, and an 8-bit strength value, ensuring complete and traceable information.

[0040] S1031. Based on the warning signal, the collection tank liquid level data is collected. The target capacity is calculated based on the waste acid production rate and safety margin. The adaptive controller generates adjustment instructions and adjusts the opening of the liquid level control valve, while also recording the adjustment process parameters. In this embodiment of the present invention, a liquid level sensor is installed in the collection tank to monitor liquid level changes in real time. Under normal operating conditions, the liquid level is maintained between 60% and 75%. For a 50-cubic-meter capacity, this is between 30 and 37.5 cubic meters. When a Level 3 warning signal is received, the liquid level sensor detects a current value of 35 cubic meters, while the waste acid production rate detector measures a real-time rate of 12 cubic meters per hour. The safety margin calculator sets a base capacity of 15 cubic meters based on historical data. If the warning persists for more than 30 minutes, the required capacity is calculated based on 1.5 times the production rate, or 18 cubic meters per hour. Based on the current liquid level and the required margin, the adaptive controller determines a target capacity of 25 cubic meters and issues a command to the liquid level control valve to increase its opening by 20%, gradually lowering the liquid level to allow for abnormal operating conditions. During the adjustment process, the liquid level sensor collects data at a frequency of once per second and generates a dynamic curve for subsequent analysis.

[0041] S1032: Based on the target capacity and waste acid flow rate data, a proportional-integral algorithm is used to calculate the transfer pump frequency control parameters. The closed-loop controller corrects for any deviations and sends frequency adjustment commands to the frequency converter to optimize transfer efficiency. In this embodiment of the present invention, a flow meter at the collection tank outlet monitors the waste acid flow rate. The normal value is 8 cubic meters per hour, which may rise to 12 cubic meters per hour during an alert. The proportional-integral algorithm calculates the target frequency of the transfer pump by analyzing the rate of liquid level decline and the required flow rate. For example, if the liquid level drops from 70% to 50%, the frequency needs to be increased from 35 Hz to 45 Hz. The algorithm's core principle is to ensure rapid response to changes in the proportional term and eliminate steady-state errors in the integral term. In this implementation, the proportional coefficient is set to 0.8, and the integration time is set to 10 seconds. The pressure sensor at the transfer pump outlet simultaneously collects pressure values, increasing from 0.3 MPa to 0.4 MPa to verify the improved transfer capacity. The closed-loop controller calculates the deviation between the liquid level and the target value every 5 seconds. If the rate of decline exceeds 2% per minute, the frequency correction amplitude is controlled to within 2 Hz to ensure a smooth transition. The adjustment command is sent to the inverter in a 24-bit format, including the target frequency and acceleration and deceleration time.

[0042] In actual operation, monitoring points operate independently yet interconnectedly. When an upstream monitoring point triggers regulation, downstream nodes respond three to five minutes in advance, creating a tiered control model to mitigate overall system fluctuations. For example, when the upstream liquid level drops rapidly, the downstream transfer pump frequency is preemptively increased by 5 Hz, increasing the waste acid flow rate by 3 cubic meters per hour and the pressure by 0.1 MPa. This strategy, based on historical data analysis, demonstrates that the optimal time difference between adjacent nodes effectively mitigates sudden changes in parameters.

[0043] In an embodiment of the present invention, the adjustment process data is recorded throughout. The data recorder stores the liquid level value, flow rate value and pressure value at intervals of 30 seconds to form a monitoring point adjustment record and feed it back to the upstream controller. For example, in a level 3 early warning response, the liquid level dropped from 75% to 48%, which took 15 minutes. The delivery pump frequency increased by 12 Hz, the flow rate increased by 5 cubic meters per hour, and the pressure value changed by 0.15 MPa. These data not only reflect the adjustment effect, but also provide a basis for process optimization. The generation of adjustment instructions adopts a standardized format. The capacity instruction contains a 32-bit control word, the upper 16 bits are the target capacity, and the lower 16 bits define the adjustment rate; the frequency instruction specifies the acceleration time to ensure accurate execution.

[0044] Through this approach, each monitoring point can flexibly adjust parameters based on the warning level. For example, in a moderate warning, the level control valve opening increment is controlled at 10% to 15%, and the frequency adjustment range is 5 to 10 Hz. In a severe warning, the opening can be increased to 25%, and the frequency is increased to the upper limit of 50 Hz.

[0045] S104. Send the instructions for adjusting the collection tank capacity and the delivery pump frequency to the corresponding execution device, generate an execution data packet through the fieldbus network, and drive the liquid level valve and the frequency converter to dynamically adjust the parameters. At the same time, collect the adjustment process data to verify the execution effect.

[0046] In an embodiment of the present invention, after the adjustment instruction is generated by the upstream controller, it is transmitted to the collection pool control cabinet and the delivery pump frequency conversion cabinet through the instruction distributor via the field bus network. The field bus adopts the Profibus-DP protocol with a communication rate of up to 12 megabits per second, ensuring that the instruction is delivered within 100 milliseconds. The execution adjustment data packet is designed to be in a standardized format, including a 16-bit device address for identifying the target unit, a 32-bit instruction code for defining the specific adjustment action, and a 16-bit check code for verifying data integrity. This design not only improves transmission efficiency, but also reduces the risk of misoperation. The instruction distributor supports multi-channel parallel processing and can send independent instructions to multiple monitoring points at the same time to ensure the overall coordination of the system.

[0047] S1041. According to the collection tank capacity adjustment instruction, the liquid level comparator is used to calculate the difference between the current liquid level and the target liquid level, and the opening instructions of the water inlet valve and the water outlet valve are generated by the liquid level valve controller. The deviation correction is performed in combination with the real-time liquid level data to achieve precise adjustment. In an embodiment of the present invention, the collection tank liquid level sensor monitors the liquid level status in real time. For example, for a collection tank with a capacity of 50 cubic meters, the liquid level is maintained between 30 and 37.5 cubic meters during normal operation. If an instruction is received to adjust the target liquid level to 25 cubic meters, the liquid level comparator detects that the current liquid level is 35 cubic meters, and the difference is 10 cubic meters. Based on this difference, the liquid level valve controller adopts a dual-valve linkage strategy: close the water inlet valve, and at the same time increase the opening of the water outlet valve from the initial 20% to 45%. During the adjustment process, the action time difference is controlled within 2 seconds to avoid rapid fluctuations in the liquid level. The liquid level monitor collects data every 100 milliseconds. If the rate of change exceeds 2% per minute, the valve compensator generates a correction value, for example, adjusting the outlet valve opening to 40% to smooth the transition. This design goal is to maintain stable liquid level regulation through real-time feedback while leaving sufficient space for the subsequent inflow of waste acid.

[0048] The delivery pump frequency is adjusted in response to changes in the liquid level. Upon receiving the frequency adjustment command, the delivery pump frequency converter generates speed control curve parameters using a frequency calculator and sends speed control data to the frequency converter. For example, when the command requires a frequency increase from 35 Hz to 45 Hz, the speed control curve is set with an acceleration time of 30 seconds, and the curve exhibits an S-shaped curve to reduce mechanical shock. The frequency calculator considers the target flow rate of 12 cubic meters per hour and the pipeline resistance, and calculates that the speed must reach 1800 rpm. When the frequency converter executes the command, the speed sensor collects real-time values ​​at a frequency of four times per second, ensuring that the adjustment accuracy is controlled within 0.1 Hz. This soft start method not only extends the life of the equipment but also ensures the continuity of waste acid transportation.

[0049] During the adjustment process, the execution status recorder simultaneously collects key parameters, including the opening values ​​of the inlet and outlet valves, the speed of the delivery pump, and the pipeline pressure, forming a complete data set for the adjustment process. For example, in a typical adjustment, the liquid level drops from 75% to 60% in approximately 10 minutes, the outlet valve opening increases from 20% to 40%, the delivery pump speed increases from 1500 to 1800 rpm, and the pipeline pressure increases from 0.3 MPa to 0.4 MPa. This data is recorded with a sampling period of 100 milliseconds, generating approximately 6000 data points in total, detailing the dynamic changes in liquid level, speed, and pressure. When the parameters stabilize and the deviation is less than 1%, the recorder sends a completion signal to the host computer, providing a basis for subsequent analysis.

[0050] In an embodiment of the present invention, the reliability of the adjustment process relies on high-precision sensors and feedback loops. The opening of the liquid level valve is measured by an absolute encoder with a resolution of 0.1 degrees, ensuring the precise execution of each adjustment. The speed of the delivery pump is monitored by a Hall sensor, with an error of less than 1 rpm at 1500 rpm, reflecting the effect of frequency changes in real time. The pipeline pressure is detected by a piezoresistive sensor with a range of 0 to 1 MPa and an accuracy of 0.1%, which can keenly capture pressure fluctuations. These sensor data are used to correct parameters in real time through a closed-loop control system. For example, when the pressure exceeds 0.4 MPa, the frequency converter automatically reduces the speed by approximately 50 rpm to maintain system balance. Such a precise feedback mechanism significantly improves the stability and safety of the adjustment.

[0051] S1042: Based on the pump speed and pipeline pressure data, a pressure monitor calculates the deviation from the target value. The speed is dynamically adjusted via a frequency converter to optimize delivery efficiency, while the adjustment data is recorded to track process status. In actual operation, the liquid level and pump regulation are linked. When the liquid level drops at a rate of 3% per minute, the pressure sensor detects that the pipeline pressure has risen to 0.45 MPa, exceeding the target by 0.4 MPa. The pressure monitor calculates the deviation and generates a speed correction value. For example, it can slightly reduce the frequency from 45 Hz to 43 Hz to bring the pressure back to a safe range. This correction process is executed by the frequency converter, with the speed adjustment step size controlled within 20 revolutions per minute to avoid sudden flow rate fluctuations. This dynamic optimization ensures that the waste acid delivery capacity matches the liquid level fluctuations while keeping pressure fluctuations within 0.05 MPa, improving system operation stability.

[0052] In an embodiment of the present invention, the execution effect of the adjustment instruction is verified through data logging. A complete adjustment may involve a 15% drop in liquid level, a 300 rpm increase in speed, and a 0.15 MPa increase in pressure. The process data records in detail the response time and parameter change amplitude of each node. For example, after the outlet valve opening is adjusted, the liquid level change curve shows a steady downward trend, the flow rate gradually stabilizes after the speed is increased, and the fluctuation converges rapidly after the pressure is adjusted. These data not only prove the success of the adjustment, but also provide rich information for optimizing the control strategy. In practical applications, the valve opening and speed control curve parameters can be adjusted according to the waste acid treatment volume and pipeline characteristics to meet the requirements of different working conditions.

[0053] S105. Real-time collection of the adjusted collection tank capacity and delivery pump frequency data, comparison with the preset target value to calculate the deviation, and generation of a deviation trend curve through filtering and smoothing to evaluate the adjustment effect. In an embodiment of the present invention, the data acquisition link relies on high-precision sensors. The liquid level sensor is deployed at multiple height positions in the collection tank, using the ultrasonic ranging principle, with a sampling frequency set to 10 Hz and a measurement accuracy of 0.1%, capable of capturing slight changes in the liquid level. The delivery pump speed is monitored by a Hall sensor installed at the end of the pump shaft, which generates 60 pulse signals per revolution based on the magnetic encoder to ensure high resolution of the speed data. The data collector synchronously records pipeline flow and pressure data, such as the flow value measured by the electromagnetic flowmeter and the pressure value obtained by the piezoresistive pressure sensor, to form a multi-dimensional original operation data set. These data provide comprehensive process status information for subsequent analysis.

[0054] The collected raw data needs to be processed to improve reliability. The signal filter uses a low-pass filtering algorithm with a cutoff frequency set to 1 Hz to remove 2 to 5 Hz high-frequency noise caused by the flow of waste acid. For example, liquid level data is often interfered by liquid level fluctuations, and the signal is smoother after filtering. The data validity judge eliminates outliers based on physical constraints, sets the liquid level valid range to 15% to 95%, and the speed range to 500 to 2000 revolutions per minute. Data outside the range is considered invalid and removed. After filtering and screening, the first processed data obtained can more truly reflect the operating status of the equipment and lay the foundation for parameter conversion.

[0055] S1051. Using the processed data, the real-time capacity and frequency values ​​are calculated and compared with the target values ​​to generate deviation data. This data is then smoothed using a Kalman filter algorithm to generate a trend curve. In this embodiment of the present invention, the capacity calculator converts the real-time capacity based on the liquid level data and the geometric characteristics of the collection tank. For example, a 50-cubic-meter tank with a liquid level of 60% corresponds to 30 cubic meters. This calculation takes into account the nonlinear effect of the tank bottom shape on the capacity, ensuring accurate results. The frequency converter converts the speed data into a frequency value based on the relationship between speed and the number of motor pole pairs. For example, 1500 rpm corresponds to 50 Hz. The value comparator compares the real-time value with the target value on a second-by-second basis. If the real-time capacity is 28 cubic meters and the target is 32 cubic meters, the deviation is negative 4 cubic meters, accounting for 12.5%. If the real-time frequency is 42 Hz and the target is 45 Hz, the deviation is negative 3 Hz, accounting for 6.7%. These deviation data directly quantify the degree of regulation deviation. The Kalman filter algorithm then intervenes to smooth the deviation data through prediction and update steps. The algorithm assumes a measurement noise standard deviation of 0.5% and a system noise of 0.2%. Each iteration combines historical data with current measurements to generate a continuous deviation trend curve. This approach effectively reduces the impact of short-term fluctuations and makes the trend more valuable.

[0056] In practical applications, generating deviation trend curves is crucial for process optimization. For example, during a single adjustment, the initial capacity deviation was -12.5% ​​and the frequency deviation was -6.7%. After 10 minutes of adjustment, the capacity deviation decreased to -2.3% and the frequency deviation decreased to -1.5%. A data logger samples data every five minutes, calculates the average deviation, and generates a device adjustment deviation report. These reports document how deviations change over time, such as a rapid decrease in capacity deviation during the initial adjustment period and a more stable trend later in the process, reflecting the control system's responsiveness. This data analysis not only verifies the effectiveness of adjustments but also provides a basis for fine-tuning parameters.

[0057] S1052. The flow and pressure data collected by multi-point sensors are used to assist in verifying the adjustment effect, and the system stability is evaluated in combination with the deviation data. In an embodiment of the present invention, the pipeline flow meter and pressure sensor monitor the waste acid flow rate and pipeline pressure respectively. The flow meter has an accuracy of 0.5%, and the pressure sensor has a range of 0 to 1 MPa with an accuracy of 0.1%. For example, when the delivery pump frequency increases from 42 Hz to 45 Hz, the flow rate increases from 10 cubic meters per hour to 12 cubic meters per hour, and the pressure increases from 0.35 MPa to 0.4 MPa. These data are correlated with the liquid level and speed data for analysis. If the flow and pressure changes are consistent with expectations, it indicates that the adjustment action is effective. If a deviation occurs, for example, the pressure rises to 0.45 MPa beyond expectations, it may indicate that the pipeline resistance is abnormal and further inspection is required. Combined with the deviation trend curve, when the average deviation of all parameters is continuously less than 3%, the system operation is considered to have entered a stable state.

[0058] In the embodiments of the present invention, rigorous data processing and deviation calculation ensure accurate process control. For example, the Kalman filter's predictive model, by continuously updating state estimates, maintains the smoothness of the deviation curve even with large liquid level fluctuations. In actual operation, the sensor's high-frequency sampling and filtering processing work together to enable the system to quickly respond to anomalies and restore stability. The trend data recorded in the deviation report can also be used for long-term optimization. For example, it was found that the lag time of the frequency adjustment on flow rate is approximately 15 seconds, which can guide future improvements to the control strategy.

[0059] S106. If the deviations between the adjusted collection tank capacity and delivery pump frequency and the target values ​​are less than a preset threshold, it is determined that the distributed control effect meets expectations, and the relevant data is encoded and stored in a historical database for traceability and analysis.

[0060] In an embodiment of the present invention, deviation analysis is a key link in verifying the control effect. The deviation calculator obtains the collection tank capacity deviation value and the delivery pump frequency deviation value in real time, and uses the root mean square method to calculate the overall level of deviation. For example, for the collection tank capacity, the system collects data once per second and continuously takes 60 data points to calculate the mean square value. If the actual capacity is 48 cubic meters and the target is 50 cubic meters, the deviation value fluctuates within plus or minus 2 cubic meters, and the mean square value is about 0.9%, which is lower than the preset threshold of 1%. Similarly, if the actual value of the delivery pump frequency is 42 Hz and the target is 43 Hz, the deviation mean square value is calculated to be 0.6%, which also meets the threshold requirement. The root mean square method comprehensively reflects the amplitude and distribution characteristics of the deviation through square and square root operations, and can better reflect the overall stability of the parameters than a single difference. The threshold comparator then determines whether the mean square value meets the standard to ensure that the control result meets the process requirements.

[0061] Stability judgment further verifies the reliability of the parameters. For the case where the mean square value is less than the threshold, the stability judge continuously collects deviation data points, sets a 10-minute observation window, records about 600 data points, and calculates the fluctuation amplitude. Taking capacity as an example, if the fluctuation range is controlled within plus or minus 0.5% and lasts for more than 8 minutes, it is considered to have reached a stable state. The stability requirements for frequency are higher, and the fluctuation must be within plus or minus 0.3% and last for at least 5 minutes. This strict criterion stems from the high requirements for parameter consistency in waste acid treatment to avoid short-term stability covering up potential fluctuations. After stability is confirmed, the parameter collector records the capacity value and frequency value at the corresponding moment. For example, the capacity is stable at 49 cubic meters and the frequency is stable at 42.8 Hz to form an operating parameter record.

[0062] S1061. Through data encoding and classification, stable parameters are generated into time-stamped historical data and stored in a database. Storage integrity is verified to ensure data availability. In this embodiment of the present invention, a data verifier performs multi-dimensional checks on operating parameter records, including whether the values ​​fall within the 15% to 95% liquid level range and the 500 to 2000 rpm speed range, whether the timestamps are continuous, and whether the physical correlation between capacity and frequency is reasonable. After verification, the data encoder adds a 32-bit timestamp to each record, such as 14:25:37 on March 11, 2024, combined with a 16-bit device number and a 32-bit parameter value, to generate standardized historical data records. The data classifier then creates an index table based on parameter type and time series, for example, dividing the time index by hour, generating approximately 360 records per hour. Separate parameter indexes are also created for capacity and frequency. This hierarchical index structure greatly improves data retrieval efficiency; for example, querying capacity change trends for a particular day takes only seconds. The data writer stores the index tables and records in the historical database. The monthly data volume is approximately 120 megabytes, containing 3 million records, which is sufficient to support long-term analysis.

[0063] In actual operation, the sophisticated design of deviation analysis and data storage has significantly improved the reliability of process control. For example, during a single adjustment, the capacity deviation decreased from the initial 2.5% to 0.8%, and the frequency deviation decreased from 1.2% to 0.5%. The mean squared error (RMS) was less than 1%, and the fluctuations stabilized within 5 minutes. The recorded stable parameters showed a capacity of 49.5 cubic meters and a frequency of 42.9 Hz, which were highly consistent with the target values, demonstrating the high efficiency of distributed control. The stored data is not only used to verify the current results but also provides a basis for equipment status monitoring. If the mean squared error of the frequency deviation increases from 0.6% to 0.9% over a long period of time, it may indicate a decrease in pump efficiency and the need for early maintenance.

[0064] The integrity of the storage process is verified by a data verifier, which checks the number of records and index consistency after each write. For example, a single storage of 4,000 records takes approximately 15 seconds, generating a 24-hour index and one daily index with an 80% compression rate, ensuring efficient storage. A data statistics tool generates a storage report, documenting the storage time, data volume, and index structure. For example, storage begins at 15:00:00 on March 11, 2024, with 3,800 records and complete index nodes. This information provides reliable support for data management and process optimization.

[0065] In the embodiments of the present invention, the accumulation of historical data creates conditions for process improvement. By analyzing the deviation trends in different time periods, the response patterns of parameter adjustments can be discovered. For example, if the capacity stabilization time is extended from 8 minutes to 12 minutes, it may reflect a decrease in the sensitivity of the liquid level control loop, and the valve response speed needs to be optimized. Long-term tracking of frequency data can also reveal the wear trend of the pump and provide a decision-making basis for preventive maintenance. The flexibility of the storage system allows the index granularity or threshold setting to be adjusted according to actual needs to adapt to diverse production scenarios.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed control method for preparing waste nitric acid from RDX, characterized in that: The method comprises: The reaction parameters and process parameters of the waste nitric acid preparation process using RDX are obtained in real time, and are associated with the pre-stored historical reaction parameters, process parameters, and historical waste nitric acid characteristic data to form an associated data set. The associated data set is then input into a pre-built waste acid characteristic prediction model for training. The trained waste acid characteristic prediction model is obtained, including: The waste acid concentration, pH, temperature and reactant conversion values ​​are collected through multi-point sensing equipment to obtain the original waste acid monitoring data set with time stamps; Data truncation is performed according to the valid range of the waste acid concentration value, pH value and temperature value in the original waste acid monitoring data set, and the missing data points are interpolated using the sliding average method to obtain a standardized waste acid data set; Performing a Pearson correlation analysis on the waste acid concentration values, pH values, temperature values, and reactant conversion rate values ​​in the standardized waste acid data set, selecting characteristic parameters with correlation coefficients greater than a preset threshold, and obtaining a key feature set of waste acid characteristics; The key feature set of the waste acid characteristics is trained using a random forest algorithm, and the feature weights of the waste acid concentration value, pH value and temperature value are adjusted using a back propagation algorithm to obtain a waste acid characteristic prediction model; The trained waste acid characteristic prediction model is used to predict upstream reaction parameters and process parameters, thereby obtaining the probability of carrying incompletely reacted lead nitrate particles. The predicted probability is compared with a preset threshold. If the predicted probability is greater than the threshold, it is determined that there is a risk of the waste nitric acid carrying incompletely reacted lead nitrate particles, and an early warning signal is triggered. The early warning signal is sent to each monitoring point. Each monitoring point receives the early warning signal as an independent control node and adjusts the collection tank capacity and delivery pump frequency respectively; Sending adjustment instructions for adjusting the collection tank capacity and the delivery pump frequency to the corresponding execution device, and dynamically adjusting the collection tank capacity and the delivery pump frequency according to the adjustment instructions; Collect the adjusted collection tank capacity and delivery pump frequency data in real time, compare the collected data with the pre-set target value, and calculate the deviation between the adjusted collection tank capacity and delivery pump frequency and the target value; If the deviation between the adjusted collection tank capacity and delivery pump frequency and the target value is less than the preset threshold, it is determined that the advance distributed control effect has achieved the expected result, and the adjusted collection tank capacity and delivery pump frequency data are stored in the historical database.

2. The method according to claim 1, characterized in that The trained waste acid characteristic prediction model is used to predict upstream reaction parameters and process parameters, thereby obtaining a probability of carrying incompletely reacted lead nitrate particles. The predicted probability is compared with a preset threshold. If the predicted probability is greater than the threshold, it is determined that there is a risk of the waste nitric acid carrying incompletely reacted lead nitrate particles, and an early warning signal is triggered, including: The output values ​​of the waste acid characteristic prediction model were analyzed using a gradient boosting algorithm to obtain an upstream process prediction parameter set consisting of reactor pressure, stirring rate, feed rate, and reactant ratio. According to the upstream process prediction parameter set, combined with the waste acid viscosity coefficient obtained by the online viscometer and the flow rate data measured by the pipeline flow meter, a waste acid flow characteristic parameter set consisting of the waste acid flow resistance value is calculated; According to the waste acid flow characteristic parameter set and the upstream process prediction parameter set, a fluid dynamics sedimentation model is used to calculate the theoretical sedimentation rate value and spatial distribution density value of the lead nitrate particles; The transmittance data of the waste acid solution is collected by a photoelectric turbidity sensor, and combined with the lead ion concentration value obtained by ion selective electrode detection, the predicted probability value of incompletely reacted lead nitrate particles is calculated. If the predicted probability value exceeds the preset threshold, a digital warning signal is sent to the decentralized controller.

3. The method according to claim 1, characterized in that The warning signal is sent to each monitoring point, and each monitoring point receives the warning signal as an independent control node and adjusts the collection tank capacity and the delivery pump frequency respectively, including: Receiving a warning data packet sent by the optical fiber communication network, wherein the warning data packet carries a monitoring point number and a warning signal value; The liquid level value obtained by the liquid level sensor of the collection tank is collected according to the warning signal value, and the capacity control parameter of the collection tank is obtained by using an adaptive controller; The waste acid flow rate value is collected by the collection tank capacity control parameter, and the delivery pump frequency control parameter is obtained by using a proportional integral algorithm; A closed-loop controller is used to calculate an adjustment deviation value based on the liquid level value of the collection tank and the operating parameters of the delivery pump, and a compensation control instruction is obtained according to the adjustment deviation value.

4. The method according to claim 3, characterized in that Also includes: After analyzing and processing the warning signals at each monitoring point, key information is extracted and the warning level and impact range are evaluated. Based on the warning situation, it is determined whether the collection pool capacity should be adjusted. If adjustment is required, the target capacity value is calculated based on the waste nitric acid generation rate, warning duration, and safety margin. At the same time, it is determined whether the delivery pump frequency should be adjusted. If adjustment is required, the target frequency value is calculated based on the target capacity value, warning duration, delivery pump characteristics, and pipeline resistance. Specifically, the following are included: A signal analyzer is used to digitally process the warning signal of the monitoring point, obtain the monitoring point number, warning time and signal strength value from the warning signal, and obtain the warning level according to the warning classification rules; The current capacity value is obtained by the collection tank capacity sensor, and the real-time generation rate is obtained according to the warning level and the waste acid generation rate detector; Using a capacity safety margin calculator to calculate a required safety capacity value according to the current capacity value and the real-time generation rate, and obtain a capacity adjustment instruction; A regulating instruction package is constructed according to the capacity regulating instruction, the instruction package is sent to the collecting tank liquid level regulating valve, and regulating process parameters are collected through the collecting tank liquid level sensor.

5. The method according to claim 1, wherein The adjusting instructions for adjusting the collection tank capacity and the delivery pump frequency are sent to corresponding execution devices, and the collection tank capacity and the delivery pump frequency are dynamically adjusted according to the adjustment instructions, including: Receive collection tank capacity adjustment instructions and delivery pump frequency adjustment instructions, and use the instruction distributor to generate execution adjustment data packets through the fieldbus network; According to the execution adjustment data packet, a liquid level comparator is used to calculate the difference between the current liquid level and the target liquid level, and an inlet valve opening instruction and an outlet valve opening instruction are obtained through a liquid level valve controller; A frequency calculator is used to generate speed control curve parameters according to the frequency adjustment instruction of the delivery pump, and the speed control data of the delivery pump is obtained through the frequency converter; An execution status recorder is used to collect the water inlet valve opening instruction, the water outlet valve opening instruction and the speed control data to obtain adjustment process data.

6. The method according to claim 1, characterized in that The real-time collection of the adjusted collection tank capacity and delivery pump frequency data, comparing the collected data with a preset target value, and calculating the deviation between the adjusted collection tank capacity and delivery pump frequency and the target value includes: A liquid level sensor is used to obtain liquid level data of the collection tank and a speed sensor is used to obtain speed data of the delivery pump. A data collector records pipeline flow data and pressure data based on the liquid level data and speed data. According to the flow data and pressure data, removing data noise by the signal filter to obtain first processed data, and removing over-limit data points from the first processed data by the data validity judgement device to obtain second processed data; The capacity calculator calculates a real-time capacity value based on the liquid level data in the second processed data, and calculates a real-time frequency value based on the speed data in the second processed data; The value comparator compares the real-time capacity value and the real-time frequency value with a preset target value to obtain deviation data.

7. The method according to claim 1, characterized in that If the deviations between the adjusted collection tank capacity and delivery pump frequency and the target values ​​are less than a preset threshold, it is determined that the advance distributed control effect has achieved the expected result, and the adjusted collection tank capacity and delivery pump frequency data are stored in the historical database, including: The deviation calculator is used to obtain the collection tank capacity deviation and the delivery pump frequency deviation, and the mean square value of the deviation is obtained by the root mean square method; Determine whether the mean square value of the deviation is less than a preset threshold value according to the mean square value of the deviation, and use a stability determiner to obtain a deviation data point for the mean square value of the deviation that is less than the preset threshold value; A parameter collector is used to record the collection tank capacity value and the delivery pump frequency value according to the deviation data point, and a data encoder is used to time-code the capacity value and the frequency value to obtain a historical data record with a time stamp; For the historical data records, a data classifier is used to establish a data index table according to parameter type and time series, and the data index table is stored in a historical database through a data writer.

Images