Gas phase method white carbon black tail gas absorption system and method with online monitoring adjustment function

Abstract

The application provides a fumed white carbon black tail gas absorption system and method with online monitoring and adjusting functions, and the method comprises the following steps: S1: fumed white carbon black tail gas collection and initial parameter detection; S2: tail gas pretreatment; S3: primary absorption; S4: secondary absorption; S5: online monitoring system data integration and threshold setting; S6: automatic adjusting control based on monitoring data; S7: standard tail gas emission; and S8: byproduct collection and treatment. The application is aimed at tail gas containing pollutants such as hydrogen chloride, silicon tetrachloride and silicon dioxide dust generated in the fumed white carbon black production process, and a full-process closed-loop treatment system of collection-pretreatment-two-stage absorption-online regulation-standard emission-byproduct recovery is constructed, and through online monitoring and multi-algorithm fusion adjustment, precise removal of tail gas pollutants and efficient recovery of resources are realized.

Description

Technical Field

[0001] This invention relates to the fields of environmental protection and waste gas treatment technology, and in particular to a gas-phase silica tail gas absorption system and method with online monitoring and regulation functions. Background Technology

[0002] Fumed silica (also known as fumed silicon dioxide) is an inorganic nanomaterial produced by the reaction of silicon tetrachloride and hydrogen at high temperatures. It is widely used in rubber, coatings, cosmetics, and other fields. The production process generates a large amount of exhaust gas, the main component of which is hydrogen chloride (…). ), unreacted silicon tetrachloride ( ),hydrogen( ), nitrogen ( ) and a small amount of silica ( Dust. Among them, hydrogen chloride is a highly corrosive gas, and silicon tetrachloride reacts with water to produce hydrochloric acid and silicon dioxide. If directly emitted, it will seriously pollute the atmosphere, water bodies and soil, endangering the ecological environment and human health, and must be purified.

[0003] Existing exhaust gas absorption technologies mostly employ a two-stage absorption process of "first-stage dilute hydrochloric acid absorption + second-stage water absorption," but this has the following shortcomings:

[0004] 1. Monitoring lag: The detection of parameters such as exhaust gas flow, temperature, and pollutant concentration is mostly offline or intermittently online, which cannot reflect changes in operating conditions in real time, resulting in untimely adjustments;

[0005] 2. Low adjustment accuracy: The adjustment of absorbent spray volume, cooling water volume, etc., relies on manual experience or simple PID control, which is difficult to adapt to the dynamic fluctuations of tail gas parameters (such as a sudden increase in tail gas flow caused by changes in reactor load), and is prone to incomplete absorption (excessive emissions) or waste of absorbent.

[0006] 3. Poor coordination: There is a lack of data linkage between pretreatment (dust filtration, temperature control), absorption, and by-product treatment. For example, the dust filtration effect is not correlated with the load of the absorption tower, which increases the risk of packing blockage.

[0007] 4. Weak adaptability: Traditional control algorithms cannot optimize adjustment strategies based on historical data, and are ineffective in handling complex nonlinear absorption processes (such as...). The coupled reaction of hydrolysis and HCl absorption is not sufficiently adaptable. Therefore, there is an urgent need for a tail gas absorption method that integrates online real-time monitoring, multi-stage data linkage, and intelligent algorithm optimization to solve the problems of poor stability, low efficiency, and insufficient adaptability in existing technologies. Summary of the Invention

[0008] This invention provides a gas phase silica tail gas absorption system and method with online monitoring and adjustment functions. Through the collaborative design of the entire process of "collection-pretreatment-absorption-monitoring-adjustment-emission-byproduct treatment" and combined with intelligent algorithm optimization, it achieves efficient purification of tail gas and resource recovery.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A gas-phase silica tail gas absorption method with online monitoring and adjustment function includes the following steps:

[0011] S1: Collect the exhaust gas generated by the reactor, detect the flow rate, temperature, and pressure data of the exhaust gas, and obtain the initial parameters;

[0012] S2: Filter the silica dust in the exhaust gas and detect the dust concentration. Then, adjust the cooling water flow of the heat exchanger in combination with the exhaust gas temperature in S1 to cool down the filtered exhaust gas and detect the temperature of the cooled exhaust gas.

[0013] S3: The cooled exhaust gas is introduced into the first-stage absorption tower, and dilute hydrochloric acid solution is used as the absorbent to absorb hydrogen chloride in the exhaust gas. The concentration of circulating dilute hydrochloric acid, pH value and residual hydrogen chloride concentration after first-stage absorption are detected.

[0014] S4: The tail gas after primary absorption is introduced into the secondary absorption tower. Water is used as the absorbent to absorb the residual hydrogen chloride and to react the silicon tetrachloride in the tail gas with water. The pH value, silicon content and liquid level of the circulating water and the liquid level of the circulating tank are detected. The residual concentrations of hydrogen chloride and silicon tetrachloride in the tail gas after secondary absorption are detected.

[0015] S5: Based on the initial parameters, the dust concentration and exhaust gas temperature detected by S2, the dilute hydrochloric acid concentration, pH value and residual hydrogen chloride concentration after primary absorption detected by S3, the pH value, silicon content and circulating tank level of the circulating water detected by S4, and the residual concentrations of hydrogen chloride and silicon tetrachloride in the exhaust gas after secondary absorption, real-time monitoring is performed and safety thresholds and exceedance thresholds for the parameters are set. When the parameters reach the exceedance threshold, an audible and visual alarm is triggered.

[0016] S6: Based on the monitoring data of S5, automatic adjustment is achieved by integrating particle swarm optimization algorithm, adaptive neuro-fuzzy inference system, and fuzzy PID control algorithm: Particle swarm optimization algorithm optimizes the membership function parameters of adaptive neuro-fuzzy inference system and fuzzy PID control algorithm; Adaptive neuro-fuzzy inference system predicts ideal adjustment amount based on real-time data of S1-S4, as feedforward correction amount of fuzzy PID control algorithm; Fuzzy PID control algorithm combines the predicted value of adaptive neuro-fuzzy inference system with real-time adjustment error, and outputs adjustment command to corresponding equipment to control the cooling water flow rate of S2, the dilute hydrochloric acid spray rate of S3, and the clean water spray rate of S4; At the same time, the adjustment error of fuzzy PID control algorithm is fed back to particle swarm optimization algorithm to dynamically update optimization parameters;

[0017] S7: The exhaust gas after secondary absorption is introduced into the exhaust gas buffer tank to detect the concentration of hydrogen chloride, silicon tetrachloride and dust in the exhaust gas. If it meets the standard, it is discharged through the chimney. If it does not meet the standard, it is returned to S2 for reprocessing.

[0018] S8: Collect silica dust, concentrated hydrochloric acid, and silicon-containing waste liquid generated during the exhaust gas absorption process, and recycle or treat them respectively.

[0019] In this specification, in step S6, the particle dimension of the particle swarm optimization algorithm is the sum of the parameter dimensions of the adaptive neural fuzzy inference system and the parameter dimensions of the fuzzy PID control algorithm. The parameters of the adaptive neural fuzzy inference system include the membership function centers and widths of the 6 input variables and the output weights of 729 fuzzy rules. The parameters of the fuzzy PID control algorithm include the membership function centers and widths of the error and the error change rate.

[0020] In this specification, in step S6, the input variables of the adaptive neural fuzzy inference system include exhaust gas flow rate, residual hydrogen chloride concentration after primary absorption, residual hydrogen chloride concentration after secondary absorption, pH value of secondary absorption circulating water, silicon content of secondary absorption circulating water, and exhaust gas temperature after cooling. The output variable is the ideal adjustment amount.

[0021] In this specification, in step S6, the inputs to the fuzzy PID control algorithm include the adjustment error and its rate of change, as well as the ideal adjustment amount output by the adaptive neural fuzzy inference system. The PID parameter correction amount is output through the fuzzy rule table, and the corrected PID parameters are combined with the feedforward correction amount to calculate the final adjustment command.

[0022] In this specification, in step S6, the interaction process of the three algorithms is as follows: the optimized parameters output by the particle swarm optimization algorithm are used as the initial parameters of the adaptive neural fuzzy inference system and the fuzzy PID control algorithm, respectively; the predicted value of the adaptive neural fuzzy inference system is used as the feedforward input of the fuzzy PID control algorithm; the adjustment error of the fuzzy PID control algorithm is fed back to the fitness function of the particle swarm optimization algorithm, and the particle swarm optimization algorithm is triggered to iterate and optimize the parameters again every 10 minutes to form a closed loop.

[0023] In this manual, the closing condition is: after adjustment by the fuzzy PID control algorithm, all parameters set in S5 return to the safe threshold and do not exceed the limit for 30 consecutive seconds.

[0024] In this instruction manual, the criterion for determining whether the exhaust gas meets emission standards in step S7 is: hydrogen chloride concentration ≤ 10. Silicon tetrachloride concentration ≤5 Dust concentration ≤5 The exhaust valve is opened when all three conditions are met; otherwise, the return valve is opened to send the exhaust gas back to S2.

[0025] In this manual, step S2, the specific operation of cooling the filtered exhaust gas is as follows: the filtered exhaust gas is introduced into a shell-and-tube heat exchanger, and the cooling water flow rate is adjusted in conjunction with the exhaust gas temperature data in S1 to reduce the exhaust gas temperature to 40-60℃. The cooling water flow rate adjustment logic of the shell-and-tube heat exchanger is as follows: if the exhaust gas temperature detected in S1 is >80℃, the opening of the cooling water flow rate regulating valve is linearly increased according to the difference between the exhaust gas temperature and 80℃; if it is ≤80℃, the valve opening is maintained at 50%.

[0026] In this specification, the treatment process of silicon-containing waste liquid in step S8 is as follows: after being introduced into the waste liquid sedimentation tank, sodium hydroxide solution is added to adjust the pH to 7-8, and after standing for 2 hours, it is filtered through a plate and frame filter press. After the filter cake is dried, it is used as industrial packing. The filtrate is reused or discharged after chloride ions are removed by an ion exchange resin column.

[0027] A gas-phase silica tail gas absorption system with online monitoring and regulation function, employing any one of the above-mentioned gas-phase silica tail gas absorption methods with online monitoring and regulation function, wherein the gas-phase silica tail gas absorption system with online monitoring and regulation function comprises:

[0028] The exhaust gas collection module is used to collect the exhaust gas generated by the reactor, detect the flow rate, temperature, and pressure data of the exhaust gas, and obtain initial parameters.

[0029] The exhaust gas pretreatment module is used to filter silica dust in the exhaust gas and detect the dust concentration. Combined with the exhaust gas temperature adjustment of S1, the cooling water flow of the heat exchanger is adjusted to cool down the filtered exhaust gas and the temperature of the cooled exhaust gas is detected.

[0030] The primary absorption module is used to introduce the cooled exhaust gas into the primary absorption tower, using dilute hydrochloric acid solution as the absorbent to absorb hydrogen chloride in the exhaust gas, and to detect the concentration of circulating dilute hydrochloric acid, pH value and residual hydrogen chloride concentration after primary absorption.

[0031] The secondary absorption module is used to introduce the exhaust gas after the primary absorption into the secondary absorption tower. It uses clean water as the absorbent to absorb the residual hydrogen chloride and reacts the silicon tetrachloride in the exhaust gas with water. It detects the pH value, silicon content and liquid level of the circulating clean water, and detects the residual concentration of hydrogen chloride and silicon tetrachloride in the exhaust gas after the secondary absorption.

[0032] The real-time monitoring module is used to monitor and set safety thresholds and exceedance thresholds for parameters based on initial parameters, dust concentration and exhaust gas temperature after cooling detected by S2, dilute hydrochloric acid concentration, pH value and residual hydrogen chloride concentration after primary absorption detected by S3, pH value, silicon content and circulating tank level of circulating water detected by S4, and residual concentrations of hydrogen chloride and silicon tetrachloride in exhaust gas after secondary absorption. When the parameters reach the exceedance threshold, an audible and visual alarm is triggered.

[0033] The regulation and control module, based on the monitoring data from S5, employs a fusion of particle swarm optimization (PSO) algorithm, adaptive neural fuzzy inference system (ANSI), and fuzzy PID control algorithm to achieve automatic regulation. The PSO algorithm optimizes the membership function parameters of the ASI and fuzzy PID control algorithms. The ASI predicts the ideal regulation amount based on real-time data from S1-S4, serving as the feedforward correction for the fuzzy PID control algorithm. The fuzzy PID control algorithm combines the predicted value from the ASI with the real-time regulation error to output regulation commands to the corresponding devices, controlling the cooling water flow rate in S2, the dilute hydrochloric acid spray rate in S3, and the clean water spray rate in S4. Simultaneously, the regulation error of the fuzzy PID control algorithm is fed back to the PSO algorithm to dynamically update the optimization parameters.

[0034] The exhaust gas emission module is used to introduce the exhaust gas after secondary absorption into the exhaust gas buffer tank, and detect the concentration of hydrogen chloride, silicon tetrachloride and dust in the exhaust gas. If it meets the standard, it is discharged through the chimney; if it does not meet the standard, it is returned to S2 for reprocessing.

[0035] The by-product processing module is used to collect the silica dust generated by filtration, the concentrated hydrochloric acid discharged from S3, and the silicon-containing waste liquid discharged from S4, and to recycle or treat them respectively.

[0036] In summary, the present invention has at least the following beneficial effects:

[0037] 1. Improve exhaust gas purification stability: By real-time monitoring of exhaust gas parameters (flow rate, temperature, pollutant concentration, etc.) and combining with a fusion algorithm to dynamically adjust absorbent dosage, cooling load, etc., the concentrations of hydrogen chloride, silicon tetrachloride, and dust in the exhaust gas are kept stable within the emission standards, avoiding intermittent excessive emissions;

[0038] 2. Improve absorption efficiency and resource utilization: By precisely adjusting parameters such as absorbent spray volume and circulating liquid concentration, excessive consumption of absorbent (dilute hydrochloric acid, water) is reduced; at the same time, by-products (silica dust, concentrated hydrochloric acid, silica-containing filter cake) are recycled in a targeted manner to "turn waste into treasure" and reduce environmental treatment costs.

[0039] 2. Enhanced system adaptability: The fusion algorithm (particle swarm optimization, adaptive neural fuzzy inference, fuzzy PID) can autonomously optimize the adjustment strategy according to the fluctuation of exhaust gas parameters, adapt to complex working conditions such as changes in reactor load and fluctuations in raw material purity, and reduce manual intervention;

[0040] 3. Ensure safe operation of equipment: By monitoring parameters such as dust concentration after filtration, liquid level in the circulating tank, and pH value of the absorbent in real time, actions such as backflushing, draining, and replenishing are triggered in a timely manner to avoid equipment blockage, corrosion, or overload, thus extending the service life of the equipment. Attached Figure Description

[0041] Figure 1 This is a schematic flowchart of the gas-phase silica tail gas absorption method with online monitoring and adjustment function involved in this invention.

[0042] Figure 2 This is a schematic diagram of the two-stage absorption process involved in this invention.

[0043] Figure 3 This is a schematic diagram of the automatic adjustment and control process involved in this invention.

[0044] Figure 4 This is a schematic diagram of the process for achieving compliant exhaust emissions and by-product treatment involved in this invention. Detailed Implementation

[0045] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0046] like Figure 1 As shown, this embodiment provides a gas-phase silica tail gas absorption method with online monitoring and adjustment function, including the following steps:

[0047] S1: Collect the exhaust gas generated by the reactor, detect the flow rate, temperature, and pressure data of the exhaust gas, and obtain the initial parameters;

[0048] S2: Filter the silica dust in the exhaust gas and detect the dust concentration. Then, adjust the cooling water flow of the heat exchanger in combination with the exhaust gas temperature in S1 to cool down the filtered exhaust gas and detect the temperature of the cooled exhaust gas.

[0049] S3: The cooled exhaust gas is introduced into the first-stage absorption tower, and dilute hydrochloric acid solution is used as the absorbent to absorb hydrogen chloride in the exhaust gas. The concentration of circulating dilute hydrochloric acid, pH value and residual hydrogen chloride concentration after first-stage absorption are detected.

[0050] S4: The tail gas after primary absorption is introduced into the secondary absorption tower. Water is used as the absorbent to absorb the residual hydrogen chloride and to react the silicon tetrachloride in the tail gas with water. The pH value, silicon content and liquid level of the circulating water and the liquid level of the circulating tank are detected. The residual concentrations of hydrogen chloride and silicon tetrachloride in the tail gas after secondary absorption are detected.

[0051] S5: Based on the initial parameters, the dust concentration and exhaust gas temperature detected by S2, the dilute hydrochloric acid concentration, pH value and residual hydrogen chloride concentration after primary absorption detected by S3, the pH value, silicon content and circulating tank level of the circulating water detected by S4, and the residual concentrations of hydrogen chloride and silicon tetrachloride in the exhaust gas after secondary absorption, real-time monitoring is performed and safety thresholds and exceedance thresholds for the parameters are set. When the parameters reach the exceedance threshold, an audible and visual alarm is triggered.

[0052] S6: Based on the monitoring data of S5, automatic adjustment is achieved by integrating particle swarm optimization algorithm, adaptive neuro-fuzzy inference system, and fuzzy PID control algorithm: Particle swarm optimization algorithm optimizes the membership function parameters of adaptive neuro-fuzzy inference system and fuzzy PID control algorithm; Adaptive neuro-fuzzy inference system predicts ideal adjustment amount based on real-time data of S1-S4, as feedforward correction amount of fuzzy PID control algorithm; Fuzzy PID control algorithm combines the predicted value of adaptive neuro-fuzzy inference system with real-time adjustment error, and outputs adjustment command to corresponding equipment to control the cooling water flow rate of S2, the dilute hydrochloric acid spray rate of S3, and the clean water spray rate of S4; At the same time, the adjustment error of fuzzy PID control algorithm is fed back to particle swarm optimization algorithm to dynamically update optimization parameters;

[0053] S7: The exhaust gas after secondary absorption is introduced into the exhaust gas buffer tank to detect the concentration of hydrogen chloride, silicon tetrachloride and dust in the exhaust gas. If it meets the standard, it is discharged through the chimney. If it does not meet the standard, it is returned to S2 for reprocessing.

[0054] S8: Collect silica dust, concentrated hydrochloric acid, and silicon-containing waste liquid generated during the exhaust gas absorption process, and recycle or treat them respectively.

[0055] The technical concept of this invention is as follows:

[0056] The solution uses a central control system as its core hub. First, the corrosion-resistant gas collection device and sensors in S1 collect initial parameters such as exhaust gas flow, temperature, and pressure in real time; then, the exhaust gas is filtered and removed by bag filters in S2. Dust is cooled using a shell-and-tube heat exchanger, laying the foundation for subsequent absorption; then, 60%–70% of the dust is initially removed through the S3 primary absorption tower (using dilute hydrochloric acid as the absorbent). Simultaneously, concentrated hydrochloric acid discharge and replenishment control is implemented (when the concentration of circulating dilute hydrochloric acid > 25%, the liquid is discharged to S8 for recovery, and the prepared solution is replenished simultaneously to maintain a concentration of 15%–25%); the S4 secondary absorption tower (using industrial water as the absorbent) deeply removes residual substances. and make Reaction with water to produce Sedimentation further reduces pollutant concentration; S5 integrates all process parameters, sets safety / exceedance thresholds, and triggers abnormal alarms.

[0057] The core innovation of the solution lies in the multi-algorithm fusion adjustment of S6: the particle swarm optimization algorithm optimizes the core parameters of the adaptive neurofuzzy inference system and the fuzzy PID control algorithm; the adaptive neurofuzzy inference system is based on real-time data from S1-S4 (such as exhaust gas flow rate, etc.). The ideal regulating value is predicted by the concentration and pH value of the circulating liquid; the fuzzy PID control algorithm combines this predicted value with the real-time error (such as concentration and pH value of the circulating liquid). (Concentration and threshold deviation), output pump frequency, valve opening and other commands; at the same time, the fuzzy PID control algorithm adjusts the error feedback to the particle swarm optimization algorithm for dynamic optimization, ensuring that the parameters are stable within the safe range;

[0058] Ultimately, S7 is discharged only after its exhaust gas meets the standards, verified by a buffer tank and integrated testing instrument; exhaust gas that does not meet the standards is returned for retreatment; S8 recovers the exhaust gas from S2. Dust, concentrated hydrochloric acid (S3), and substances containing S4 The filter cake achieves the dual goals of resource reuse and environmental compliance, and is suitable for tail gas treatment scenarios in the large-scale production of fumed silica.

[0059] The details are as follows:

[0060] S1: Collection of tail gas and detection of initial parameters for fumed silica production

[0061] In the fumed silica production process, silicon tetrachloride reacts with hydrogen at high temperatures in a reactor to produce silicon dioxide (silicon dioxide), while simultaneously generating hydrogen chloride (SiCl₂). silicon tetrachloride ( ),hydrogen( ), nitrogen ( ) and a small amount of silica ( Dust-laden exhaust gas. To prevent exhaust gas leaks and pollution, key parameters need to be collected and tested using specialized equipment to provide basic data for subsequent treatment.

[0062] 1. Deployment of exhaust gas collection device: Install a corrosion-resistant gas collection hood (made of Hastelloy alloy, resistant to corrosion) at the end of the reactor outlet flue. (Corrosion), the gas collection hood and flue are connected by a flange seal (the gasket is made of polytetrafluoroethylene to ensure airtightness). The outlet of the gas collection hood is connected to a sealed delivery pipe (made of 316L stainless steel, with an inner diameter of 500mm designed according to the maximum exhaust gas flow rate). The pipe is fully insulated (using aluminum silicate cotton, 50mm thick) to prevent corrosion in the exhaust gas. The pipes were blocked due to premature condensation caused by a sudden drop in temperature.

[0063] 2. Initial Parameter Online Detection: At a distance of 1.5m from the outlet of the gas collection hood on the delivery pipeline, three types of sensors are installed radially along the pipeline: an online gas flow meter (model: MF-800, accuracy ±1%FS) to detect the exhaust gas volumetric flow rate, defined as... (unit: / h);

[0064] An online temperature sensor (model: TT-300, platinum resistance thermometer Pt100, measurement range 0-200℃, accuracy ±0.5℃) is used to detect exhaust gas temperature and is defined as... (Unit: °C); Online pressure sensor (Model: PT-500, measuring range -50 to 50 kPa, accuracy ±0.2 kPa), detects exhaust gas pressure in pipelines, defined as... (Unit: kPa).

[0065] 3. Data transmission to the central control system: Sensor detection data is transmitted in real time to the central control system (hardware: Siemens S7-1200 PLC, equipped with a 128GB data storage module) via industrial Ethernet (using Profinet protocol, transmission rate 100Mbps). This data is the core basis for subsequent preprocessing, absorption, and adjustment—for example, This directly affects the calculation of the absorbent spray volume. Determine the cooling load of the heat exchanger. Used to determine if a pipe is blocked (a sudden increase in pressure indicates a risk of blockage).

[0066] S2: Exhaust gas pretreatment (dust filtration and temperature control)

[0067] exhaust gas If dust enters the absorption tower, it will adhere to the surface of the packing material, reducing absorption efficiency, and the high-temperature exhaust gas will accelerate the evaporation of the absorbent (such as the evaporation of water in dilute hydrochloric acid). Therefore, it is necessary to remove the dust first and adjust the temperature to a suitable range. The pre-processed parameters will be used as input variables for the S6 algorithm.

[0068] 1. Dust filtration: The exhaust gas collected by S1 is introduced into a corrosion-resistant bag filter (shell material: 316L stainless steel, internally equipped with 10 filter bag units). The filter bag material is polytetrafluoroethylene (PTFE), with a filtration accuracy of 1μm (capable of intercepting 99.9% of dust particles larger than 1μm). An automatic backflushing device is installed on the top of the filter (backflushing medium is dry compressed nitrogen, pressure: 0.6MPa). The default backflushing cycle is 30 minutes / time. If S6 detects that the dust concentration exceeds the standard, it can trigger immediate backflushing.

[0069] Install an online dust concentration detector (model: DM-600, laser scattering method, measurement range 0-10) on the filter outlet pipeline. Accuracy ±0.1 The dust concentration in the filtered exhaust gas is defined as... (unit: The data is transmitted to the central control system in real time.

[0070] 2. Exhaust gas temperature regulation: The filtered exhaust gas enters a shell-and-tube heat exchanger (heat exchange area 20... The tube side is made of titanium alloy, and the shell side uses industrial cooling water. Through heat exchange, the exhaust gas temperature is reduced to 40-60℃ (this range can reduce...). Evaporation, while ensuring (Absorption efficiency).

[0071] An online cooling water flow regulating valve (model: V-400, adjustment range 0-50) is installed on the inlet cooling water pipeline of the heat exchanger. (Accuracy ±1%), valve opening degree and S1 transmission Linkage: If >80℃, the central control system increases the valve opening (opening and...) Linear relationship: (unit: %), increase cooling water flow (maximum 50). / h); if The valve opening is maintained at 50%, and the cooling water flow rate is 25. / h. An online temperature sensor (same as the TT-300 of S1) is installed at the heat exchanger outlet to detect the temperature of the exhaust gas after cooling, defined as... (Unit: °C) The data is fed back to the central control system in real time and serves as the input variable for the adaptive neurofuzzy inference system in S6. .

[0072] S3: Primary absorption (initial absorption of hydrogen chloride)

[0073] exhaust gas Concentrations are typically as high as 1000-3000 Preliminary removal is required first using dilute hydrochloric acid absorbent (utilizing the common ion effect to reduce hydrogen chloride solubility and decrease absorbent consumption). Key parameters in the absorption process will provide a basis for adjusting S6. The two-stage absorption process is as follows: Figure 2 As shown.

[0074] 1. Absorption tower structure and operation: the exhaust gas after cooling ( (40-60℃) Enters from the bottom into the primary packed absorption tower (tower height 8m, diameter 1.2m, material 316L stainless steel), which is filled with polypropylene corrugated packing (model BX500, specific surface area 500). (90% porosity) to increase the gas-liquid contact area.

[0075] The absorbent is dilute hydrochloric acid with a mass concentration of 15%–20%, which is circulated from a primary absorption tank (volume 5). (with stirring device) is circulated by a primary circulation pump (model: P-100, flow rate 0-100). (Voltage controlled) The gas is fed to the top spray device (using spiral nozzles with atomized particle size of 100-200μm) to form a liquid film that comes into countercurrent contact with the exhaust gas, absorbing 60% to 70% of the exhaust gas. .

[0076] 2. Key Parameter Detection and Transmission: An online hydrochloric acid concentration detector (model: CM-200, refractive index method, measurement range 10%~30%, accuracy ±0.5%) is installed on the outlet pipeline of the primary absorption circulation tank to detect the concentration of the circulating liquid, defined as... (Unit: %); An online pH meter (model: PH-300, measuring range 0-7, accuracy ±0.02pH) is installed on the same pipeline to detect the pH value of the circulating liquid, defined as... ;

[0077] Online installation of tail gas outlet pipeline at the top of the primary absorption tower Concentration detector (Model: HC-500, UV absorption method, measurement range 0-1000) Accuracy ±5 ), detect the exhaust gas after absorption Residual concentration, defined as (Unit: mg / m³).

[0078] Central control system receiving After the data is collected, it is compared with the safety threshold (15%~25%) set by S5: If If the concentration exceeds 25% (exceeding the threshold), immediately start the discharge pump of the primary absorption circulation tank (model: P-110, made of corrosion-resistant fluoroplastic material, flow rate 0-10). A portion of the concentrated hydrochloric acid is transported to the hydrochloric acid storage tank in S8 via a corrosion-resistant pipeline; simultaneously, the dilute hydrochloric acid replenishment pump (model: P-120, flow rate 0-5) is started. Add a 31% (w / w) mixture of industrial hydrochloric acid and water to the circulating tank until... Restored to the safe threshold of 15% to 25%.

[0079] The above parameters ( , , The data is transmitted in real time to the central control system, whereby... As input variables of ANFIS in S6 , (From S1) and Together they are used in S6 to determine whether the frequency of the primary circulating pump needs to be adjusted.

[0080] S4: Secondary absorption (deep absorption of hydrogen chloride and treatment with silicon tetrachloride)

[0081] Even after primary absorption, 30%–40% of the exhaust gas still remains. and a small amount (Boiling point 57.6℃, readily soluble in water and reacts to form) and This requires deep absorption with clean water to ensure that pollutants meet standards. The parameters for this step are the core inputs for S6 regulation.

[0082] 1. Absorption Tower Structure and Reaction Mechanism: The tail gas after primary absorption enters the secondary packed absorption tower (10m high, 1.5m in diameter, made of 316L stainless steel) from the bottom. The tower is filled with the same polypropylene corrugated packing (BX500) as the primary tower. The absorbent is industrial clean water (conductivity ≤50μS / cm), which is circulated by the secondary absorption tank (volume 8). (with a bottom conical sedimentation zone) is circulated by a two-stage circulation pump (model: P-200, flow rate 0-150). (Voltage controlled) The gas is fed to the top spray device of the tower and comes into countercurrent contact with the exhaust gas.

[0083] Residue It is absorbed by water, and the reaction is as follows: ; Reaction with water: , generated It settles at the bottom of the circulation tank.

[0084] 2. Key parameter detection and transmission:

[0085] An online pH meter (same as PH-300) is installed on the outlet pipeline of the secondary absorption circulation tank to detect the pH value of the circulating liquid, which is defined as... ;

[0086] An online silicon content analyzer (model: Si-400, molybdenum blue spectrophotometry, measurement range 0-10 g / L, accuracy ±0.1 g / L) is installed in the pipeline to detect the mass concentration of Si in the circulating fluid, defined as... (Unit: g / L);

[0087] An online level gauge (model: LT-300, ultrasonic, measuring range 0-2m, accuracy ±0.01m) is installed on the top of the circulating tank to detect the liquid level height, defined as... (Unit: m);

[0088] Online installation of tail gas outlet pipeline at the top of the secondary absorption tower Concentration detector (same as HC-500) and online Concentration detector (Model: SC-600, gas chromatography, measurement range 0-50) Accuracy ±0.5 ), respectively detect the exhaust gas Final concentration (unit: )and Residual concentration (unit: ).

[0089] The above parameters ( , , , , The data is transmitted in real time to the central control system, whereby... As input variables of ANFIS , As , As This is the core basis for S6 to adjust the frequency of the secondary circulation pump and replenish clean water.

[0090] S5: Online Monitoring System Data Integration and Threshold Setting

[0091] The central control system needs to integrate all real-time data from S1 to S4, establish a monitoring interface, and set safety thresholds and exceedance thresholds for each parameter to provide judgment criteria for the algorithm adjustment of S6.

[0092] 1. Data Integration: The central control system (equipped with WinCC monitoring software) integrates the following data into a dynamic trend curve (update cycle 1 second):

[0093] S1: , , S2: , S3: , , S4: , , , , The interface simultaneously displays the operating status of each device (such as pump frequency and valve opening), and supports historical data query (storage period of 1 year).

[0094] 2. Threshold Setting Basis and Specific Values: The thresholds are set based on the comprehensive emission standards for air pollutants and the requirements for safe operation of equipment.

[0095] Safety threshold ≤ 5 (To avoid packing blockage), threshold > 5 ; Safety threshold: 40-60℃ (to ensure absorption efficiency); exceedance threshold: <40℃. (decreased solubility) or >60℃ (aggravated evaporation of absorbent); The safety threshold is 15%–25% (too low a concentration will result in insufficient absorption, while too high a concentration will lead to…). (Volatile), exceeding the standard threshold by <15% or >25%; Safety threshold ≤ 50 (Reduce secondary absorption load), exceeding the threshold > 50 ; Safety threshold 2-5 (too strong acidity corrodes equipment, too weak acidity...) (decreased absorption capacity), exceeding the threshold by <2 or >5; Safety threshold ≤ 5g / L (avoid) Excessive sedimentation can clog pipes, exceeding the standard threshold of >5g / L; Safety threshold: 0.5-1.5m (below 0.5m, evacuation is easy; above 1.5m, overflow is easy); exceeding threshold: <0.5m or >1.5m. : Qualification threshold ≤ 10 (Meets national emission standards), exceeding the threshold by more than 10. ; : Qualification threshold ≤ 5 (Referencing similar enterprise standards), the threshold for exceeding the standard is >5. .

[0096] 3. Alarm Mechanism: When any parameter reaches the threshold exceeding the limit, the central control system triggers an audible and visual alarm (control cabinet buzzer + red warning light), and the monitoring interface highlights the parameter exceeding the limit and the detection location (e.g., the outlet of the secondary absorption tower). (If the concentration exceeds the standard), the alarm time and parameter value are recorded simultaneously to provide a trigger signal for the adjustment of S6.

[0097] S6: Automatic adjustment and control based on monitoring data

[0098] In the tail gas absorption system of fumed silica, parameters such as tail gas composition (e.g., hydrogen chloride, silicon tetrachloride), flow rate, and temperature often change dynamically due to fluctuations in reactor operating conditions. Traditional single control algorithms struggle to balance adjustment accuracy and response speed. This step integrates three algorithms—fuzzy PID control, adaptive neural fuzzy inference system, and particle swarm optimization—to construct a closed-loop control system encompassing parameter optimization, trend prediction, precise adjustment, and dynamic feedback.

[0099] Based on the threshold values ​​of each parameter set in S5 (e.g., hydrogen chloride concentration ≤ 10 after secondary absorption) With the goal of receiving real-time monitoring data (such as exhaust gas flow rate, absorbent pH value, etc.) from S1 to S4.

[0100] The core parameters of the other two algorithms are optimized using the particle swarm optimization algorithm to improve model adaptability;

[0101] The adaptive neurofuzzy inference system predicts the ideal adjustment amount based on real-time data, reducing adjustment lag.

[0102] The fuzzy PID control algorithm combines predicted values ​​and real-time errors to output the final adjustment command (such as pump frequency and valve opening), and feeds the adjustment effect back to the particle swarm optimization algorithm to achieve dynamic iteration.

[0103] Automatic adjustment and control process such as Figure 3 As shown.

[0104] Algorithm 1: Particle Swarm Optimization Algorithm – Global Parameter Optimizer

[0105] Core function: By simulating the optimization behavior of swarm particles, it provides optimal initial parameters (such as membership function centers and fuzzy rule weights) for fuzzy PID control algorithms and adaptive neurofuzzy inference systems, solving the problems of low accuracy and poor adaptability of manually tuned parameters, and ensuring that the two algorithms can still work stably when exhaust gas parameters fluctuate.

[0106] 1.1 Model Construction: Particle and Parameter Space Definition

[0107] The core of the particle swarm optimization algorithm is to abstract the "parameters to be optimized" into "particles" and find the optimal combination of parameters by updating the positions of the particles.

[0108] Particle dimension definition: Each particle corresponds to a set of parameters to be optimized, and the dimension is the sum of the parameters of the two algorithms, i.e.: ;

[0109] The parameter dimensions of the adaptive neural fuzzy inference system (including the membership function centers of 6 input variables) ,width and the output weights of 729 fuzzy rules The total dimensions are 6×3×2+729×7=5199, where "6" represents the number of input variables, "3" represents the number of fuzzy sets for each variable, "2" represents the center and width, and "7" represents the number of output parameters for each rule.

[0110] The parameter dimensions of the fuzzy PID control algorithm (including error e, error rate of change) Membership function center ,width The total dimensions are 2 × 7 × 2 = 28, where "2" represents the number of input variables, "7" represents the number of fuzzy sets for each variable, and "2" represents the center and width.

[0111] Particle position and velocity:

[0112] Particle position : The parameter combination of the i-th particle, where The specific value of the k-th parameter (e.g.) (can represent the center of the first fuzzy set of the first input variable in an adaptive neural fuzzy inference system); particle velocity. The update step size of particle parameters determines the magnitude of parameter adjustment. Optimal position definition: the optimal position of an individual particle. The optimal combination of parameters for the i-th particle from the initial iteration to the current iteration; the global optimal position. The optimal parameter combination in history among all particles, i.e., the optimized parameters of the final output.

[0113] 1.2 Model Training: Parameter Optimization Based on Historical Data

[0114] The core of training is to find the parameter combination that minimizes the system's regulation error through iterative optimization. The specific steps are as follows:

[0115] 1. Training sample selection: Collect system operation data from the past 3 months (1000 sets in total). Each set of samples includes:

[0116] Input: Real-time monitoring data of S1-S4 (6 parameters including exhaust gas flow rate and hydrogen chloride concentration after primary absorption);

[0117] Output: The ideal adjustment value at the corresponding moment (such as the frequency of the secondary circulation pump, which is obtained from the optimal record of manual debugging and the parameter statistics when the system is stable).

[0118] 2. Fitness Function Design: The goal is to minimize the combined error of the two algorithms. The formula is as follows:

[0119] ;

[0120] Prediction error of the adaptive neural fuzzy inference system (N=1000 is the number of samples). The adjustment amount predicted by the algorithm. (The ideal adjustment amount in the sample).

[0121] Output error of fuzzy PID control algorithm ( (The adjustment amount output by the algorithm).

[0122] Weighting coefficients (prioritize prediction accuracy and reduce subsequent adjustment burden).

[0123] 3. Particle iterative update:

[0124] Initialization: The number of particles is set to 50 (balancing optimization efficiency and accuracy), and the number of iterations is 100; initial positions are randomly generated. (Parameter values ​​are within a physically reasonable range, such as the membership function center being within the threshold range of the monitored parameters) and initial velocity. (The value is 10% of the parameter range).

[0125] Speed ​​update formula: ;

[0126] Inertia weight (initially 0.9, decreasing linearly to 0.4 with each iteration; early stage for global exploration, later stage for local refinement). Learning factor (guides particles toward individual and global optima). Random numbers (to increase the diversity of optimization);

[0127] t: Current iteration number.

[0128] Position update formula: If the updated parameters exceed a reasonable range, force truncation to the boundary value.

[0129] 4. Optimal position update: After each iteration, calculate the fitness of each particle. If it is less than the individual's optimal fitness, then update. If the fitness is less than the global optimal fitness, then update. After the iteration is complete, These are the optimal initial parameters for both algorithms.

[0130] 1.3 Model Application: Dynamic Parameter Correction

[0131] After training, the global optimal position It is broken down into two parts:

[0132] Parameters of the adaptive neural fuzzy inference system: Used to initialize its membership function and fuzzy rules; fuzzy PID control algorithm parameters: This is used to initialize its membership function. Simultaneously, the system collects the real-time adjustment error of the fuzzy PID control algorithm every 10 minutes (such as the deviation between the hydrogen chloride concentration after secondary absorption and the target value), updating the fitness function accordingly. This triggers a new iteration of the particle swarm optimization algorithm (reducing the number of iterations to 20, improving real-time performance), dynamically adjusting parameters to adapt to changes in operating conditions.

[0133] Algorithm 2: Adaptive Neural Fuzzy Inference System – Adjustment Trend Predictor

[0134] Core function: It integrates the self-learning ability of neural networks with the nonlinear mapping ability of fuzzy logic, and predicts the "ideal adjustment amount" (such as the frequency of circulating pump and valve opening) in advance based on the real-time monitoring data of S1-S4, so as to provide feedforward signals for fuzzy PID control algorithm and reduce overshoot problems caused by system lag.

[0135] 2.1 Model Construction: Input, Output, and Fuzzy Rule Design

[0136] The model is based on the nonlinear relationship between "exhaust gas parameters and adjustment amount", and its structure is as follows:

[0137] Input variables (6, all from real-time data from S1-S4): Exhaust gas flow rate (unit: (Online gas flow meter from S1); Hydrogen chloride concentration after primary absorption (unit: (from an online HCl concentration detector from S3). Hydrogen chloride concentration after secondary absorption (unit: (from the online HCl concentration detector of S4). pH value of the secondary absorption circulating solution (from the online pH meter in S4); Silicon content in the secondary absorption circulating liquid (unit: g / L, from the online silicon content analyzer in S4); : Exhaust gas temperature after cooling (unit: °C, from the online temperature sensor at the heat exchanger outlet of S2).

[0138] Output variables:

[0139] Ideal adjustment (taking "secondary circulation pump frequency correction value" as an example, unit: Hz, which is the value that needs to be increased or decreased based on the current frequency).

[0140] Fuzzy sets and membership functions:

[0141] Each input variable is divided into 3 fuzzy sets: {low (L), medium (M), high (H)}, for example:

[0142] Fuzzy set of (hydrogen chloride concentration after primary absorption): low (≤30) ), middle (30-70) ), high (≥70) The membership function uses a Gaussian function (with parameters optimized by the particle swarm optimization algorithm).

[0143] ;

[0144] : The j-th fuzzy set of the i-th input variable (i=1...6, j=1...3, corresponding to L, M, and H respectively);

[0145] : Center of Gaussian function (e.g. The fuzzy set center of the "middle" =50); : Gaussian function width (e.g. The width of the fuzzy set in the middle =10).

[0146] Fuzzy rule base: Total Based on expert experience and historical data, some typical rules are as follows: 1. If (Traffic) is high, and (Primary HCl) is high, and (Level 2) If it is high, then (Pump frequency correction value) High; 2. If (Traffic) is low, and (pH value) low, and High silicon content means (Pump frequency correction value); 3. If (Temperature) High, and (Level 2) If it is low, then (Pump frequency correction value) is low.

[0147] 2.2 Model Training: Parameter Optimization Based on Hybrid Learning Algorithm

[0148] The training objective is to adjust the membership function parameters. , and rule output weights , so that the predicted value Approaching the ideal adjustment amount The specific process is as follows:

[0149] 1. Fuzzification (First Layer): Calculate the membership degree of each input variable to each fuzzy set, for example... At that time, the membership degree of the fuzzy set belonging to the "middle" The membership degree of a "high" fuzzy set .

[0150] 2. Rule Strength Calculation (Second Layer): The strength of the k-th rule is the product of the input membership degrees (representing the degree of rule triggering): ;

[0151] For example, for rule 1 ( high, high, (High), if , , If the membership degree of other inputs is 1 (default "medium"), then... .

[0152] 3. Normalized Strength (Third Layer): Standardizes the strength of the rules, eliminating differences in magnitude. ;

[0153] 4. Output Calculation (Fourth Layer): The output of each rule is a linear combination of the input variables, and the total output is a weighted sum: ;in The output parameters for the k-th rule (i.e., the parameters optimized by the particle swarm optimization algorithm) ).

[0154] 5. Parameter optimization: Parameters were obtained using the particle swarm optimization algorithm. , , As initial values, the output parameters are then fine-tuned using the least squares method. , This causes prediction error Minimum.

[0155] 2.3 Model Application: Real-time Adjustment Quantity Prediction

[0156] The system collects 6 input variables in real time (sampling interval of 1 second). Substitute the values ​​into the trained model: 1. Calculate the membership degree of each input variable according to step 2.2. 2. Calculate the strength of the 729 rules. and normalized intensity 3. Obtain the output formula. (For example, if it is predicted that the frequency of the secondary circulation pump needs to be increased by 3Hz), it is used as the feedforward input of the fuzzy PID control algorithm.

[0157] Algorithm 3: Fuzzy PID Control Algorithm – Final Regulator

[0158] Core function: By combining the predicted values ​​of the adaptive neurofuzzy inference system with the real-time adjustment error, the proportional gain of the PID controller is dynamically corrected. ),integral( ),differential( The parameters are set to output precise adjustment commands (such as pump frequency and valve opening) to ensure that exhaust gas parameters are quickly stabilized within safe thresholds.

[0159] 3.1 Model Construction: Input, Output, and Fuzzy Rule Design

[0160] The model uses "error-parameter correction-adjustment" as its logical chain, and its structure is as follows:

[0161] Input variable: Adjustment error ( The target threshold is, for example, the hydrogen chloride concentration after secondary absorption ≤ 10. ; For real-time detection values, from S4 online Concentration detector); Error change rate ( =1s, reflecting the trend of error change); feedforward correction amount (Predicted values ​​from an adaptive neurofuzzy inference system).

[0162] Output variable: PID parameter correction amount , , (Used to correct proportional, integral, and derivative coefficients respectively); final adjustment command (e.g., the frequency of a secondary circulation pump, in Hz).

[0163] Fuzzy sets and membership functions:

[0164] and Each is divided into 7 fuzzy sets: {NB (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), PB (positive large)}, for example:

[0165] Fuzzy sets: NB (≤-10), NM (-10~-5), NS (-5~-2), ZO (-2~2), PS (2~5), PM (5~10), PB (≥10) (unit: mg / m³);

[0166] The membership function uses a trigonometric wave function (with parameters optimized by the particle swarm optimization algorithm):

[0167] ;

[0168] : The m-th fuzzy set of error e (m=1...7, corresponding to NB to PB); : Center of the triangular wave (e.g. (PS) center =3.5); , The center of adjacent fuzzy sets (e.g.) left neighbor =0, right neighbor =6.5).

[0169] Fuzzy rule table: Based on PID parameter tuning experience, 7×7=49 rules were formulated (some are shown below):

[0170]

[0171] Rule logic: When the error is large and changes slowly, increase the proportional coefficient to quickly reduce the error; when the error is small, decrease the integral coefficient to avoid overshoot.

[0172] 3.2 Model Training: PID Parameter Initialization and Optimization

[0173] 1. Initial PID parameter settings: Based on system characteristics, the initial parameters are set to... =5.0 (proportioning factor) =0.1 (integral coefficient) =0.5 (differential coefficient).

[0174] 2. Membership function optimization: Optimize the center of the triangular wave using the particle swarm optimization algorithm. and width ( The goal is to minimize the root mean square of the adjustment error. ).

[0175] 3.3 Model Application: Real-time Adjustment Command Output

[0176] 1. Parameter Correction: Calculate real-time error and error change rate ; blurring to obtain and The fuzzy set to which it belongs (e.g.) It belongs to Photoshop. Belongs to PS); Query the fuzzy rule table to obtain the parameter correction amount (e.g. =0.5, =0.02, =0);

[0177] Corrected parameters: ;

[0178] 2. Adjustment command calculation: Combining the feedforward correction, the final adjustment command is: ;

[0179] For example: if , , , The integral term = 5 (cumulative error). , ,but: ;

[0180] 3. Execution and Feedback: Adjustment Instructions The signal is sent to the frequency converter of the secondary circulation pump, and the pump frequency is adjusted to 32.1Hz; a new signal is detected after 1 second. The new error is calculated and fed back to the particle swarm optimization algorithm for parameter updates.

[0181] Algorithm fusion and interaction process: Taking the excessive concentration of hydrogen chloride after secondary absorption as an example

[0182] Assume S5 monitors the "hydrogen chloride concentration after secondary absorption" (Exceeding the standard, target) The collaborative process of the three algorithms is as follows:

[0183] 1. Data Input: The central control system inputs real-time data from S1-S4 ( , , , , L, Synchronous input adaptive neural fuzzy inference system and fuzzy PID control algorithm.

[0184] 2. The particle swarm optimization algorithm provides initial parameters: the optimized adaptive neurofuzzy inference system has been obtained. ( (the fuzzy center of the "middle") Fuzzy PID control algorithm (e's PS center).

[0185] 3. Adaptive Neural Fuzzy Reasoning System Prediction: Calculation Membership degree: (middle), (High); Trigger rule 1 (High traffic, Level 1) High and Level 2 High), rule strength Output predicted value (It is recommended to increase the pump frequency by 3Hz).

[0186] 4. Fuzzy PID control algorithm adjustment: calculation (PS) (PS); The rule table is consulted. , After correction , ; Calculate adjustment instructions It is sent to the secondary circulation pump.

[0187] 5. Feedback and Optimization: After 10 minutes, S4 detected... (Meets the standard) The adjustment error of the fuzzy PID is reduced, and the particle swarm optimization algorithm updates the parameters according to the new error, thereby improving the adjustment accuracy in the next adjustment.

[0188] The core value of algorithm fusion

[0189] The three algorithms achieve three major improvements through closed-loop collaboration: "parameter optimization using particle swarm optimization algorithm → trend prediction using adaptive neuro-fuzzy inference system → precise execution using fuzzy PID control algorithm → iterative optimization using error feedback": 1. Response speed: Feedforward prediction reduces the regulation lag time from 5 seconds to 1 second; 2. Regulation accuracy: The fluctuation range of hydrogen chloride concentration after secondary absorption is reduced from ±5... Reduced to ±1m 3. Adaptability: Under operating conditions where exhaust gas flow fluctuates by ±30%, it can still stably control parameters within the safe threshold, solving the problems of easy overshooting and difficulty in stabilization of traditional control algorithms.

[0190] This integrated system ensures that the S6 regulation commands can be precisely applied to the absorption processes of S3 and S4, providing a core guarantee for S7 to meet emission standards.

[0191] S7: Meets exhaust emission standards

[0192] The exhaust gas after secondary absorption needs to be tested again to confirm compliance before being discharged. Exhaust gas that fails to meet standards must be recycled and retreated to ensure emissions comply with environmental requirements. The process for compliant exhaust gas emission and byproduct treatment is as follows: Figure 4 As shown.

[0193] 1. Exhaust Gas Buffer and Final Detection: The exhaust gas after secondary absorption first enters the exhaust gas buffer tank (volume 3... The material is 316L stainless steel. Its function is to stabilize the exhaust gas pressure (to prevent pressure fluctuations during emission from affecting detection accuracy). An online integrated gas analyzer (model: GA-800, integrated) is installed on the outlet pipeline of the buffer tank. , The dust detection module (with the same detection accuracy as the previous dedicated instrument) re-detects the exhaust gas. , , The detection data is transmitted to the central control system in real time.

[0194] 2. Emissions and Recirculation Control: The central control system compares the final test data with the compliance thresholds set by S5.

[0195] like , , If the exhaust gas meets the standards, the system will automatically open the exhaust valve (model: V-600, pneumatic ball valve, 300mm diameter) at the outlet of the buffer tank, and the exhaust gas will be discharged into the atmosphere through a steel chimney (20m high, more than 5m above surrounding buildings);

[0196] If the standard is not met, immediately close the exhaust valve and simultaneously open the return valve (model: V-610, same specification as the exhaust valve) to send the exhaust gas back to the pretreatment process through the return pipe (300mm diameter, connected to the inlet pipe of the S2 heat exchanger) for filtration, cooling and absorption again until the test meets the standard.

[0197] The switching logic between emission and recirculation is dynamically adjusted by the central control system based on the adjustment results of S6. For example, if S6 has increased the frequency of the secondary circulation pump, the system can extend the residence time of the buffer tank (default 30 seconds) and detect emission after the parameters stabilize.

[0198] S8: By-product collection and treatment

[0199] The exhaust gas produced during the process of absorption Dust, concentrated hydrochloric acid, and Si-containing waste liquid must be properly collected and treated to achieve resource recycling and environmental compliance.

[0200] 1. Dust recovery: generated by backflushing the S2 bag filter. Dust (purity ≥99%) is conveyed to a dust storage tank (volume 5) via a closed screw conveyor (material 316L stainless steel, conveying capacity 0-50kg / h). (Conical bottom design). An online level gauge (model: LT-500, radar type, measuring range 0-3m, accuracy ±0.02m) is installed on the top of the storage tank. The level data is transmitted to the central control system: when the level is ≥80% (i.e., 4... When the dust is discharged, the system triggers a tank replacement reminder (audio-visual alarm + interface prompt). The dust is periodically transported out by tanker truck and reused as raw material for the production of fumed silica (it needs to be crushed to a particle size ≤1μm) or sold as a rubber reinforcing agent.

[0201] 2. Concentrated hydrochloric acid recovery: Concentrated hydrochloric acid discharged from the S3 primary absorption circulation tank ( (Used via a corrosion-resistant centrifugal pump (model: P-300, fluoroplastic material, flow rate 0-20)) ) is sent to the hydrochloric acid storage tank (volume 10) (Lined with polytetrafluoroethylene). The storage tank is equipped with an online level gauge (same as LT-500) and an online concentration detector (same as CM-200), and the data is transmitted to the central control system: when the liquid level ≥ 90% (i.e., 9... When the concentration is insufficient, a transfer reminder is triggered, and the solution is transported by tanker truck to a chemical plant for sale as industrial-grade hydrochloric acid (concentration ≥ 31%). If the concentration is insufficient, the solution is returned to the primary absorption circulation tank and mixed with newly replenished 31% industrial hydrochloric acid.

[0202] 3. Treatment of Si-containing waste liquid: Si-containing waste liquid discharged from the S4 secondary absorption circulation tank ( First, it enters the waste liquid sedimentation tank (volume 15). (Divided into 3 compartments connected in series), a 5% sodium hydroxide solution is added via a metering pump (model: M-400, flow rate 0-50L / h) to adjust the pH of the waste liquid to 7-8 (reaction: Excess NaOH makes sodium silicate stable.

[0203] After settling and standing for 2 hours, the mixture is passed through a plate and frame filter press (filtration area 20 mm). Filtration at a working pressure of 0.6 MPa yielded a SiO2 filter cake with a moisture content ≤30% (the main components of which were sodium silicate and unreacted silicate). The filter cake is dried in a hot air dryer (inlet temperature 120℃) and then sold as a filler for construction.

[0204] The filtered filtrate (mainly containing The sample enters an ion exchange resin column (filled with D201 strong basic anion exchange resin, with a processing capacity of 10 m³ / h) to remove residual substances. (control The treated water is recycled through pipelines to the secondary absorption and circulation tank of S4 as supplemental clean water, or discharged after pH adjustment to 6-9 (in compliance with integrated wastewater discharge standards).

[0205] Through the above steps, the tail gas of gas-phase silica is collected, pretreated, and absorbed in two stages to achieve compliant emissions, forming a closed loop. Combined with the S6 fusion algorithm for real-time adjustment, it not only ensures that pollutant emissions meet standards but also realizes the resource utilization of by-products, resulting in significant environmental and economic benefits.

[0206] In some embodiments, to further improve the regulation accuracy of multivariable nonlinear systems, a surface fitting algorithm (SF) is added to the existing fuzzy PID control algorithm (F-PID), adaptive neural fuzzy inference system (ANFIS), and particle swarm optimization algorithm (PSO). The collaborative logic of the four algorithms is as follows:

[0207] 1. Surface fitting algorithm: Based on historical monitoring data of S1-S4, fit the nonlinear surface relationship of "input parameter - ideal adjustment amount", output the fitting correction value, and make up for the prediction bias of ANFIS in small sample scenarios.

[0208] 2. PSO: Not only optimizes the parameters of ANFIS and F-PID, but also optimizes the polynomial coefficients of the surface fitting algorithm to improve the fitting accuracy;

[0209] 3. ANFIS and surface fitting algorithm: The outputs of the two are fused into a "comprehensive feedforward quantity", which is used as the input of F-PID to reduce the prediction risk of a single model;

[0210] 4. F-PID: Combining the comprehensive feedforward quantity and real-time error, the final adjustment command is output. The adjustment error is simultaneously fed back to the PSO and surface fitting algorithm to dynamically update the model.

[0211] Algorithm 4: Surface Fitting Algorithm (SF) – Nonlinear Relation Fitter

[0212] Core function: To fit key exhaust gas parameters (such as flow rate, etc.) using multivariate polynomial fitting. The nonlinear surface relationship between concentration and ideal adjustment (such as pump frequency) provides supplementary correction for ANFIS predictions, especially when historical data is sparse or operating conditions change abruptly, thus improving the reliability of the feedforward signal.

[0213] Model building: Input, output, and surface equation design

[0214] The surface fitting algorithm, based on a "multiple inputs - single output" nonlinear mapping, focuses on key parameters that affect the adjustment effect.

[0215] Input variables (4, from the core parameters of S1-S4): Exhaust gas flow rate (i.e., in S1) ,unit: ); After primary absorption Concentration (i.e., in S3) ,unit: ); After secondary absorption Concentration (i.e., in S4) ,unit: ); : Exhaust gas temperature after cooling (i.e., in S2) (Unit: ℃)

[0216] Output variables:

[0217] : Fitted correction value of ideal adjustment (taking "secondary circulation pump frequency correction value" as an example, unit: Hz).

[0218] The surface equation is in the form of a multivariate quadratic polynomial (balancing fitting accuracy and computational complexity), and its expression is as follows: ;

[0219] : Coefficient of the constant term; : Coefficients of the first-order term (p=1,2,3,4, corresponding to 4 input variables); : Quadratic coefficient ( To avoid overlapping terms, there are a total of 4 + 6 = 10 quadratic terms, such as (etc.); all coefficients The parameters to be optimized are solved using the PSO algorithm.

[0220] Model Training: Coefficient Optimization Based on Least Squares and PSO

[0221] The training objective is to solve for the coefficients using historical data. , so that the fitted value Approaching the ideal adjustment amount (Using the same training samples as ANFIS), the steps are as follows:

[0222] 1. Training Sample Preparation: Select 1000 sets of historical data, identical to those used in ANFIS, with each set containing input data. and output (t=1...1000).

[0223] 2. Preliminary Least Squares Fitting: Substitute the samples into the formula to construct the matrix equation. ,in:

[0224] (1000×1 vector);

[0225] A 1000×15 matrix (each row corresponds to a set of sample input items: );

[0226] (15×1 coefficient vector).

[0227] The initial values ​​of the coefficients are obtained using the least squares method: .

[0228] 3. PSO optimization coefficient: [The remaining text appears to be incomplete and requires further context.] The initial particle positions for PSO are used as the optimization coefficients with the goal of minimizing the fitting error.

[0229] Particle Dimension (Corresponding to 15 coefficients);

[0230] Fitness function: (Mean squared error);

[0231] Optimized output of optimal coefficients Substituting these values ​​into the formula yields the final surface equation.

[0232] Model Application: Real-time Fitting Correction Value Calculation

[0233] The system collects input variables in real time (sampling interval of 1 second). Substitute the optimized surface equation: ;

[0234] Output As a supplement to the ANFIS predictions, it participates in the fusion of feedforward values.

[0235] The fusion and interaction logic and formula of the four algorithms

[0236] The four algorithms form a closed loop through "parameter optimization - dual-model prediction fusion - execution adjustment - error feedback", with the core interaction as follows:

[0237] 1. Interaction between PSO and three other algorithms: Global parameter optimization

[0238] The optimization dimensions of PSO are expanded to (New) ), particle position Includes ANFIS parameters, F-PID parameters, and surface fitting coefficients. The fitness function is updated as follows: ; : Mean absolute error of surface fitting; Weighting coefficients (prioritizing ANFIS prediction accuracy, followed by surface fitting).

[0239] PSO iteratively updates all parameters and outputs the globally optimal parameters. These are respectively assigned to ANFIS, F-PID, and surface fitting algorithm.

[0240] 2. Interaction between ANFIS and surface fitting algorithms: Feedforward fusion

[0241] ANFIS Predictions Correction value for surface fitting Integrate into a "comprehensive feedforward quantity" The formula is: ;

[0242] in The weights are dynamic and determined by the real-time errors of the two models: ;

[0243] The prediction error of ANFIS at the previous time step ( (This refers to the actual adjustment amount at the previous moment).

[0244] : The fitting error of the surface at the previous moment.

[0245] Logic: If ANFIS predictions are more accurate ( ),but Increasing the value (e.g., to 0.8) assigns a higher weight to ANFIS; conversely, decreasing the value increases the weight of surface fitting, achieving a "better fitter gets more weight" approach.

[0246] 3. Integration of feedforward and F-PID: Calculation of control commands

[0247] The input to the F-PID is updated to the combined feedforward. The final adjustment command formula is revised as follows:

[0248] ;

[0249] in , , , , The definition remains the same as before, except that the feedforward term is changed from a single ANFIS output to a fused value.

[0250] 4. Feedback loop: Adjustment error drives model update

[0251] F-PID adjustment error ( (The adjusted detection value) is also fed back to:

[0252] PSO: Updates the fitness function every 10 minutes. The trigger parameters were re-optimized;

[0253] Surface fitting algorithm: using newly acquired data every hour The training set is updated by fine-tuning the coefficients using the least squares method. It adapts to drifting under operating conditions.

[0254] Fusion example: after second-order absorption Concentration exceeding the standard adjustment

[0255] With "secondary absorption" concentration (Exceeded target, target r=10) Taking ")" as an example, the collaborative process of the four algorithms is demonstrated:

[0256] 1. Parameter initialization: PSO has been optimized to obtain:

[0257] Surface fitting coefficients (example): , ( (a first term) ( (quadratic terms);

[0258] The ANFIS and F-PID parameters are the same as before.

[0259] 2. Real-time data input: , , , .

[0260] 3. ANFIS and Surface Fitting Prediction:

[0261] ANFIS output ;

[0262] Surface fitting calculation: .

[0263] 4. Feedforward quantity fusion:

[0264] Error at the previous time step: ;

[0265] Dynamic weights: ;

[0266] Overall feedforward quantity: .

[0267] 5. F-PID control:

[0268] , After correction , , ;

[0269] Adjustment instructions:

[0270] .

[0271] 6. Feedback optimization: Detect after 10 minutes. (Meets standards), adjustment error PSO updates all parameters based on the new error, and the surface fitting algorithm fine-tunes the coefficients using the new samples.

[0272] The core contribution of surface fitting algorithms

[0273] 1. Enhanced nonlinear fitting capabilities: Compared to ANFIS, which relies on fuzzy rules, surface fitting directly models multivariate interaction relationships (such as exhaust gas flow rate and...) using polynomials. (The synergistic effect of concentration) reduces the fitting error by 20% to 30% in scenarios with strong parameter coupling;

[0274] 2. Improve robustness to small samples: When a certain parameter (such as...) When historical data on concentration is insufficient, surface fitting can be achieved through other relevant parameters (such as concentration). The quadratic term of the concentration indirectly captures the pattern, avoiding the predicted mutations of ANFIS;

[0275] 3. Dynamic weight fusion mechanism: By adaptively allocating weights based on real-time error, the problem of insufficient adaptability of a single model when switching operating conditions is solved, and the standard deviation of the feedforward quantity is reduced from ±1.5Hz to ±0.8Hz, providing a more stable input for F-PID.

[0276] After the four algorithms are integrated, the system can handle exhaust gas flow fluctuations of ±30%. Even under extreme operating conditions with concentration fluctuations of ±50%, it can still absorb secondary adsorption. Concentration controlled at 8-12 (Within the acceptable range), the adjustment response time is shortened to 0.8 seconds, a 25% improvement compared to the fusion of the three algorithms.

[0277] A gas-phase silica tail gas absorption system with online monitoring and regulation function, applying any one of the above-mentioned gas-phase silica tail gas absorption methods with online monitoring and regulation function, the gas-phase silica tail gas absorption system with online monitoring and regulation function includes: a tail gas collection module, used to collect the tail gas generated by the reactor, detect the flow rate, temperature, and pressure data of the tail gas, and obtain initial parameters; a tail gas pretreatment module, used to filter silica dust in the tail gas and detect the dust concentration, and then, in conjunction with the tail gas temperature in S1, adjust the cooling water flow rate of the heat exchanger to... The filtered exhaust gas is cooled, and the temperature of the cooled exhaust gas is monitored. A primary absorption module introduces the cooled exhaust gas into the primary absorption tower, using dilute hydrochloric acid solution as the absorbent to absorb hydrogen chloride in the exhaust gas. The concentration of circulating dilute hydrochloric acid, pH value, and the concentration of residual hydrogen chloride after primary absorption are monitored. A secondary absorption module introduces the exhaust gas after primary absorption into the secondary absorption tower, using clean water as the absorbent to absorb residual hydrogen chloride and react silicon tetrachloride in the exhaust gas with water. The pH value, silicon content, and circulating tank level of the clean water are monitored. The concentration of chloride in the exhaust gas after secondary absorption is also monitored. The module monitors the residual concentrations of hydrogen chloride and silicon tetrachloride. A real-time monitoring module monitors the parameters based on initial parameters, dust concentration detected by S2, exhaust gas temperature after cooling, dilute hydrochloric acid concentration and pH value detected by S3, residual hydrogen chloride concentration after primary absorption, pH value of circulating water, silicon content, and circulating tank level detected by S4, and the residual concentrations of hydrogen chloride and silicon tetrachloride in the exhaust gas after secondary absorption. It sets safety and exceedance thresholds for these parameters and triggers an audible and visual alarm when the parameters reach the exceedance threshold. A control module, based on the monitoring data from S5, uses a fusion of particle swarm optimization algorithm, adaptive neurofuzzy inference system, and fuzzy PID control algorithm to achieve automatic adjustment. The particle swarm optimization algorithm optimizes the membership function parameters of the adaptive neurofuzzy inference system and the fuzzy PID control algorithm. The adaptive neurofuzzy inference system predicts the ideal adjustment amount based on real-time data from S1-S4, which serves as the feedforward correction for the fuzzy PID control algorithm. The fuzzy PID control algorithm combines the predicted value from the adaptive neurofuzzy inference system with the real-time adjustment error to output adjustment commands to the corresponding equipment for controlling S2. The system controls the cooling water flow rate, the dilute hydrochloric acid spray rate in S3, and the clean water spray rate in S4. Simultaneously, the adjustment error of the fuzzy PID control algorithm is fed back to the particle swarm optimization algorithm to dynamically update the optimization parameters. The exhaust gas emission module introduces the exhaust gas after secondary absorption into the exhaust gas buffer tank, detects the concentrations of hydrogen chloride, silicon tetrachloride, and dust in the exhaust gas, and if the standards are met, it is discharged through the chimney; otherwise, it is returned to S2 for reprocessing. The by-product treatment module collects the silica dust generated during filtration, the concentrated hydrochloric acid discharged from S3, and the silicon-containing waste liquid discharged from S4, and recycles or treats them respectively.

Claims

1. A gas-phase silica tail gas absorption method with online monitoring and adjustment function, characterized in that, Includes the following steps: S1: Collect the exhaust gas generated by the reactor, detect the flow rate, temperature, and pressure data of the exhaust gas, and obtain the initial parameters; S2: Filter the silica dust in the exhaust gas and detect the dust concentration. Then, in combination with the exhaust gas temperature in S1, adjust the cooling water flow of the heat exchanger to cool down the filtered exhaust gas and detect the temperature of the cooled exhaust gas. S3: The cooled exhaust gas is introduced into the first-stage absorption tower, and dilute hydrochloric acid solution is used as the absorbent to absorb hydrogen chloride in the exhaust gas. The concentration of circulating dilute hydrochloric acid, pH value and residual hydrogen chloride concentration after first-stage absorption are detected. S4: The tail gas after primary absorption is introduced into the secondary absorption tower. Water is used as the absorbent to absorb the residual hydrogen chloride and to react the silicon tetrachloride in the tail gas with water. The pH value, silicon content and liquid level of the circulating water and the liquid level of the circulating tank are detected. The residual concentrations of hydrogen chloride and silicon tetrachloride in the tail gas after secondary absorption are detected. S5: Based on the initial parameters, the dust concentration and exhaust gas temperature detected by S2, the dilute hydrochloric acid concentration, pH value and residual hydrogen chloride concentration after primary absorption detected by S3, the pH value, silicon content and circulating tank level of the circulating water detected by S4, and the residual concentrations of hydrogen chloride and silicon tetrachloride in the exhaust gas after secondary absorption, real-time monitoring is performed and safety thresholds and exceedance thresholds for the parameters are set. When the parameters reach the exceedance threshold, an audible and visual alarm is triggered. S6: Based on the monitoring data of S5, automatic adjustment is achieved by integrating particle swarm optimization algorithm, adaptive neuro-fuzzy inference system, and fuzzy PID control algorithm: Particle swarm optimization algorithm optimizes the membership function parameters of adaptive neuro-fuzzy inference system and fuzzy PID control algorithm; Adaptive neuro-fuzzy inference system predicts ideal adjustment amount based on real-time data of S1-S4, as feedforward correction amount of fuzzy PID control algorithm; Fuzzy PID control algorithm combines the predicted value of adaptive neuro-fuzzy inference system with real-time adjustment error, and outputs adjustment command to corresponding equipment to control the cooling water flow rate of S2, the dilute hydrochloric acid spray rate of S3, and the clean water spray rate of S4; At the same time, the adjustment error of fuzzy PID control algorithm is fed back to particle swarm optimization algorithm to dynamically update optimization parameters; S7: The exhaust gas after secondary absorption is introduced into the exhaust gas buffer tank to detect the concentration of hydrogen chloride, silicon tetrachloride and dust in the exhaust gas. If it meets the standard, it is discharged through the chimney. If it does not meet the standard, it is returned to S2 for reprocessing. S8: Collect silica dust, concentrated hydrochloric acid and silicon-containing waste liquid generated during the exhaust gas absorption process, and recycle or treat them respectively.

2. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S6, the particle dimension of the particle swarm optimization algorithm is the sum of the parameter dimensions of the adaptive neurofuzzy inference system and the parameter dimensions of the fuzzy PID control algorithm. The parameters of the adaptive neurofuzzy inference system include the membership function centers and widths of the 6 input variables and the output weights of 729 fuzzy rules. The parameters of the fuzzy PID control algorithm include the membership function centers and widths of the error and the error change rate.

3. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S6, the input variables of the adaptive neural fuzzy inference system include exhaust gas flow rate, residual hydrogen chloride concentration after primary absorption, residual hydrogen chloride concentration after secondary absorption, pH value of secondary absorption circulating water, silicon content of secondary absorption circulating water, and exhaust gas temperature after cooling. The output variable is the ideal adjustment amount.

4. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S6, the inputs to the fuzzy PID control algorithm include the adjustment error and its rate of change, as well as the ideal adjustment amount output by the adaptive neural fuzzy inference system. The PID parameter correction amount is output through the fuzzy rule table, and the corrected PID parameters are combined with the feedforward correction amount to calculate the final adjustment command.

5. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S6, the interaction process of the three algorithms is as follows: the optimization parameters output by the particle swarm optimization algorithm are used as the initial parameters of the adaptive neurofuzzy inference system and the fuzzy PID control algorithm, respectively; the predicted value of the adaptive neurofuzzy inference system is used as the feedforward input of the fuzzy PID control algorithm; the adjustment error of the fuzzy PID control algorithm is fed back to the fitness function of the particle swarm optimization algorithm, and the particle swarm optimization algorithm is triggered to iterate and optimize the parameters again every 10 minutes to form a closed loop.

6. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 5, characterized in that, The closed-loop termination condition is: after adjustment by the fuzzy PID control algorithm, all parameters set in S5 return to the safe threshold and do not exceed the standard for 30 consecutive seconds.

7. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S7, the criterion for determining whether the exhaust gas meets the emission standards is: hydrogen chloride concentration ≤ 10. Silicon tetrachloride concentration ≤5 Dust concentration ≤5 The exhaust valve is opened when all three conditions are met; otherwise, the return valve is opened to send the exhaust gas back to S2.

8. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S2, the specific operation for cooling the filtered exhaust gas is as follows: the filtered exhaust gas is introduced into a shell-and-tube heat exchanger, and the cooling water flow rate is adjusted based on the exhaust gas temperature data in S1 to reduce the exhaust gas temperature to 40-60℃. The cooling water flow rate adjustment logic of the shell-and-tube heat exchanger is as follows: if the exhaust gas temperature detected in S1 is >80℃, the opening of the cooling water flow rate regulating valve is linearly increased according to the difference between the exhaust gas temperature and 80℃; if it is ≤80℃, the valve opening is maintained at 50%.

9. The gas-phase silica tail gas absorption method with online monitoring and adjustment function according to claim 1, characterized in that, In step S8, the treatment process of silicon-containing waste liquid is as follows: after being introduced into the waste liquid sedimentation tank, sodium hydroxide solution is added to adjust the pH to 7-8. After standing for 2 hours, it is filtered through a plate and frame filter press. The filter cake is dried and used as industrial packing. The filtrate is reused or discharged after chloride ions are removed by an ion exchange resin column.

10. A gas-phase silica tail gas absorption system with online monitoring and adjustment function, characterized in that, The gas-phase silica tail gas absorption method with online monitoring and regulation function according to any one of claims 1 to 9, wherein the gas-phase silica tail gas absorption system with online monitoring and regulation function comprises: The exhaust gas collection module is used to collect the exhaust gas generated by the reactor, detect the flow rate, temperature, and pressure data of the exhaust gas, and obtain initial parameters. The exhaust gas pretreatment module is used to filter silica dust in the exhaust gas and detect the dust concentration. Combined with the exhaust gas temperature adjustment of S1, the cooling water flow of the heat exchanger is adjusted to cool down the filtered exhaust gas and the temperature of the cooled exhaust gas is detected. The primary absorption module is used to introduce the cooled exhaust gas into the primary absorption tower, using dilute hydrochloric acid solution as the absorbent to absorb hydrogen chloride in the exhaust gas, and to detect the concentration of circulating dilute hydrochloric acid, pH value and residual hydrogen chloride concentration after primary absorption. The secondary absorption module is used to introduce the tail gas after the primary absorption into the secondary absorption tower. It uses clean water as the absorbent to absorb the residual hydrogen chloride and reacts the silicon tetrachloride in the tail gas with water. It detects the pH value, silicon content and liquid level of the circulating clean water, and detects the residual concentration of hydrogen chloride and silicon tetrachloride in the tail gas after the secondary absorption. The real-time monitoring module is used to monitor and set safety thresholds and exceedance thresholds for parameters based on initial parameters, dust concentration and exhaust gas temperature after cooling detected by S2, dilute hydrochloric acid concentration, pH value and residual hydrogen chloride concentration after primary absorption detected by S3, pH value, silicon content and circulating tank level of circulating water detected by S4, and residual concentrations of hydrogen chloride and silicon tetrachloride in exhaust gas after secondary absorption. When the parameters reach the exceedance threshold, an audible and visual alarm is triggered. The regulation and control module, based on the monitoring data from S5, employs a fusion of particle swarm optimization (PSO) algorithm, adaptive neural fuzzy inference system (ANSI), and fuzzy PID control algorithm to achieve automatic regulation. The PSO algorithm optimizes the membership function parameters of the ASI and fuzzy PID control algorithms. The ASI predicts the ideal regulation amount based on real-time data from S1-S4, serving as the feedforward correction for the fuzzy PID control algorithm. The fuzzy PID control algorithm combines the predicted value from the ASI with the real-time regulation error to output regulation commands to the corresponding devices, controlling the cooling water flow rate in S2, the dilute hydrochloric acid spray rate in S3, and the clean water spray rate in S4. Simultaneously, the regulation error of the fuzzy PID control algorithm is fed back to the PSO algorithm to dynamically update the optimization parameters. The exhaust gas emission module is used to introduce the exhaust gas after secondary absorption into the exhaust gas buffer tank, and detect the concentration of hydrogen chloride, silicon tetrachloride and dust in the exhaust gas. If it meets the standard, it is discharged through the chimney; if it does not meet the standard, it is returned to S2 for reprocessing. The by-product processing module is used to collect the silica dust generated by filtration, the concentrated hydrochloric acid discharged from S3, and the silicon-containing waste liquid discharged from S4, and to recycle or treat them respectively.

Images