Intelligent control method of multi-frequency modified adsorption system

Abstract

The invention discloses an intelligent control method for a multi-frequency modified adsorption system, and belongs to the field of automatic control technology. The multi-frequency modified adsorption system comprises a plurality of adsorption towers and valves. A plurality of sensors are first installed on each adsorption tower to collect data and valve switch status in real time; historical data are collected, the relationship between material type and multi-frequency modification parameters is analyzed, and a knowledge base is constructed; real-time data is intelligently identified and classified to determine the material type and adsorption stage, and the frequency, intensity and action time of the physical field are adjusted in combination with the knowledge base, and then the adsorption tower and valve operation are controlled according to the real-time status data to dynamically adjust the adsorption system; finally, the adsorption tower outlet data is collected to compare the actual adsorption effect with the expected adsorption effect, and when the deviation exceeds the set threshold, an adaptation rule is re-searched from the knowledge base, and if the adaptation rule is not found, a warning is issued in time, and the parameters are manually adjusted and recorded and stored, so as to realize the intelligent control of the multi-frequency modified adsorption system.

Description

Technical Field

[0001] The invention relates to the technical field of automatic control, in particular to an intelligent control method for a multi-frequency modified adsorption system. Background Art

[0002] In many industrial fields such as chemical industry, environmental protection, and pharmaceuticals, adsorption technology is a key means to achieve material separation and purification. Its principle is to use the difference in adsorption capacity of adsorbents for different substances to separate the target component from other coexisting components. Traditional adsorption systems mainly rely on changing the physical and chemical properties of the adsorbent itself to improve the adsorption effect, but as the industry's requirements for separation efficiency and selectivity continue to increase, its limitations are becoming increasingly prominent.

[0003] The multi-frequency modified adsorption system is a new type of adsorption technology. It regulates the interaction between the adsorbent and the adsorbate by applying a physical field (such as an electromagnetic field, ultrasonic field, etc.) with a specific frequency combination, intensity and action time during the adsorption process. The physical field can change the electron cloud distribution, molecular vibration state and active sites on the adsorbent surface, enhance the adsorption capacity and selectivity, and promote the diffusion of adsorbate molecules in the adsorbent pores, thereby accelerating the adsorption rate. However, due to the complex nature of the material and the combined influence of multiple factors on the adsorption process, a large amount of experiments and experience are often required to adjust the parameters, which is time-consuming and cost-intensive, and it is difficult to ensure the stability and consistency of the adsorption effect; in addition, the existing adsorption system control methods are mostly based on manual experience or simple feedback control, lacking intelligent and automated decision-making mechanisms. In the face of complex and changeable working conditions, it is impossible to adjust the physical field parameters and operating conditions in a timely manner according to real-time data. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent control method for a multi-frequency modified adsorption system to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: an intelligent control method of a multi-frequency modified adsorption system, wherein the multi-frequency modified adsorption system comprises a plurality of adsorption towers and valves, and the control method comprises:

[0006] Step S100: installing a variety of sensors on each adsorption tower to collect various data and the switch status of each valve in real time, and pre-processing the collected data;

[0007] Step S200: collecting historical data of the adsorption system, analyzing the corresponding relationship between different material types and multi-frequency modification parameters, and constructing a multi-frequency modification knowledge base;

[0008] Step S300: Intelligently identify and classify the data collected in real time to identify the material type and adsorption stage; based on the intelligent identification results, combined with the multi-frequency modification knowledge base, adjust the frequency, intensity and action time of the physical field to generate a multi-frequency modification strategy;

[0009] Step S400: According to the state data of the adsorption system collected in real time and the multi-frequency modification strategy, the working state of the adsorption tower and the valve is controlled, and the multi-frequency modification adsorption system is dynamically adjusted to perform corresponding operations;

[0010] Step S500: Collect data at the outlet of the adsorption tower, analyze the actual adsorption effect, and compare it with the expected effect, set a deviation threshold, if the deviation exceeds the threshold, indicating that the actual effect deviates from the expected, then retrieve the adaptation rules from the multi-frequency modification knowledge base again; if the adaptation rules are not found, timely warning, manual adjustment of system parameters and record storage.

[0011] Furthermore, in step S100, temperature sensors, pressure sensors, flow sensors and concentration sensors are installed at different heights and positions of each adsorption tower, respectively for real-time acquisition and monitoring of the temperature T, pressure C, material inlet and outlet flow L and concentration N of the target component in the material in the adsorption tower; position sensors are installed at each valve to obtain the switch status of the valve; spectral sensors are installed at different positions inside the adsorption tower and at the inlet and outlet of the material to obtain spectral data G; the collected data is filtered by a sliding window filtering algorithm to remove noise data, and the different types of data are unified into the [0,1] interval using the Min-Max normalization method;

[0012] Furthermore, the step S200 includes:

[0013] Step S201: Collect historical data from the adsorption system over a period of time in the past, including material type M, frequency combination F=[f1, f2, ..., fi] of the physical field, intensity I, action time t, spectral data, and corresponding adsorption amount Q and adsorption selectivity Z, wherein fi represents the i-th frequency component; the material type M represents the type of material entering the adsorption tower; the frequency combination F of the physical field represents the frequency combination of the physical field applied during the adsorption process; the intensity I is the intensity of the physical field, which represents the energy of the physical field acting on the material and the adsorbent; the action time t represents the duration of the continuous action of the physical field; the spectral data G includes the wavelength and intensity of the spectral curve, which is used to identify the material type; the adsorption amount Q is the amount of the target component absorbed by the adsorbent; the adsorption selectivity Z is the ratio of the adsorbent to the target component compared to other coexisting components.

[0014] According to the relationship between the adsorption capacity and the action time, each record in the historical data is marked with the corresponding adsorption stage S, including the initial adsorption stage, the intermediate adsorption stage, and the adsorption equilibrium stage; calculate the adsorption rate: v(t)=[Q(t+Δt)-Q(t)] / Δt; where v(t) represents the adsorption rate at time t, Q(t) represents the adsorption capacity at time t, Δt is the set sampling interval, and Q(t+Δt) represents the adsorption capacity at time t+Δt; set the threshold V1 as the adsorption rate threshold in the initial stage, and V2 as the adsorption rate threshold in the adsorption equilibrium stage; when v(t)>=V1, it is set that the adsorption process belongs to the initial adsorption stage; when V2<=v(t)<V1, it is set that the adsorption process enters the intermediate adsorption stage; when v(t)<=V2, it is set that the adsorption process reaches the adsorption equilibrium stage.

[0015] Step S202: First, set the support threshold smin to measure the frequency of occurrence of candidate item sets in the dataset; generate candidate set Ck, where k represents the number of elements in the item set; in the initial stage, generate candidate sets containing a single element: {material type}, {adsorption stage}, {frequency combination}, {intensity}, {action time}, {spectral data}, {temperature}, {pressure}, {flow rate}, {concentration}; according to the historical data of the adsorption system, calculate the frequency s of each candidate item set in the dataset: s = number of records containing this item set / total number of records; filter out the item sets with frequency s greater than smin to obtain the frequent item set Lk; as the value of k increases, generate the final candidate set {material type, adsorption stage, frequency combination, intensity, action time, spectral data, temperature, pressure, flow rate, concentration} by combining the existing frequent item sets; generate association rules from the frequent item sets, with the antecedent including {material type, adsorption stage, spectral data, temperature, pressure, flow rate, concentration} and the consequent including {frequency combination, intensity, action time, adsorption capacity, adsorption selectivity}; set the confidence threshold cmin, and for each generated association rule, calculate its confidence c = number of records containing the antecedent and the consequent / number of records containing the antecedent; filter out the association rules with confidence greater than the threshold cmin, and store the filtered association rules in the multi-frequency modification knowledge base in a structured manner.

[0016] Further, the step S300 includes:

[0017] Step S301: The data collected in step S100 is used as input, recorded as vector X=[x1,x2,...,xn], where xn represents the nth feature, including temperature, pressure, flow, concentration, and spectral data; a deep neural network DNN model is used for identification and classification, and the model consists of an input layer, multiple hidden layers, and an output layer, and the hidden layer uses a ReLU activation function; the historical data collected in step S200 is used as a training set, where the material type and adsorption stage are used as target outputs; the learning rate and the number of training rounds are set to train the DNN, and the weights and biases of the network are adjusted by the back propagation algorithm, and the model parameters are optimized by the cross entropy loss function; the real-time data vector X is input into the trained DNN model, and the predicted category ML of the material type and the adsorption stage S are output;

[0018] Step S302: According to the identified material type and adsorption stage, search for matching association rules in the multi-frequency modification knowledge base constructed in step S200; set the association rule set in the knowledge base to R={r1, r2, ..., rj}, where rj represents the jth rule in the set R, and the rule rj is represented by a twelve-tuple (Mj, Sj, Fj, Ij, tj, Gj, Tj, Cj, Lj, Nj, Qj, Zj), including the corresponding material type, adsorption stage, physical field frequency combination, physical field intensity, action time, spectral data, temperature, pressure, flow rate, concentration, adsorption amount, and adsorption selectivity; traverse the rule set R to find the rule that satisfies Mj=ML and Sj=S. If there are multiple matching rules, select the rule with the highest confidence as the final rule; extract the physical field frequency combination F, intensity I, and action time t from the final rule to generate a multi-frequency modification strategy.

[0019] In the above technical solution, a deep neural network DNN model is used for intelligent identification and classification. It can utilize the collected multi-dimensional data and extract key features through the neural network structure, so as to realize accurate identification of the system state. Based on the intelligent identification results and the multi-frequency modification knowledge base, a multi-frequency modification strategy is generated to adjust parameters such as the frequency, intensity and action time of the physical field to adapt to the needs of different material types and adsorption stages, thereby improving the flexibility and adaptability of the system.

[0020] Furthermore, the step S400 includes:

[0021] Step S401: The frequency combination F in the multi-frequency modification strategy is used to generate a multi-frequency electromagnetic field through a frequency synthesizer and sent to the physical field generator of the adsorption tower through a PLC; at the same time, according to the intensity I in the strategy, the power output of the physical field transmitter is adjusted; a timer is set to control the duration of the physical field action; according to the real-time flow sensor, the opening of the first type of valve is adjusted to control the real-time flow of the material; when switching between adsorption stages, the first type of valve is opened and closed in a preset order to maintain the continuous flow of the material; according to the concentration data of the target component in the material fed back by the concentration sensor, the opening of the second type of valve is dynamically adjusted to control the injection amount of the adsorption agent; the first type of valve is a related valve for controlling the flow of the fluid, and the second type of valve is a related valve for adjusting the modification process.

[0022] Step S402: During the adsorption process, the adsorption amount Q of each adsorption tower is monitored in real time, and the adsorption stage of the adsorption tower is dynamically switched according to the change of the adsorption amount; in the initial adsorption stage, the maximum flow rate of the material is set to Qmax, and in the initial adsorption stage, the material flow rate Q=Qmax; the physical field intensity and action time are set to the corresponding values ​​in the multi-frequency modification strategy; in the mid-term adsorption stage, the material flow rate is reduced and the physical field frequency is increased; in the adsorption equilibrium stage, the first type of valve is closed, the material input is stopped, and the physical field action is maintained until the set target time t is reached, and the modification process of the adsorbent is completed.

[0023] In the above technical scheme, the working status of the adsorption tower and the valve are dynamically adjusted according to the real-time collected adsorption system status data and the multi-frequency modification strategy, and precise control of the adsorption process is achieved by controlling key variables such as the material flow rate and the parameters of the physical field.

[0024] Furthermore, the step S500 includes:

[0025] Step S501: collect the concentration data of the target component at the outlet of the adsorption tower in real time, calculate the adsorption amount of the current adsorption process = (inlet target component concentration - outlet target component concentration) × material flow × adsorption time; determine the selective adsorption index Z = (target component adsorption amount / coexisting impurity adsorption amount); sort out the three data of outlet concentration, adsorption amount and adsorption selectivity index to form a new data vector Y = [outlet concentration, adsorption amount, adsorption selectivity];

[0026] Step S502: taking the preprocessed data vector Y as input, inputting it into the DNN model in step S300, outputting the predicted material type and adsorption stage, and matching the corresponding adsorption amount and adsorption selectivity according to the multi-frequency modification knowledge base, and taking the model prediction result as the actual adsorption effect; comparing the actual adsorption effect with the expected result in the multi-frequency modification strategy, and calculating the deviation between the two; setting a deviation threshold, when the adsorption amount deviation or adsorption selectivity deviation exceeds the threshold, it is determined that the actual adsorption effect deviates from the expectation, and the strategy is re-planned; returning to step S302, taking the currently identified material type and adsorption stage as conditions, re-retrieve new rules from the knowledge base, and search for new rules in the association rule set R of the knowledge base;

[0027] Step S503: If no matching rule is found in the knowledge base, an early warning will be issued to promptly notify relevant personnel to manually adjust the relevant parameters of the adsorption system, and the relevant data of the abnormal situation will be recorded and stored.

[0028] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0029] The present invention realizes intelligent control of the multi-frequency modified adsorption system by combining real-time data collection, a multi-frequency modification knowledge base and a deep neural network model, can dynamically adjust the working state of the adsorption tower and the valve, and automatically optimize the frequency, intensity and action time of the physical field, thereby improving the accuracy and efficiency of the adsorption process; through the analysis and intelligent recognition of historical data, it can accurately identify the material type and adsorption stage, and generate corresponding multi-frequency modification strategies based on this, thereby improving the adaptive ability of the adsorption process;

[0030] The present invention introduces a multi-frequency modification strategy, which not only improves the adsorption amount and selectivity of the adsorption process, but also timely adjusts the control strategy by comparing the actual adsorption effect with the expected result in real time, thereby effectively reducing the reliance on manual intervention; at the same time, by combining the deep learning model with historical data, it can more accurately match materials and adsorption processes, optimize the operating efficiency of the adsorption tower, and ensure the efficient and stable operation of the adsorption system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0032] Figure 1 It is a method flow chart of the intelligent control method of the multi-frequency modified adsorption system. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] See also Figure 1 The present invention provides a technical solution: an intelligent control method for a multi-frequency modified adsorption system, wherein the multi-frequency modified adsorption system comprises a plurality of adsorption towers and valves, and the control method comprises:

[0035] Step S100: installing a variety of sensors on each adsorption tower to collect various data and the switch status of each valve in real time, and pre-processing the collected data;

[0036] Step S200: collecting historical data of the multi-frequency modification adsorption system, analyzing the corresponding relationship between different material types and multi-frequency modification parameters, and constructing a multi-frequency modification knowledge base;

[0037] Step S300: Intelligently identify and classify the data collected in real time to identify the material type and adsorption stage; based on the intelligent identification results, combined with the multi-frequency modification knowledge base, adjust the frequency, intensity and action time of the physical field to generate a multi-frequency modification strategy;

[0038] Step S400: According to the state data of the adsorption system collected in real time and the multi-frequency modification strategy, the working state of the adsorption tower and the valve is controlled, and the adsorption system is dynamically adjusted to perform corresponding operations;

[0039] Step S500: Collect data at the outlet of the adsorption tower, analyze the actual adsorption effect, and compare it with the expected effect, set a deviation threshold, if the deviation exceeds the threshold, indicating that the actual effect deviates from the expected, then retrieve the adaptation rules from the multi-frequency modification knowledge base again; if the adaptation rules are not found, timely warning, manual adjustment of system parameters and record storage.

[0040] Furthermore, in step S100, temperature sensors, pressure sensors, flow sensors and concentration sensors are installed at different heights and positions of each adsorption tower, respectively for real-time acquisition and monitoring of the temperature T, pressure C, material inlet and outlet flow L and concentration N of the target component in the material in the adsorption tower; position sensors are installed at each valve to obtain the switch status of the valve; spectral sensors are installed at different positions inside the adsorption tower and at the inlet and outlet of the material to obtain spectral data G; the collected data is filtered by a sliding window filtering algorithm to remove noise data, and the different types of data are unified into the [0,1] interval using the Min-Max normalization method;

[0041] Further, the step S200 includes:

[0042] Step S201: Collect historical data within a past period from the adsorption system, including material type M, frequency combination F = [f1, f2,..., fi] of the physical field, intensity I, action time t, spectral data, and corresponding adsorption capacity Q and adsorption selectivity Z, where fi represents the i-th frequency component; the material type M represents the type of material entering the adsorption tower; the frequency combination F of the physical field represents the frequency combination of the physical field applied during the adsorption process; the intensity I is the intensity of the physical field, indicating the energy magnitude of the physical field acting on the material and the adsorbent; the action time t represents the duration of the continuous action of the physical field; the spectral data G includes the wavelength and intensity of the spectral curve, used to identify the material type; the adsorption capacity Q is the quantity of the target component absorbed by the adsorbent; the adsorption selectivity Z is the ratio of the adsorbent to the target component compared to other co-existing components.

[0043] According to the relationship between the adsorption capacity and the action time, mark each record in the historical data with the corresponding adsorption stage S, including the initial adsorption stage, the intermediate adsorption stage, and the adsorption equilibrium stage; calculate the adsorption rate: v(t) = [Q(t + Δt) - Q(t)] / Δt; where v(t) represents the adsorption rate at time t, Q(t) represents the adsorption capacity at time t, Δt is the set sampling interval, and Q(t + Δt) represents the adsorption capacity at time t + Δt; set the threshold V1 as the adsorption rate threshold in the initial stage, and V2 as the adsorption rate threshold in the adsorption equilibrium stage; when v(t) >= V1, set the adsorption process to belong to the initial adsorption stage; when V2 <= v(t) < V1, set the adsorption process to enter the intermediate adsorption stage; when v(t) <= V2, set the adsorption process to reach the adsorption equilibrium stage.

[0044] Step S202: First, set the support threshold smin to measure the frequency of candidate item sets in the data set; generate a candidate set Ck, where k represents the number of elements in the item set; in the initial stage, generate a candidate set containing a single element: {material type}, {adsorption stage}, {frequency combination}, {intensity}, {action time}, {spectral data}, {temperature}, {pressure}, {flow rate}, {concentration}; based on the historical data of the adsorption system, calculate the frequency of each candidate item set in the data set s = the number of records containing this item set / the total number of records; filter out the item sets with a frequency s greater than smin to obtain the frequent item set Lk; as the k value increases, by combining the existing frequent item sets, the frequent item sets Lk are obtained. The item set generates the final candidate set {material type, adsorption stage, frequency combination, intensity, action time, spectral data, temperature, pressure, flow rate, concentration}; association rules are generated from the frequent item set, the antecedent includes {material type, adsorption stage, spectral data, temperature, pressure, flow rate, concentration}, and the consequent includes {frequency combination, intensity, action time, adsorption amount Q, adsorption selectivity Z}; the confidence threshold cmin is set, and for each generated association rule, its confidence c=the number of records containing the antecedent and the consequent / the number of records containing the antecedent is calculated; the association rules with confidence greater than the threshold cmin are screened out, and the screened association rules are stored in a structured manner in the multi-frequency modification knowledge base.

[0045] Furthermore, the step S300 includes:

[0046] Step S301: The data collected in step S100 is used as input, recorded as vector X=[x1,x2,...,xn], where xn represents the nth feature, including temperature, pressure, flow, concentration, and spectral data; a deep neural network DNN model is used for identification and classification, and the model consists of an input layer, multiple hidden layers, and an output layer, and the hidden layer uses a ReLU activation function; the historical data collected in step S200 is used as a training set, where the material type and adsorption stage are used as target outputs; the learning rate and the number of training rounds are set to train the DNN, and the weights and biases of the network are adjusted by the back propagation algorithm, and the model parameters are optimized by the cross entropy loss function; the real-time data vector X is input into the trained DNN model, and the predicted category ML of the material type and the adsorption stage S are output;

[0047] Step S302: According to the identified material type and adsorption stage, search for matching association rules in the multi-frequency modification knowledge base constructed in step S200; set the association rule set in the knowledge base to R={r1, r2, ..., rj}, where rj represents the jth rule in the set R, and the rule rj is represented by a twelve-tuple (Mj, Sj, Fj, Ij, tj, Gj, Tj, Cj, Lj, Nj, Qj, Zj), including the corresponding material type, adsorption stage, physical field frequency combination, physical field intensity, action time, spectral data, temperature, pressure, flow rate, concentration, adsorption amount, and adsorption selectivity; traverse the rule set R to find the rule that satisfies Mj=ML and Sj=S. If there are multiple matching rules, select the rule with the highest confidence as the final rule; extract the physical field frequency combination F, intensity I, and action time t from the final rule to generate a multi-frequency modification strategy.

[0048] In the above technical solution, a deep neural network DNN model is used for intelligent identification and classification. It can utilize the collected multi-dimensional data and extract key features through the neural network structure, so as to realize accurate identification of the system state. Based on the intelligent identification results and the multi-frequency modification knowledge base, a multi-frequency modification strategy is generated to adjust parameters such as the frequency, intensity and action time of the physical field to adapt to the needs of different material types and adsorption stages, thereby improving the flexibility and adaptability of the system.

[0049] Furthermore, the step S400 includes:

[0050] Step S401: The frequency combination F in the multi-frequency modification strategy is used to generate a multi-frequency electromagnetic field through a frequency synthesizer and sent to the physical field generator of the adsorption tower through a PLC; at the same time, according to the intensity I in the strategy, the power output of the physical field transmitter is adjusted; a timer is set to control the duration of the physical field action; according to the real-time flow sensor, the opening of the first type of valve is adjusted to control the real-time flow of the material; when switching between adsorption stages, the first type of valve is opened and closed in a preset order to maintain the continuous flow of the material; according to the concentration data of the target component in the material fed back by the concentration sensor, the opening of the second type of valve is dynamically adjusted to control the injection amount of the adsorption agent; the first type of valve is a related valve for controlling the flow of the fluid, and the second type of valve is a related valve for adjusting the modification process.

[0051] Step S402: During the adsorption process, the adsorption amount Q of each adsorption tower is monitored in real time, and the adsorption stage of the adsorption tower is dynamically switched according to the change of the adsorption amount; in the initial adsorption stage, the maximum flow rate of the material is set to Qmax, and in the initial adsorption stage, the material flow rate Q=Qmax; the physical field intensity and action time are set to the corresponding values ​​in the multi-frequency modification strategy; in the mid-term adsorption stage, the material flow rate is reduced and the physical field frequency is increased; in the adsorption equilibrium stage, the first type of valve is closed, the material input is stopped, and the physical field action is maintained until the set target time t is reached, and the modification process of the adsorbent is completed.

[0052] In the above technical scheme, the working status of the adsorption tower and the valve are dynamically adjusted according to the real-time collected adsorption system status data and the multi-frequency modification strategy, and precise control of the adsorption process is achieved by controlling key variables such as the material flow rate and the parameters of the physical field.

[0053] Furthermore, the step S500 includes:

[0054] Step S501: collect the concentration data of the target component at the outlet of the adsorption tower in real time, calculate the adsorption amount of the current adsorption process = (inlet target component concentration - outlet target component concentration) × material flow × adsorption time; determine the selective adsorption index Z = (target component adsorption amount / coexisting impurity adsorption amount); sort out the three data of outlet concentration, adsorption amount and adsorption selectivity index to form a new data vector Y = [outlet concentration, adsorption amount, adsorption selectivity];

[0055] Step S502: taking the preprocessed data vector Y as input, inputting it into the DNN model in step S300, outputting the predicted material type and adsorption stage, and matching the corresponding adsorption amount and adsorption selectivity according to the multi-frequency modification knowledge base, and taking the model prediction result as the actual adsorption effect; comparing the actual adsorption effect with the expected result in the multi-frequency modification strategy, and calculating the deviation between the two; setting a deviation threshold, when the adsorption amount deviation or adsorption selectivity deviation exceeds the threshold, it is determined that the actual adsorption effect deviates from the expectation, and the strategy is re-planned; returning to step S302, taking the currently identified material type and adsorption stage as conditions, re-retrieve new rules from the knowledge base, and search for new rules in the association rule set R of the knowledge base;

[0056] Step S503: If no matching rule is found in the knowledge base, an early warning will be issued to promptly notify relevant personnel to manually adjust the relevant parameters of the adsorption system, and the relevant data of the abnormal situation will be recorded and stored.

[0057] Embodiments of the present invention:

[0058] A chemical plant uses a multi-frequency modified adsorption system to treat industrial wastewater containing bisphenol A. The system includes three adsorption towers (numbered B1-B3) and matching control valves. The goal is to reduce the concentration of bisphenol A to ≤0.1mg / L;

[0059] Temperature sensor: one is installed at the bottom, middle and top of the adsorption tower to collect temperature data in real time; taking B1 tower as an example, the bottom temperature T1=25℃, the middle temperature T2=28℃, the top temperature T3=30℃, and the average temperature T=27.67℃; Pressure sensor: installed at the inlet and outlet of the adsorption tower to monitor pressure changes, the inlet pressure of B1 tower C1=0.4MPa, the outlet pressure C2=0.38MPa; Flow sensor: installed at the wastewater inlet pipe and the adsorption reagent injection pipe to measure the flow rate of materials in and out, the wastewater inlet flow rate L1=100m 3 / h, the adsorption reagent is a modified activated carbon suspension, and the injection flow rate is L=5L / h; concentration sensor: installed at the inlet and outlet of the adsorption tower, HPLC is used to detect the concentration of bisphenol A, and the inlet concentration N=50mg / L; position sensor: a position sensor is installed at each valve to obtain the switch status of the valve; spectral sensor: a spectral sensor is installed at the bottom, middle, top and wastewater inlet and outlet of the adsorption tower to obtain spectral data; the characteristic wavelengths of bisphenol A are detected at 278nm and 310nm, and spectral data are obtained at the inlet of the B1 tower: 278 nm intensity 0.8, 310 nm intensity 0.5, which are organized as G=[278,310,0.8,0.5]; the collected data is filtered by a sliding window to remove noise data, the sliding window size is set to 5 data points, and then the Min-Max normalization method is used to unify different types of data into the [0,1] interval;

[0060] Step S200: Collect historical data of the multi-frequency modified adsorption system for the past 6 months, including material type (wastewater containing bisphenol A), frequency combination of physical field F=[120Hz, 180Hz, 600Hz], intensity I=25W / m 2, action time t=45min, spectral feature vector G= [278,310,0.8,0.5] and corresponding adsorption amount and adsorption selectivity; according to the relationship between adsorption amount and action time, mark each record in the historical data with the corresponding adsorption stage; calculate the adsorption rate v(t)=[Q(t+Δt)-Q(t)] / Δt, set the sampling interval Δt=5min; set the threshold V1=800g / h as the adsorption rate threshold in the initial stage, V2=200g / h as the adsorption rate threshold in the adsorption equilibrium stage, obtain the adsorption amount Q(t)=1000g at time t, the adsorption amount Q (t+Δt)=1067g at time t +5min, then the adsorption rate v(t)=(1067-1000) / (5 / 60)=804g / h; since v(t)>=V1, it is judged that the adsorption process belongs to the initial adsorption stage;

[0061] First, the support threshold smin=0.2 is set. In the initial stage, a candidate set containing a single element is generated: {wastewater containing bisphenol A}, {initial adsorption stage}, {[120Hz,180Hz,600Hz]}, {25W / m²}, {45min}, {[278,310,0.8,0.5]}, {27.67℃}, {0.38MPa}, {100m 3 / h}, {50mg / L}; According to the historical data of the adsorption system, the frequency of each candidate item set in the data set is calculated as s = the number of records containing this item set / the total number of records. The candidate item set {wastewater containing bisphenol A} appears 60 times in 200 records, then s = 0.3>smin, and the candidate item set becomes a frequent item set; With the increase of k value, the final candidate set {material type, adsorption stage, frequency combination, intensity, action time, spectral data, temperature, pressure, flow rate, concentration} is generated by combining the existing frequent item sets; Association rules are generated from the frequent item sets, the antecedents include {material type, adsorption stage, spectral data, temperature, pressure, flow rate, concentration}, and the consequents include {frequency combination, intensity, action time, adsorption amount Q, adsorption selectivity Z}; The confidence threshold cmin is set to 0.7, for each generated association rule: {wastewater containing bisphenol A, initial adsorption stage, [278,310,0.8,0.5], 27.67℃, 0.38MPa, 100m 3 / h, 50mg / L}→{[120Hz,180Hz,600Hz], 25W / m², 45min, adsorption capacity Q=3500g, adsorption selectivity Z=8.5}, the number of records containing the antecedent and the consequent is 30, the number of records containing the antecedent is 35, and the confidence c=the number of records containing the antecedent and the consequent / the number of records containing the antecedent=30 / 35=0.86>cmin, then the selected association rules are stored in the multi-frequency modification knowledge base in a structured manner;

[0062] Step S300: Use the data collected in step S100 as the input vector X = [x1, x2, x3, x4, x5], where x1 is the standardized temperature, x2 is the standardized pressure, x3 is the standardized flow, x4 is the standardized concentration, and x5 is the standardized spectral data; use a deep neural network DNN model for recognition and classification, the model consists of an input layer (5 neurons), 3 hidden layers (12 neurons in each hidden layer) and an output layer (2 neurons, corresponding to the material type and adsorption stage, respectively), and the hidden layer uses a ReLU activation function; use the historical data collected in step S200 as the training set, set the learning rate to 0.0015, and the number of training rounds to 120, and train the DNN. The back propagation algorithm adjusts the weights and biases of the network, and the cross entropy loss function is used to optimize the model parameters; the real-time data vector X is input into the trained DNN model, and the predicted category of the material type is output, ML = wastewater containing bisphenol A, and the adsorption stage S = initial adsorption stage; according to the identified material type and adsorption stage, the matching association rules are searched in the multi-frequency modification knowledge base. The association rule set in the knowledge base is R = {r1, r2, r3}, where r1 is the above-mentioned rule {wastewater containing bisphenol A, initial adsorption stage, [278, 310, 0.8, 0.5], 27.67℃, 0.38MPa, 100m 3 / h, 50mg / L}→{[120Hz,180Hz,600Hz], 25W / m², 45min, adsorption amount Q=3500g, adsorption selectivity Z=8.5} traverse the rule set R, find the rule that satisfies Mj=ML and Sj=S, and extract the corresponding frequency combination F=[120Hz,180Hz,600Hz], intensity I=25W / m², action time t=45min of the physical field from rule r1 to generate a multi-frequency modification strategy;

[0063] Step S400: The frequency combination F=[120Hz, 180Hz, 600Hz] in the multi-frequency modification strategy is used to generate a multi-frequency electromagnetic field through a frequency synthesizer, and is sent to the physical field generator of the adsorption tower B1 through a programmable logic controller (PLC); at the same time, according to the intensity I=25W / m² in the strategy, the power output of the physical field transmitter is adjusted; a timer is set to control the duration of the physical field to 45 minutes;

[0064] According to the real-time flow sensor, adjust the opening of the first type of valve to control the real-time flow of wastewater to Qmax=100m 3 / h; When switching the adsorption stage, open and close the first type of valves in a preset order to maintain the continuous flow of the wastewater; According to the concentration data of bisphenol A in the wastewater fed back by the concentration sensor, dynamically adjust the opening degree of the second type of valves to control the injection amount of the adsorption reagent;

[0065] In the initial adsorption stage, the material flow rate Q = Qmax = 100 m 3 / h, and the physical field intensity and action time are set to the corresponding values in the multi-frequency modification strategy; When the adsorption rate v(t) drops to V2 < v(t) <= V1, enter the intermediate adsorption stage, reduce the wastewater flow rate to 60 m 3 / h, and at the same time increase the physical field frequency to F = [150 Hz, 200 Hz, 700 Hz]; When the adsorption rate v(t) <= V2, reach the adsorption equilibrium stage, close the first type of valves, stop the input of the wastewater, and maintain the physical field action until the set target time t = 45 min is reached, completing the modification process of the adsorbent;

[0066] Step S500: Real-time collect the concentration data of bisphenol A at the outlet of adsorption tower B1 as N = 0.08 mg / L; Calculate the adsorption capacity Q of the current adsorption process as Q = (50 - 0.08) × 100 × (45 / 60) = 3744 g; The adsorption capacity of coexisting impurities (phenol) is 400 g, then the selective adsorption index Z = 3744 / 400 ≈ 9.36. Organize the three data of the outlet concentration, adsorption capacity, and adsorption selectivity index to form a new data vector Y = [0.08 mg / L, 3744 g, 9.36]; Combine the preprocessed data vector Y with the real-time collected data such as temperature and pressure as the input, input it into the DNN model in step S300, output the predicted material type and adsorption stage, and match the corresponding adsorption capacity Q = 3500 g and adsorption selectivity Z = 8.5 according to the multi-frequency modification knowledge base; Calculate the deviation between the two, the adsorption capacity deviation ΔQ = |3744 - 3500| / 3500 = 0.07, and the adsorption selectivity deviation ΔZ = |9.36 - 8.5| / 8.5 ≈ 0.10; Set the deviation threshold to 0.15. Since both the adsorption capacity deviation and the adsorption selectivity deviation do not exceed the threshold, it is determined that the actual adsorption effect meets the expectation, and the system continues to operate according to the current strategy.

[0067] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. Intelligent control method of multi-frequency modified adsorption system, characterized by: The multi-frequency modified adsorption system includes a plurality of adsorption towers and valves, and the control method includes: Step S100: installing a variety of sensors on each adsorption tower to collect various data and the switch status of each valve in real time, and pre-processing the collected data; Step S200: collecting historical data of the multi-frequency modification adsorption system, analyzing the corresponding relationship between different material types and multi-frequency modification parameters, and constructing a multi-frequency modification knowledge base; Step S300: Intelligently identify and classify the data collected in real time to identify the material type and adsorption stage; based on the intelligent identification results, combined with the multi-frequency modification knowledge base, adjust the frequency, intensity and action time of the physical field to generate a multi-frequency modification strategy; Step S400: According to the real-time collected state data of the multi-frequency modified adsorption system and the multi-frequency modification strategy, the working state of the adsorption tower and the valve is controlled, and the multi-frequency modified adsorption system is dynamically adjusted to perform corresponding operations; Step S500: Collect data at the outlet of the adsorption tower, analyze the actual adsorption effect, and compare it with the expected effect, set a deviation threshold, if the deviation exceeds the threshold, indicating that the actual effect deviates from the expected, then retrieve the adaptation rules from the multi-frequency modification knowledge base again; if the adaptation rules are not found, timely warning, manual adjustment of system parameters and record storage.

2. The intelligent control method of the multi-frequency modified adsorption system according to claim 1, characterized in that: In step S100, temperature sensors, pressure sensors, flow sensors and concentration sensors are installed at different heights and positions of each adsorption tower, which are used to collect and monitor the temperature T, pressure C, material inlet and outlet flow L and concentration N of the target component in the material in real time; position sensors are installed at each valve to obtain the switch status of the valve; spectral sensors are installed at different positions inside the adsorption tower and at the inlet and outlet of the material to obtain spectral data G; the collected data is filtered through a sliding window to remove noise data, and the Min-Max normalization method is used to unify different types of data into the [0,1] interval.

3. The intelligent control method of the multi-frequency modified adsorption system according to claim 2, characterized in that: The step S200 includes: Step S201: collecting historical data from the adsorption system over a period of time in the past, including material type M, frequency combination F=[f1, f2, ..., fi] of the physical field, intensity I, action time t, spectral data, and corresponding adsorption amount Q and adsorption selectivity Z, wherein fi represents the i-th frequency component; the material type M represents the type of material entering the adsorption tower; the frequency combination F of the physical field represents the frequency combination of the physical field applied during the adsorption process; the intensity I is the intensity of the physical field, which represents the energy of the physical field acting on the material and the adsorbent; the action time t represents the duration of the continuous action of the physical field; the spectral data G includes the wavelength and intensity of the spectral curve, which is used to identify the material type; the adsorption amount Q is the amount of the target component absorbed by the adsorbent; the adsorption selectivity Z is the ratio of the adsorbent to the target component compared to other coexisting components; According to the relationship between the adsorption capacity and the action time, each record in the historical data is marked with the corresponding adsorption stage S, including the initial adsorption stage, the intermediate adsorption stage, and the adsorption equilibrium stage; calculate the adsorption rate: v(t) = [Q(t + Δt) - Q(t)] / Δt; where v(t) represents the adsorption rate at time t, Q(t) represents the adsorption capacity at time t, Δt is the set sampling interval, and Q(t + Δt) represents the adsorption capacity at time t + Δt; set the threshold V1 as the adsorption rate threshold in the initial stage and V2 as the adsorption rate threshold in the adsorption equilibrium stage; when v(t) >= V1, it is set that the adsorption process belongs to the initial adsorption stage; when V2 <= v(t) < V1, it is set that the adsorption process enters the intermediate adsorption stage; when v(t) <= V2, it is set that the adsorption process reaches the adsorption equilibrium stage.

4. The intelligent control method of the multi-frequency modified adsorption system according to claim 1, characterized in that: The step S200 further includes: Step S202: First, set the support threshold smin to measure the frequency of occurrence of candidate item sets in the dataset; generate candidate sets Ck, where k represents the number of elements in the item set; in the initial stage, generate candidate sets containing a single element: {material type}, {adsorption stage}, {frequency combination}, {intensity}, {action time}, {spectral data}, {temperature}, {pressure}, {flow rate}, {concentration}; according to the historical data of the adsorption system, calculate the frequency s of each candidate item set in the dataset: s = number of records containing this item set / total number of records; filter out the item sets with frequency s greater than smin to obtain frequent item sets Lk; as the value of k increases, generate the final candidate set {material type, adsorption stage, frequency combination, intensity, action time, spectral data, temperature, pressure, flow rate, concentration} by combining the existing frequent item sets; generate association rules from the frequent item sets, with the antecedent including {material type, adsorption stage, spectral data, temperature, pressure, flow rate, concentration} and the consequent including {frequency combination, intensity, action time, adsorption capacity, adsorption selectivity}; set the confidence threshold cmin, and for each generated association rule, calculate its confidence c = number of records containing the antecedent and the consequent / number of records containing the antecedent; filter out the association rules with confidence greater than the threshold cmin and store the filtered association rules in the multi-frequency modification knowledge base in a structured manner.

5. The intelligent control method of the multi-frequency modified adsorption system according to claim 1, characterized in that: The step S300 includes: Step S301: The data collected in step S100 is used as input, recorded as vector X=[x1,x2,...,xn], where xn represents the nth feature, including temperature, pressure, flow, concentration, and spectral data; a deep neural network DNN model is used for identification and classification, and the model consists of an input layer, multiple hidden layers, and an output layer, and the hidden layer uses a ReLU activation function; the historical data collected in step S200 is used as a training set, where the material type and adsorption stage are used as target outputs; the learning rate and the number of training rounds are set to train the DNN, and the weights and biases of the network are adjusted by the back propagation algorithm, and the model parameters are optimized by the cross entropy loss function; the real-time data vector X is input into the trained DNN model, and the predicted category ML of the material type and the adsorption stage S are output; Step S302: According to the identified material type and adsorption stage, search for matching association rules in the multi-frequency modification knowledge base constructed in step S200; set the association rule set in the knowledge base to R={r1, r2, ..., rj}, where rj represents the jth rule in the set R, and the rule rj is represented by a twelve-tuple (Mj, Sj, Fj, Ij, tj, Gj, Tj, Cj, Lj, Nj, Qj, Zj), including the corresponding material type, adsorption stage, physical field frequency combination, physical field intensity, action time, spectral data, temperature, pressure, flow rate, concentration, adsorption amount, and adsorption selectivity; traverse the rule set R to find the rule that satisfies Mj=ML and Sj=S. If there are multiple matching rules, select the rule with the highest confidence as the final rule; extract the physical field frequency combination F, intensity I, and action time t from the final rule to generate a multi-frequency modification strategy.

6. The intelligent control method of the multi-frequency modified adsorption system according to claim 1, characterized in that: The step S400 includes: Step S401: The frequency combination F in the multi-frequency modification strategy is used to generate a multi-frequency electromagnetic field through a frequency synthesizer and sent to the physical field generator of the adsorption tower through a PLC; at the same time, according to the intensity I in the strategy, the power output of the physical field transmitter is adjusted; a timer is set to control the duration of the physical field action; according to the real-time flow sensor, the opening of the first type of valve is adjusted to control the real-time flow of the material; when switching between adsorption stages, the first type of valve is opened and closed in a preset order to maintain the continuous flow of the material; according to the concentration data of the target component in the material fed back by the concentration sensor, the opening of the second type of valve is dynamically adjusted to control the injection amount of the adsorption agent; the first type of valve is a related valve for controlling the flow of the fluid, and the second type of valve is a related valve for adjusting the modification process.

7. The intelligent control method of the multi-frequency modified adsorption system according to claim 3, characterized in that: The step S400 further includes: Step S402: During the adsorption process, the adsorption amount Q of each adsorption tower is monitored in real time, and the adsorption stage of the adsorption tower is dynamically switched according to the change of the adsorption amount; in the initial adsorption stage, the maximum flow rate of the material is set to Qmax, and in the initial adsorption stage, the material flow rate Q=Qmax; the physical field intensity and action time are set to the corresponding values ​​in the multi-frequency modification strategy; in the mid-term adsorption stage, the material flow rate is reduced and the physical field frequency is increased; in the adsorption equilibrium stage, the first type of valve is closed, the material input is stopped, and the physical field action is maintained until the set target time t is reached, and the modification process of the adsorbent is completed.

8. The intelligent control method of the multi-frequency modified adsorption system according to claim 1, characterized in that: The step S500 includes: Step S501: collect the concentration data of the target component at the outlet of the adsorption tower in real time, calculate the adsorption amount of the current adsorption process = (inlet target component concentration - outlet target component concentration) × material flow × adsorption time; determine the selective adsorption index Z = (target component adsorption amount / coexisting impurity adsorption amount); sort out the three data of outlet concentration, adsorption amount and adsorption selectivity index to form a new data vector Y = [outlet concentration, adsorption amount, adsorption selectivity]; Step S502: taking the preprocessed data vector Y as input, inputting it into the DNN model in step S300, outputting the predicted material type and adsorption stage, and matching the corresponding adsorption amount and adsorption selectivity according to the multi-frequency modification knowledge base, and taking the model prediction result as the actual adsorption effect; comparing the actual adsorption effect with the expected result in the multi-frequency modification strategy, and calculating the deviation between the two; setting a deviation threshold, when the adsorption amount deviation or adsorption selectivity deviation exceeds the threshold, it is determined that the actual adsorption effect deviates from the expectation, and the strategy is re-planned; returning to step S302, taking the currently identified material type and adsorption stage as conditions, re-retrieve new rules from the knowledge base, and search for new rules in the association rule set R of the knowledge base; Step S503: If no matching rule is found in the knowledge base, an early warning will be issued to promptly notify relevant personnel to manually adjust the relevant parameters of the adsorption system, and the relevant data of the abnormal situation will be recorded and stored.

Images