A terminal ultrapure water quality intelligent prediction device based on an RBF neural network algorithm model and a prediction and control method

Abstract

The application discloses a terminal ultrapure water quality intelligent prediction device and a prediction and control method based on an RBF neural network algorithm model, the prediction device is arranged on a bypass before an original water pool of an ultrapure water preparation system, and comprises a processing system, a detection system and a water quality prediction system; the processing system is used for pretreating original water; the detection system is used for detecting pretreated water in real time, and acquiring pretreated water parameter real-time data; the water quality prediction system adopts an RBF neural network to predict the TOC and B content in terminal ultrapure water according to the pretreated water parameter real-time data, obtains a prediction result, and adjusts the amount of a reagent in the ultrapure water preparation system in combination with historical reagent dosing amounts, so that the TOC and B content in the terminal ultrapure water are stable and up to the standard. The application can predict the TOC and B content in ultrapure water produced at the original water end, can compare historical data, can regulate and control the reagent dosing amount, and can ensure that the TOC and B content in the terminal ultrapure water are stable and up to the standard.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electronic-grade ultra-pure water production, and particularly relates to a terminal ultra-pure water quality intelligent prediction device based on an RBF neural network algorithm model and a prediction and control method. BACKGROUND

[0002] With the improvement of the process, the integrated circuit industry has increasingly strict requirements on the quality of ultra-pure water. Trace pollutants such as total organic carbon (TOC) and boron (B) are prone to affect the water quality of the electronic ultra-pure water system. Taking TOC as an example, the TOC requirement of ultra-pure water used in 12-inch integrated circuit production is less than 1 μg / L or even 0.5 μg / L at present, and a Sievers M500e instrument is usually used to monitor the terminal water production. The B requirement is even lower, and the B in the ultra-pure water usually needs to be less than 50 ng / L or even less than 10 ng / L at present, and only a Sievers boron meter can be used for online detection at the terminal.

[0003] The detection data at the terminal can reflect the water quality at the present stage, but there is a high lag. For example, CN107758942A introduces an ultra-pure water manufacturing system, which has an ultra-pure water detection system in the patent. The detection system is located after the multi-stage resin column and the filter, that is, at the terminal water production position. When using this detection system, if the water quality of the terminal ultra-pure water is poor or even there is a risk of TOC or B exceeding the standard, the system operator will control the TOC and B indexes by adding a fixed proportion of oxidizing agent to the front-end raw water tank or adjusting the pH value of the RO (reverse osmosis) section. However, in the field operation, the overall preparation process of the ultra-pure water has a time span. The raw water stays in the raw water tank for nearly 20 h, and the water body needs to pass through the pretreatment system, the process system and finally the polishing system from the raw water tank to the terminal water outlet. The water body stays in the RO water tank and the pure water tank for nearly 10 h. That is to say, after the TOC and B of the terminal ultra-pure water do not meet the standard, adding a senior oxidizing agent to the raw water tank for TOC removal or adjusting the pH of the RO section for B removal will still affect the water quality of the subsequent 10 h or so, affect the normal water use of the integrated circuit industry, and cause great economic losses.

[0004] The above-mentioned method of feeding back the front end after judging the TOC and B content of the terminal water production has the problems of water quality prediction deficiency and control measure lag. SUMMARY

[0005] The present application aims to provide a terminal ultra-pure water quality intelligent prediction device based on an RBF (radial basis function) neural network algorithm model and a prediction and control method, for the prediction and control of the TOC and B content in the ultra-pure water.

[0006] Technical scheme: The pre-judgment device is arranged on the bypass before the raw water pool of the ultrapure water preparation system, comprising a treatment system, a detection system and a water quality pre-judgment system, the treatment system is used for pretreating raw water and outputting pretreated water meeting the requirements of the detection system; the detection system is used for real-time detection of the pretreated water to obtain pretreated water parameter real-time data; the water quality pre-judgment system uses an RBF neural network algorithm model to pre-judge the total organic carbon TOC and boron B content in the terminal ultrapure water according to the pretreated water parameter real-time data, obtains a prediction result, and adjusts the dosage of the reagent in the ultrapure water preparation system in combination with the historical reagent dosage, so that the TOC and B content in the terminal ultrapure water are stable and up to standard. The TOC and B content in the ultrapure water can be pre-judged at the raw water end, historical data can be compared, the reagent dosage can be regulated, and the stable and up-to-standard TOC and B content in the terminal ultrapure water can be ensured.

[0007] Optionally, the treatment system comprises an ultrafiltration (UF) unit, a reverse osmosis (RO) unit and an ion exchange resin tower connected in sequence.

[0008] Optionally, the pretreated water parameter real-time data comprises resistivity, TOC content, small molecule organic matter content and B content data.

[0009] Optionally, the RBF radial basis neural network algorithm model formula is as follows:

[0010]

[0011] wherein, y j represents the jth result output after hidden layer processing, i.e., the terminal product water TOC content and B content predicted according to the pretreated water quality, ω ij represents the weight value of the ith input value to the jth output value; h is the number of hidden layer nodes; x c is the center value of the base function, i.e., the clustering center, i represents the input value number of the neural network algorithm, j represents the output value number of the neural network algorithm, c i is the water parameter real-time data of the ith sample in the base function, d m"x represents the maximum distance between the ith sample value x i and the clustering center in the input base function.

[0012] Optionally, the raw water flow rate of the bypass is 20-50 L / h, the overall residence time in the water quality pre-judgment device is 10-20 min, and the bypass outlet water is supplemented into the filter water pool of the ultrapure water preparation system.

[0013] The application also designs a terminal ultrapure water quality intelligent pre-judgment and control method based on the RBF neural network algorithm model based on the pre-judgment device, comprising the following steps:

[0014] The raw water is pretreated to obtain pretreated water parameter real-time data, the water quality is pre-judged according to the pretreated water parameter real-time data, and a pre-judgment result is output;

[0015] According to the pre-judgment result, the historical reagent dosing amount is combined to control the reagent dosing amount in the dosing system, so as to realize water quality pre-judgment and control.

[0016] Further, the raw water is pretreated to obtain pretreated water parameter real-time data, the water quality is pre-judged according to the pretreated water parameter real-time data, and a pre-judgment result is output; specifically including:

[0017] (1) Pretreating the raw water to obtain pretreated water meeting the detection requirements;

[0018] (2) Real-time detection of the pretreated water to obtain pretreated water parameter real-time data, including resistivity, TOC content, small molecule organic matter content and B content data;

[0019] (3) Using an RBF neural network algorithm model to process the pretreated water parameter real-time data to obtain the predicted values of the TOC content and B content in the terminal ultrapure water, i.e. the pre-judgment result.

[0020] Further, the pretreatment of the raw water in step (1) is specifically:

[0021] The UF membrane in the UF unit removes colloids and suspended particles in the raw water, the RO membrane in the RO unit removes ions, bacteria and viruses in the raw water, and the ion exchange resin further removes salt, which is used for pretreating the raw water.

[0022] Further, step (3) is specifically:

[0023] The pretreated water parameter real-time data is processed by a normalization function; the normalization function is:

[0024]

[0025] Wherein, X i ' is the data obtained after the pretreated water parameter real-time data is normalized, x i is the water parameter real-time data of the i-th sample in the base function; x min is the minimum value of the sample data; x max is the maximum value of the sample data;

[0026] The cluster center of the normalized pretreated water parameter real-time data is obtained, and the number of cluster centers is used as the number of hidden layer nodes in the RBF neural network algorithm model;

[0027] The variance σ of the base function is solved;

[0028] According to the cluster center point xC , data distance clustering center maximum distance d m"x And the number of clustering centers, the RBF radial basis function neural network algorithm model expression is obtained as:

[0029]

[0030] Wherein, y j Represent the output of the hidden layer processing after the jth result, that is, the terminal water TOC content and B content predicted according to the pretreated water quality, j represents the number of output values of the neural network algorithm, and h is the number of hidden layer nodes.

[0031] Let Then the above formula is rewritten as:

[0032]

[0033] An electronic device for storing and executing the method, comprising:

[0034] A memory storing executable program code;

[0035] A processor coupled to the memory;

[0036] The processor calls the executable program code stored in the memory to execute the steps of the terminal ultrapure water quality intelligent prediction and control method based on the RBF neural network algorithm model.

[0037] Beneficial effects: compared with the prior art, the significant technical effects of the present application are: (1) the TOC and B content control process is changed from problem finding-solution to risk prediction-risk avoidance, which saves a lot of TOC and B content control time and water quality recovery time; (2) a kind of intelligent risk prediction method based on RBF neural network algorithm model is proposed, which is simulated by large data training, a water quality prediction device is connected before the raw water tank, a processing system with a processing time of only 10-20 minutes and a plurality of water quality detection equipment are equipped in the prediction device, and real-time data are input into the RBF neural network algorithm model with a large number of sample data training and output result convergence, with the aid of the prediction result of the RBF neural network algorithm model, the change of TOC and B content in the terminal ultrapure water can be accurately predicted when the raw water quality fluctuates, and feedback is given to the TOC and B removal unit to adjust the dosage of reagent in real time, ensuring the stability of the water quality; (3) the RBF neural network algorithm model is trained in combination with a large amount of data in the early stage, which is suitable for ultrapure water preparation system providing training sample data, and also has reference significance for ultrapure water preparation systems with similar process flow. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1A schematic diagram of a traditional mode terminal ultrapure water quality control process;

[0039] Figure 2 A schematic diagram of a terminal ultrapure water quality control process in a smart prediction mode;

[0040] Figure 3 A terminal ultrapure water quality model structure for RBF prediction DETAILED DESCRIPTION

[0041] The application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] Figure 1 A schematic diagram of a traditional semiconductor ultrapure water preparation system, in which a senior oxidizing agent can be added at the raw water tank to ensure that the TOC value in the produced water is stable, and a pH adjuster can be added at the RO unit to ensure that the B value in the produced water is stable. In the traditional mode, the control process of TOC and B content in the terminal ultrapure water has a lag, which belongs to the process from discovering problems to solving problems. From the raw water tank to the ultrapure water use point, there is a quite long hydraulic retention time, usually more than 20h. From the RO produced water tank to the use point, there is usually more than 10h. When the abnormal signal of terminal TOC and B data is fed back to the front-stage dosing system, the water quality in the filter tank and the RO produced water tank has already been substandard. From the start of the dosing system to the time period when the produced water TOC is qualified, the produced water may have water quality problems.

[0043] As shown in Figure 2 The ultrapure water quality prediction device of the application is arranged on a bypass before the raw water tank of the traditional semiconductor ultrapure water preparation system, and a stream of raw water is introduced into the water quality prediction device. The water quality prediction device comprises a treatment system, a detection system and a water quality prediction system. The treatment system comprises UF (ultrafiltration) unit, RO unit and ion exchange resin tower connected in sequence. The UF membrane in the UF unit removes colloids and suspended particles in the raw water. The RO reverse osmosis membrane in the RO unit removes ions and bacteria viruses in the raw water. The ion exchange resin further desalts, and is used for pretreating the raw water, improving the raw water quality, preventing the precise unit in the detection system from being contaminated and clogged, and making the effluent meet the water inlet requirement of the detection system. The detection system is used for real-time detection of the pretreated water, and obtains real-time data of the pretreated water. The detection system imports the obtained real-time data of the pretreated water into the water quality prediction system. The water quality prediction system predicts the TOC and B content in the terminal ultrapure water according to the real-time data of the pretreated water, outputs accurate prediction results, and feeds back the adding amount information of the pH adjuster and bromide salt / oxidizing agent to the dosing system in combination with historical dosing amount, so as to adjust the dosing amount of the TOC removal unit (i.e. the dosing point of the raw water tank) and the B removal unit (i.e. the dosing point of the secondary reverse osmosis unit) in the system, and to ensure that the TOC and B content in the terminal ultrapure water is stable and up to standard.

[0044] In this embodiment, the bypass water flow rate before the raw water enters the raw water pool is 20-50 L / h, and the overall residence time of the water quality prediction device is 10-20 min. The bypass effluent (i.e. the detection system effluent) will be supplemented into the filtered water pool. In the treatment system composed of UF, RO and ion exchange resin tower, the residence time of the raw water is 4-7 min, which serves to improve the raw water quality and make the effluent meet the detection system influent requirements. The detection system includes resistivity detection instruments, TOC detection instruments, small molecule organic matter detection instruments and B tables, and can obtain the resistivity, TOC content, small molecule organic matter content and B content data of the pretreated raw water. The detection system will detect the real-time data of the four parameters (TOC, small molecule organic matter content, resistivity and B content) in the pretreated water according to the preset frequency (in this embodiment, detection is performed once every 15 min). The intelligent prediction model in the water quality prediction system predicts the TOC and B content in the terminal ultrapure water according to the input real-time data of the four parameters, compares with the actual terminal ultrapure water quality, performs error convergence, iteratively trains the intelligent prediction model, and outputs the accurate prediction result until the prediction model converges to the limit (error less than 0.05) and feeds back to the dosing system. In the ultrapure water preparation process, the raw water pool and the secondary RO unit will adjust the dosing amount according to the requirements of the qualified ultrapure water quality, so as to ensure the stability of the water quality of the filtered water pool, the RO water pool and the ultrapure water point. Compared with the traditional mode, the TOC and B treatment feedback time of the ultrapure water point can be shortened to within 20 min after using the intelligent prediction method.

[0045] The four parameters output by the detection system fluctuate irregularly and present a high degree of nonlinearity, with many dimensions. The intelligent prediction model in the water quality prediction system is an RBF neural network algorithm model, which is a feedforward neural network with unique optimal approximation (overcoming the problem of local minimum value), simple training, fast learning convergence speed and other good performances, and is suitable for rapid simulation of a large amount of data and fitting of multivariate data. As shown in Figure 3 , the RBF neural network algorithm model includes an input layer, a hidden layer and an output layer. The input data are TOC content, small molecule organic matter content, B content and resistivity. After hidden layer processing, TOC and B are output.

[0046] Further, the formula of the initial RBF radial basis neural network algorithm model is as follows:

[0047]

[0048] wherein ω ij represents the weight value from the i-th hidden layer node to the j-th output value; σ represents the variance of the base function (also referred to as the base function amplitude), and the base function here is φ(x i ,x c ) = ||x i -xc ||;h is the number of hidden layer nodes; x c is the center value (cluster center) of the base function, i represents the number of input values of the neural network algorithm, j represents the number of output values of the neural network algorithm, x i represents the input value of the i-th sample in the base function, y j represents the output of the j-th result after the hidden layer processing, that is, the terminal product water TOC, B content predicted according to the pretreated water sample water quality.

[0049] Before obtaining accurate prediction results, the RBF neural network algorithm model needs to be simulated and trained with a large amount of sample data. By inputting the four types of parameters obtained by the detection system into the RBF neural network algorithm model, the prediction result is obtained. According to the difference between the prediction result and the actual water quality information result of the ultrapure water use point, the neural network is iterated. When the water quality error value is less than 0.05, the iteration is ended, and it is considered that the prediction model converges to the limit. At this time, it can be considered that the prediction result has high accuracy, and the expected water quality of the terminal ultrapure water can be fed back according to the prediction result of the water quality prediction system.

[0050] The water quality prediction system feeds back the signal to the rear-end dosing program according to the prediction value of TOC and B content in the ultrapure water of the ultrapure water use point, combines the historical reagent dosing amount, and adjusts the reagent amount of the TOC and B removal unit to ensure that the TOC and B content in the terminal ultrapure water is stable and meets the standard.

[0051] The water quality prediction system stores the reagent dosing amount in the bromide salt / oxidizing agent dosing unit of the raw water pool in the historical ultrapure water preparation system for removing TOC and the pH adjuster dosing amount in the secondary reverse osmosis unit dosing unit for removing boron. The prediction result will be compared with the corresponding reagent dosing amount of TOC and B content in the historical ultrapure water, and the dosing amount will be adjusted in the raw water pool and the primary RO unit at any time.

[0052] The method for intelligently predicting and controlling the terminal ultrapure water quality by using the ultrapure water quality prediction device comprises the following steps:

[0053] S1, pretreat the raw water by the ultrapure water quality prediction device, obtain real-time data of the pretreated water, predict the water quality according to the real-time data of the pretreated water, and output the prediction result; specifically:

[0054] (1) pretreat the raw water by the treatment system composed of UF, RO and ion exchange resin, improve the water quality of the raw water, and obtain pretreated water meeting the water inlet requirements of the detection system.

[0055] (2) detect the pretreated water in real time by the detection system to obtain real-time data of the pretreated water parameters.

[0056] First, it needs to be clear which water parameters will affect the TOC content and B content in the produced water. The water parameters usually include: flow rate, conductivity, pH, small molecule organic matter content (urea, etc.), TOC, silicon, B, etc. Through a large number of sample data screening, in this embodiment, the TOC content, small molecule organic matter content (LW), B content, and resistivity of the four water parameters have a greater impact on the TOC and B content of the produced water. After the raw water is treated by the water quality prediction device, the effluent is detected by the detection system to obtain the above four parameters as the key influencing factors, and then as the input factors of the water quality prediction system.

[0057] In this embodiment, the resistivity detector is used to detect the resistivity of the pretreated water in real time, the TOC detector is used to detect the TOC content of the pretreated water in real time, the small molecule organic matter detector is used to detect the small molecule organic matter content in the pretreated water in real time, and the B table is used to detect the B content of the pretreated water in real time. The resistivity, TOC content, small molecule organic matter content, and B content data of the raw water after pretreatment are obtained.

[0058] (3) The water quality prediction system processes the real-time data of the pretreated water parameters to obtain the predicted values of the TOC and B content in the produced water at the water point, i.e. the prediction result. Specifically:

[0059] First, the four key influencing factors (i.e. the real-time data of the pretreated water parameters) obtained by the detection system are normalized by the following function:

[0060]

[0061] Wherein, X i is the data obtained after the real-time data of the pretreated water parameters is normalized, x i is the input value of the i-th sample in the base function, i.e. the real-time data of the water parameters; x min is the minimum value of the sample data; x m"x is the maximum value of the sample data. Normalization limits the data between [-1, 1], and the algorithm is easier to converge to the optimal solution.

[0062] Further, the number of base function center points in the RBF radial basis neural network algorithm model can be solved by using the K-means mean clustering method or the subtraction clustering algorithm, and the number of clustering centers x c is obtained as the number of RBF hidden layer nodes h.

[0063] Further, when the hidden layer activation base function is selected as the Gaussian function, after the clustering center x c and the number of clustering center points h are determined, the variance σ of the base function can be solved by the following formula.

[0064]

[0065] Where, d max The i-th sample value x represents the input to the basis function. i The maximum distance between the cluster center and the cluster center is calculated using the Euclidean distance function.

[0066] Furthermore, the cluster centers, d, are obtained. max After determining the number of cluster centers, substituting them into equation (1) yields the following expression:

[0067]

[0068] in, Let be the activation function expressed by the Gaussian function, with Expression. Command Equation (4) can then be rewritten as:

[0069]

[0070] Furthermore, the pseudo-inverse method is used to solve for the function weight ω. ij Let J = J(x) i Let J be the expected output value of the terminal TOC and B content for the function fitting. ij For the i-th input vector x i At the j-th output node y j The expected output value, ω kj Let be the weights from the k-th hidden node to the j-th output node, i∈(1,n), k∈(1,n), k∈(1,h). Then the weight matrix {ω kj The following formula can be used to solve it:

[0071] {ω kj}=G + J (6)

[0072]

[0073] Among them G @ G = {g} iA The pseudo-inverse matrix of} can be solved using singular value decomposition (SVD). iA Represents the i-th input vector x i The output value of the kth hidden layer node has a total of k input vectors.

[0074] Suppose G is an M-row × N-column matrix. Performing singular value decomposition on G, we get [U,S,V] = SVD(G), which yields matrices U, S, and V respectively.

[0075] wherein U is a MxM matrix, the column vectors in U are left singular vectors, V is a NxN matrix, the column vectors in V are called right singular vectors; S is a MxN diagonal matrix; the main diagonal elements of S matrix are singular values of matrix G, S1=1. / S, is an NxM matrix determined by singular values. The above matrices have the following relationship:

[0076] G=U×S×V T (8)

[0077] G + =V×S1×U T (9)

[0078] Further, based on the RBF function weight ω solved above, the expected output value y corresponding to the input parameter can be obtained by inputting formula (5).

[0079] S2, according to the prediction result, combining the historical dosage of the medicament, controlling the dosage of each medicament in the dosing system, realizing water quality prediction and control;

[0080] In this embodiment, according to the prediction result, the signal is fed back to the rear-end dosing program in combination with the historical dosage of the medicament, and the dosage of the medicament in the raw water pool bromide / oxidizing agent dosing unit for removing TOC and the dosage of the medicament in the secondary reverse osmosis unit dosing unit for removing boron are adjusted to ensure that the TOC and B content in the terminal ultrapure water meet the stability requirements.

[0081] Next, taking an electronic enterprise as an example for further illustration, the TOC content of the ultrapure water of the electronic enterprise needs to be less than 1 μg / L, and the boron content of the effluent needs to be less than 50 ng / L. The water source of the electronic enterprise is a mixture of tap water and reclaimed water. Compared with municipal tap water, the TOC and small molecule organic matter content in reclaimed water fluctuate more obviously, and the impact load problem of the ultrapure water preparation system caused by water quality fluctuation needs to be concerned at all times.

[0082] In this embodiment, by inputting 4 dimensions (TOC content, small molecule organic matter content, B content, and resistivity) parameters, a total of 200 input quantities, when the iteration number is 105 times, the RBF prediction model error value is less than 0.05, the model converges to the limit, the prediction result obtains the optimal solution, and the error with the measured value approaches the limit, and the model structure is as shown in Figure 3 .

[0083] When the RBF prediction model converges to the limit, the prediction result obtains the optimal solution, and the water TOC and B content can be predicted by the water quality data of the prediction device. When the raw water enters the prediction device, the detection system result is x (TOC content, small molecule organic matter content, B content, resistivity) = (0.5, 0.9, 0.8, 0.3), and the BRF model output prediction result y1 value is 1.2, that is, the terminal ultrapure water TOC content is predicted to be 1.2 μg / L, and the y2 value is 0.9, that is, the terminal B content is predicted to be 45 ng / L, that is, the TOC has an over-standard risk, and the B content index does not have an over-standard risk.

[0084] The prediction signal will be transmitted to the advanced oxidation system in the raw water tank, and the dosage of sodium bromide and oxidant added to the raw water tank will be increased for the specific removal of small molecules according to the comparison of historical data. If the B content index is over-standard, the pH regulator can be added in the secondary RO unit stage to increase the pH value, so as to improve the B removal rate.

[0085] After 15 minutes, the raw water passes through the pretreatment system again and enters the detection system, and the latest detection data is obtained. Then the detection data is imported into the intelligent prediction model, the expected y value is obtained according to the RBF model, and the dosage of the TOC and B content removal process is adjusted by comparing the historical data. This step is repeated to ensure that the water in the raw water tank can still meet the requirements of the terminal ultrapure water TOC and B content after being treated by the subsequent process.

[0086] The present application proposes an intelligent prediction method for the organic matter and B content in the terminal ultrapure water. The method needs to connect a water quality prediction device in front of the raw water tank. The device has a treatment system, a detection system and a water quality prediction system. The water quality prediction system can predict the TOC and B content in the terminal water by detecting the water quality after pretreatment, which has the advantages of being timely and accurate. The second method is to build an RBF neural network algorithm model in the prediction device. By inputting the water quality of the prediction device into the RBF neural network algorithm model, the water quality of the prediction device and the actual terminal ultrapure water can be compared, and the neural network can be iterated. After a large amount of sample data is simulated and trained, when the difference between the prediction result and the actual result is less than 0.05, it is considered that the function converges to the limit, and the prediction result can accurately represent the terminal ultrapure water quality. The method can predict the TOC and B content in the terminal ultrapure water at the raw water end, and can compare the historical data to adjust the dosage of the reagent in the ultrapure water preparation system, so as to ensure that the TOC and B content in the terminal ultrapure water meet the standard stably.

Claims

1. A terminal ultrapure water quality intelligent prediction device based on the RBF neural network algorithm model, characterized in that, The device is installed on the bypass line in front of the raw water tank of the ultrapure water preparation system. It includes a treatment system, a detection system and a water quality prediction system. The treatment system is used to pre-treat the raw water and output pre-treated water that meets the requirements of the detection system. The detection system is used to monitor the pretreated water in real time and obtain real-time data on the pretreated water parameters. The water quality prediction system uses the RBF neural network algorithm model to predict the total organic carbon (TOC) and boron (B) content in the terminal ultrapure water based on the real-time data of the pretreated water parameters, obtains the prediction results, and adjusts the dosage of reagents in the ultrapure water preparation system in combination with the historical reagent dosage to ensure that the TOC and B content in the terminal ultrapure water meets the standards stably. The formula for the RBF radial basis function neural network algorithm model is: ，j=1,2,3……， in, The output after hidden layer processing is the first... The results, namely the predicted TOC and B content in the final product water based on the water quality of the pretreated water samples, Representing the The input value to the first The weight values ​​of each output value; It is the number of hidden layer nodes; = , The center value of the basis function, i.e., the cluster center. The number of input values ​​for a neural network algorithm. This represents the number of output values ​​of a neural network algorithm. The first of the basis functions Real-time water parameter data for each sample The first input to the basis function represents the... Sample value The maximum distance from the cluster center; The flow rate of the raw water introduced into the bypass is 20-50 L / h, and the overall residence time in the water quality prediction device is 10-20 min. The bypass effluent is then added to the filtration tank of the ultrapure water preparation system.

2. The terminal ultrapure water quality intelligent prediction device based on the RBF neural network algorithm model according to claim 1, characterized in that, The treatment system includes an ultrafiltration (UF) unit, a reverse osmosis (RO) unit, and an ion exchange resin tower connected in sequence.

3. The terminal ultrapure water quality intelligent prediction device based on the RBF neural network algorithm model according to claim 1, characterized in that, Real-time data on pretreated water parameters include resistivity, TOC content, small molecule organic matter content, and B content.

4. A prediction and control method for a terminal ultrapure water quality intelligent prediction device based on the RBF neural network algorithm model as described in any one of claims 1 to 3, characterized in that, Includes the following steps: The system pre-treats raw water to obtain real-time data on pre-treated water parameters, makes predictions on water quality based on the real-time data on pre-treated water parameters, and outputs the prediction results. Based on the prediction results and combined with the historical dosage of chemicals, the dosage of each chemical in the dosing system is controlled to achieve water quality prediction and control.

5. The intelligent prediction and control method for terminal ultrapure water quality based on the RBF neural network algorithm model according to claim 4, characterized in that, The process involves pretreating raw water to obtain real-time data on pretreated water parameters, making predictions about water quality based on this data, and outputting the prediction results. Specifically, this includes: (1) Pre-treat the raw water to obtain pre-treated water that meets the testing requirements; (2) Real-time monitoring of pretreated water to obtain real-time data of pretreated water parameters, including resistivity, TOC content, small molecule organic matter content and B content; (3) The RBF neural network algorithm model is used to process the real-time data of pretreated water parameters to obtain the predicted values ​​of TOC content and B content in the terminal ultrapure water, i.e. the prediction results.

6. The intelligent prediction and control method for terminal ultrapure water quality based on the RBF neural network algorithm model according to claim 5, characterized in that, The pretreatment of raw water in step (1) specifically involves: The UF membrane in the UF unit removes colloids and suspended particles from the raw water, the RO membrane in the RO unit removes ions, bacteria and viruses from the raw water, and the ion exchange resin further desalinates the water, thus pretreating the raw water.

7. The intelligent prediction and control method for terminal ultrapure water quality based on the RBF neural network algorithm model according to claim 5, characterized in that, Step (3) is as follows: The real-time data of pretreated water parameters are processed using a normalization function; the normalization function is: ， in, The data is obtained after normalizing the real-time water parameter data. The first of the basis functions Real-time water parameter data for each sample; The minimum value of the sample data; The maximum value of the sample data; Find the cluster centers of the normalized preprocessed real-time water parameter data, and use the number of cluster centers as the number of hidden layer nodes in the RBF neural network algorithm model. Solving for the variance of the basis functions ; Based on cluster centers Maximum distance of data from cluster center And the number of cluster centers, yielding the RBF radial basis function neural network algorithm model expression: ，j=1,2,3……， in, The output after hidden layer processing is the first The results, namely the predicted TOC and B content in the final product water based on the water quality of the pretreated water samples, This represents the number of output values ​​of a neural network algorithm. It is the number of hidden layer nodes; make Then the above formula can be rewritten as: ， in, Representing the The input value to the first The weight values ​​of each output value.

8. An electronic device, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the steps of the terminal ultrapure water quality intelligent prediction and control method based on the RBF neural network algorithm model as described in any one of claims 4-7.

Images