Constant-pressure water supply system for water for semiconductor packaging process

Abstract

The invention relates to the technical field of water supply system control, in particular to a water constant-pressure water supply system for a semiconductor packaging process. In order to overcome the defects of an existing single-sensor constant-pressure water supply system, three pressure sensors are arranged on an external supply pipeline to obtain pressure data, an intermediate value is obtained through comparison to control the operation frequency of an external supply pump, external supply constant pressure is ensured, and sensor faults can be found in time. Meanwhile, water consumption, external supply frequency and electricity consumption information are integrated, the rotating speed of an external supply pump is predicted through a multiple linear regression model, and the rotating speed is stabilized in combination with a closed-loop PID controller. And according to electricity prices and water consumption conditions in different time periods, a linear weighting method is adopted, the frequency of the external supply pump is dynamically adjusted, and energy-saving optimization is achieved. The reliability, the stability and the energy-saving performance of the water supply system are effectively improved, and a more intelligent and efficient solution is provided for constant-pressure water supply control of water for the semiconductor packaging process.

Description

Technical Field

[0001] The present invention relates to the field of water supply system control technology, and specifically to a constant pressure water supply system for semiconductor packaging process water. Background Art

[0002] In the semiconductor packaging process, the constant pressure water supply system is extremely important for the normal operation of the equipment. The traditional constant pressure water supply system usually relies on a single pressure sensor to control the operating frequency of the external supply pump to maintain the pressure of the external supply pipeline constant. However, this method has obvious defects.

[0003] On the one hand, it is difficult for a single sensor to accurately judge its own health status. In the semiconductor packaging environment, there are various complex interference factors. Once the sensor fails, the system cannot obtain accurate pressure information in time, which can easily cause the external supply pump to stop, seriously affecting the continuity and stability of water supply, which may have an adverse effect on the semiconductor packaging process and bring great inconvenience to production.

[0004] On the other hand, existing water supply system control methods often ignore the complex relationship between factors such as water consumption, electricity consumption, and electricity price differences in different periods in the semiconductor packaging process. The operation of semiconductor packaging equipment has specific rules and requirements, and the water and electricity consumption are different from ordinary water supply scenarios. If intelligent regulation cannot be carried out according to these actual conditions, it will lead to serious energy waste. For example, during peak electricity consumption, if the operating frequency of the external supply pump cannot be reasonably adjusted, it will not only increase the energy consumption cost, but also may affect the water supply effect due to insufficient power supply pressure, thereby indirectly affecting the cooling of semiconductor equipment or other water-related process links; and during low-peak periods, if the low-priced electricity is not fully utilized to increase water supply reserves or perform equipment maintenance operations, it will also cause idle and waste of resources. Therefore, there is an urgent need for a more advanced, reliable and energy-saving constant pressure water supply control technology. Summary of the invention

[0005] The purpose of the present invention is to solve the problems in the background technology and provide a constant pressure water supply system for semiconductor packaging process water.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions:

[0007] An external supply pump, an external supply pressure sensor, a PLC, a host computer, a nitrogen-sealed water tank and an external supply pump frequency adjustment system, wherein the external supply pressure sensor is arranged on the water supply pipeline of the semiconductor packaging equipment;

[0008] The external supply pump frequency regulation system comprises the following steps:

[0009] Step 1: Obtain historical data of the semiconductor packaging process, record it once every 1 minute, including the pressure data P of the external pressure sensor, the total water consumption Q, the operating frequency f of the external pump, the current time t and the power consumption C, and perform data preprocessing. There are 3 external pressure sensors, and the pressure data is obtained by selecting the median;

[0010] Step 2: Using the data from step 1 to train a multivariate linear regression model;

[0011] Step 3: Calculate the external supply pump speed predicted by the multivariate linear regression model;

[0012] Step 4: Use a closed-loop PID controller to stabilize the speed of the external supply pump;

[0013] Step 5: Use the linear weighting method to set the final external supply pump speed.

[0014] Preferably, the data preprocessing in step 1 is specifically as follows:

[0015] Step 1.1: Query the historical data to obtain the minimum value P of the pressure data min , the maximum value of the pressure data P max , the minimum total water consumption Q total,min , the maximum total water consumption Q total,max , the minimum operating frequency of the external supply pump is f min , the maximum operating frequency f of the external supply pump max ;

[0016] Step 1.2: The data in step 1 is normalized using Min-Max. The normalization formula is: X min is the minimum value, X max is the maximum value, and X is the current value;

[0017] By querying the historical data of the semiconductor packaging process, the minimum and maximum values ​​of the pressure data, the minimum and maximum values ​​of the total water consumption, and the minimum and maximum frequencies of the external supply pump operation are obtained, providing a range reference for subsequent data processing. In the water environment of the semiconductor packaging process, the pressure data is obtained by taking the median of three pressure sensors, which effectively reduces the impact of a single sensor failure or abnormal data on the overall system. For example, in the operation of equipment in a semiconductor production workshop, there may be electromagnetic interference, temperature changes that affect sensor accuracy, and vibrations during the process can also cause fluctuations in sensor output. In this case, taking the median can avoid the impact of such abnormal values ​​on the overall system, thereby ensuring the relative stability and reliability of the pressure data, enabling the system to more accurately perceive the actual pressure state of the external supply pipeline, and providing a data basis for the constant pressure water supply control of the semiconductor packaging process water.

[0018] Preferably, the training of the multivariate linear regression model in step 2 is specifically as follows:

[0019] Step 2.1: The historical data pre-processed in step 1 is divided into a training set and a test set by a random number generator at a ratio of 70%:30%;

[0020] Step 2.2: Construct the multivariate linear regression model C = β0 + β1 × Q total +β2×f+β3×P+β4×t+∈, where C is the power consumption, β0 is the intercept term, β1 is the influence coefficient of the total water consumption, β2 is the influence coefficient of the external supply pump frequency, β3 is the influence coefficient of the pressure data, β4 is the influence coefficient of the time, and ∈ is the error term; train the model with the training set and calculate the residual sum of squares i is the observed value, C i is the actual power consumption of the training set shown;

[0021] Step 2.3: Take partial derivatives of RSS with respect to β0, β1, β2, β3 and β4 to obtain estimated values. The equation group is as follows:

[0022]

[0023] Step 2.4: Use the test set to evaluate the trained multivariate regression model and calculate the mean square error (MSE) and the coefficient of determination (R). 2 , if MSE < 0.3 and R 2 >0.75, the training is successful.

[0024] C i is the actual power consumption, The multivariate linear regression model predicts electricity consumption, is the average power consumption of the test set.

[0025] The pre-processed historical data is divided into training set and test set by a random number generator at a ratio of 70%:30%. This division method is scientific and random. The 70% training set has a large amount of data and can fully contain the system operation information under various process stages, providing a rich data basis for model training, so that the model can learn a more comprehensive input and output relationship pattern, such as pressure, water consumption, pump frequency, power consumption, and pump speed. The 30% test set can be used to independently verify the generalization ability of the model, ensure that the model can perform well on unseen data, avoid overfitting the model to the training data, and thus improve the accuracy and reliability of the model in practical applications;

[0026] Construct a multivariate linear regression model and calculate the residual sum of squares RSS. RSS reflects the degree of difference between the model prediction value and the actual value. For each coefficient in the model, including the total water consumption influence coefficient, the external supply pump frequency influence coefficient, and the pressure data influence coefficient, the partial derivative is calculated to obtain an estimated value. This is to use mathematical optimization methods to find the coefficient combination that minimizes RSS, thereby determining the optimal parameters of the model. In this way, the model can more accurately characterize the influence of each input factor on the output, such as power consumption, and improve the model's prediction accuracy for the actual system operating status. For example, in different stages of the semiconductor packaging process, it is possible to more accurately predict power consumption and pump speed based on changes in pressure, water consumption, and pump frequency, providing a more reliable basis for energy saving and constant pressure control;

[0027] The trained multivariate regression model was evaluated using the test set, and the mean square error (MSE) and the coefficient of determination were calculated. MSE measures the average deviation between the model's predicted value and the actual value. MSE < 0.3 indicates that the model's prediction error is small, indicating that the model's prediction results on the test set are closer to the actual situation. The coefficient of determination measures the degree of fit of the model to the data. > 0.75 indicates that the model can explain most of the changes in the data and has a high goodness of fit. By setting such a threshold to determine whether the model training is successful, it can ensure that the model put into use has high performance and reliability, and ensure that the system can be effectively and intelligently regulated according to the model's prediction results in actual operation, so as to achieve the goals of constant pressure water supply and energy-saving optimization.

[0028] Preferably, the external supply pump speed is predicted in step 3 as follows:

[0029] Step 3.1: collecting the pressure data P of the external supply pipeline, the total water consumption Q, the current time t and the operating frequency f of the external supply pump every 1 minute;

[0030] Step 3.2: normalize the data in step 3.1;

[0031] Step 3.3: input the data preprocessed in step 3.2 into the trained multivariate linear regression model;

[0032] Step 3.4: The multivariate linear model predicts electricity consumption Historical highest electricity consumption C high , and the water consumption during this period is Q total,real , the historical maximum water consumption Q hist,max If satisfied And Q total,real <0.9×Q hist,max , then reduce the operating frequency of the external supply pump to reduce energy consumption; if If the load is at a low point, the operating frequency of the external supply pump will be increased to avoid peak loads and save energy.

[0033] The pressure data P, total water consumption Q, current time t and external supply pump operation frequency f are collected every 1 minute. High-frequency data collection can timely capture changes in the operating status of the water supply system. In the semiconductor packaging process, there will be differences in water consumption at different stages in the actual water supply process. By collecting data every minute, the system can quickly perceive these changes and provide sufficient data support for subsequent real-time regulation, ensuring that the system can respond in time, maintain a constant external supply pressure, and improve water supply quality and stability;

[0034] After normalizing the collected data, it is input into the trained multivariate linear regression model to ensure the consistency of the data format between the real-time data and the training model. Since the model is trained based on normalized data, this allows the real-time data to be accurately processed in the model, improving the accuracy of the model prediction. For example, the pressure data collected in different time periods may have large numerical differences due to different fluctuation ranges of pipeline pressure. After normalization, the degree of its influence in the model can be kept consistent with that during training, allowing the model to accurately predict key parameters such as power consumption and external supply pump speed based on real-time data, providing a reliable basis for intelligent regulation;

[0035] Frequency adjustment is performed based on the comparison between the prediction results of the multivariate linear model and historical data, achieving intelligent energy saving. The system does not simply adjust according to the current pressure or water consumption, but comprehensively considers multiple factors such as historical electricity consumption, current actual water consumption, and model-predicted electricity consumption. For example, when water demand has not reached its peak but the model predicts that electricity consumption will exceed the historical maximum, the frequency is reduced in advance, and when electricity consumption is low during off-peak hours and there is room for energy saving, the frequency is increased. This dynamic adjustment enables the water supply system to minimize energy consumption, optimize the overall system performance, extend the service life of equipment, and reduce operating costs while meeting water demand.

[0036] Preferably, a closed-loop PID controller is used to stabilize the speed of the external supply pump;

[0037] In the water supply system, changes in water demand will cause pressure fluctuations in the external supply pipeline. The closed-loop PID controller can monitor pressure changes in real time and quickly adjust the speed of the external supply pump according to the set target pressure value. For example, when a large number of water-using equipment is suddenly turned on, causing the pressure to drop, the PID controller can quickly increase the speed of the external supply pump to replenish the pressure and quickly restore the external supply pressure to the set constant pressure value; conversely, when water demand decreases and pressure increases, the controller will reduce the speed of the external supply pump to keep the pressure stable.

[0038] Preferably, the step 5 is specifically the following steps:

[0039] Step 5.1: In combination with the standards of the local power company, the linear weighting method is used to assign different weight values ​​w and |1-w| to the multivariate linear regression model and PID controller of the peak period, flat period and valley period. During the peak period, the weight value w of the multivariate model ranges from 0.1 to 0.3, and the weight value |1-w| of the PID controller ranges from 0.7 to 0.9; during the flat period, the weight value w of the multivariate model ranges from 0.4 to 0.6, and the weight value |1-w| of the PID controller ranges from 0.4 to 0.6; during the valley period, the weight value w of the multivariate model ranges from 0.7 to 0.9, and the weight value |1-w| of the PID controller ranges from 0.1 to 0.3;

[0040] Step 5.2: Obtain the external supply pump frequency adjustment value Δf output by the multivariate linear regression model model The frequency adjustment value Δf of the external supply pump output by the PID controller PID ;

[0041] Step 5.3: Output the final external supply pump adjustment value Δf Pcombined =w×Δf model +(|1-w|)×Δf PID .

[0042] Using the linear weighting method to assign different weight values ​​to the multivariate linear regression model and PID controller according to different time periods can better adapt to the characteristics of water and electricity demand in different time periods; During peak hours, the multivariate weight value w is set to 0.1-0.3, and the PID controller weight value is set to 0.7-0.9. At this time, the water demand is large and the electricity price is high. The higher weight of the PID controller can ensure the stability and rapid response capability of the system, give priority to meeting the water demand and avoid instability caused by frequent adjustments. At the same time, the multivariate model can also provide certain auxiliary references to deal with complex situations; In the flat section, the weight values ​​of the multivariate model and PID controller are both 0.4-0.6. During this stage, the water and electricity consumption are relatively stable, and the weights of the two are equivalent, working together to maintain the stable operation of the system and reasonable energy consumption control; During the valley period, the weight value w of the multivariate model is 0.7-0.9, and the weight value of the PID controller is 0.1-0.3. At this time, the electricity price is low, and the multivariate model can be fully utilized to make more flexible adjustments to the system to achieve peak-shifting energy saving, such as increasing water supply reserves or performing equipment maintenance. At the same time, the PID controller can also provide stability protection when necessary;

[0043] Obtaining the frequency adjustment of the external supply pump output by the multivariate linear regression model and the frequency adjustment of the external supply pump output by the PID controller can fully consider the impact of the two control methods on the system. The multivariate linear regression model is based on a comprehensive analysis of historical data and real-time data. The adjustment amount of its output reflects the comprehensive consideration of the system in terms of energy saving and meeting water demand; the PID controller focuses on the rapid response adjustment of real-time feedback such as current pressure. The adjustment amounts of both have an important impact on the final operating state of the system. Through the final adjustment amount formula, the advantages of the multivariate linear regression model and the PID controller are combined, so that the system can achieve the best operating effect at different times. While meeting water demand, energy-saving optimization is achieved, the stability and reliability of the system are improved, and it is ensured that the water supply system can flexibly adjust the operating frequency of the external supply pump according to actual conditions to adapt to the complex and changeable water and electricity consumption environment.

[0044] In summary, the beneficial effects of the present invention are:

[0045] 1. By setting up three pressure sensors in the external supply pipeline and obtaining pressure data by taking the median, the redundant design of the sensor is realized. The system can accurately judge the health status of each sensor by observing the changes in the difference size, further improving the reliability and stability of the water supply system and reducing the occurrence of water supply interruptions due to equipment failure.

[0046] 2. Construct a multivariate linear regression model, taking into account the impact of multiple factors on electricity consumption, such as total water consumption, external supply pump operation frequency and pressure data. By learning and analyzing historical data, the model can predict electricity consumption and external supply pump speed under different working conditions; assign different weight values ​​to the multivariate linear regression model and PID controller according to different time periods, achieving energy-saving effects and reducing the operating costs of the water supply system.

[0047] 3. Model training and real-time data analysis based on a large amount of historical data enable the system to make intelligent decisions based on actual operating conditions and achieve intelligent regulation and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is the frequency diagram of the external supply pump of the present invention;

[0049] Figure 2 It is the electricity cost diagram of the present invention;

[0050] Figure 3 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0051] The following specific embodiments are merely explanations of the present invention and are not limitations of the present invention. After reading this specification, those skilled in the art may make modifications to the embodiments without any creative contribution as needed. However, such modifications are protected by the patent law as long as they are within the scope of the claims of the present invention.

[0052] The present invention will be described in detail below with reference to the accompanying drawings by way of embodiments.

[0053] Embodiment 1: A constant pressure water supply system for semiconductor packaging process water, comprising the following steps:

[0054] Step 1: Obtain historical data of the semiconductor packaging process, record it once every 1 minute, including the pressure data P of the external pressure sensor, the total water consumption Q, the operating frequency f of the external pump, the current time t and the power consumption C, and perform data preprocessing;

[0055] Step 1.1: Query the historical data to obtain the minimum value P of the pressure data min , the maximum value of the pressure data P max , the minimum total water consumption Q total,min , the maximum total water consumption Q total,max , the minimum operating frequency of the external supply pump is f min , the maximum operating frequency f of the external supply pump max , the pressure data is obtained by selecting the median through 3 pressure sensors;

[0056] Step 1.2: The data in step 1 is normalized using Min-Max. The normalization formula is: X min is the minimum value, X max is the maximum value, and X is the current value;

[0057] Step 2: Using the data from step 1 to train a multivariate linear regression model;

[0058] Step 2.1: The historical data pre-processed in step 1 is divided into a training set and a test set by a random number generator at a ratio of 70%:30%;

[0059] Step 2.2: Construct the multivariate linear regression model C = β0 + β1 × Q total +β2×f+β3×P+β4×t+∈, train the model with the training set and calculate the residual sum of squares i is the observed value, C i is the actual power consumption of the training set shown;

[0060] Step 2.3: Take partial derivatives of RSS with respect to β0, β1, β2, β3 and β4 to obtain estimated values. The equation group is as follows:

[0061]

[0062] Step 2.4: Use the test set to evaluate the trained multivariate regression model and calculate the mean square error (MSE) and the coefficient of determination (R). 2 , if MSE < 0.3 and R 2 >0.75, the training is successful.

[0063] C i is the actual power consumption, The multivariate linear regression model predicts electricity consumption, The average power consumption of the test set is:

[0064] Step 3: Calculate the external supply pump speed predicted by the multivariate linear regression model;

[0065] Step 3.1: collecting the pressure data P of the external supply pipeline, the total water consumption Q and the operating frequency f of the external supply pump every 1 minute;

[0066] Step 3.2: normalize the data in step 3.1;

[0067] Step 3.3: input the data preprocessed in step 3.2 into the trained multivariate linear regression model;

[0068] Step 3.4: The multivariate linear model predicts electricity consumption Historical highest electricity consumption C high , and the water consumption during this period is Q total,real , the historical maximum water consumption Q hist,max If satisfied And Q total,real <0.9×Q hist,max , then reduce the operating frequency of the external supply pump to reduce energy consumption; if If the load is at a low point, the operating frequency of the external supply pump will be increased to avoid peak loads and save energy.

[0069] Step 4: Use a closed-loop PID controller to stabilize the speed of the external supply pump;

[0070] Step 5: Use the linear weighting method to set the final external supply pump speed.

[0071] Step 5.1: assign different weight values ​​w and |1-w| to the multivariate linear regression model and PID controller of the peak period, flat period and valley period, respectively. During the peak period, the weight value w of the multivariate model ranges from 0.1 to 0.3, and the weight value |1-w| of the PID controller ranges from 0.7 to 0.9; during the flat period, the weight value w of the multivariate model ranges from 0.4 to 0.6, and the weight value |1-w| of the PID controller ranges from 0.4 to 0.6; during the valley period, the weight value w of the multivariate model ranges from 0.7 to 0.9, and the weight value |1-w| of the PID controller ranges from 0.1 to 0.3;

[0072] Step 5.2: Obtain the external supply pump frequency adjustment value Δf output by the multivariate linear regression model model The frequency adjustment value Δf of the external supply pump output by the PID controller PID ;

[0073] Step 5.3: Output the final external supply pump adjustment value Δf Pcombined =w×Δf model +(|1-w|)×Δf PID ,as follows Figure 1 , Figure 2 As shown:

[0074] like Figure 1 As shown, there is an obvious upward trend at the 3rd, 7th, 8th, 12th and 18th minutes, which shows that the multivariate linear regression model plays a leading role, while the downward trend is dominated by the PID shown to maintain a stable water supply. Figure 2 As shown, when the model was not used for optimization control, the monthly electricity bill was around 80,000-95,000 yuan. After the model was used in June, the overall energy consumption was reduced, saving 10,000-16,000 yuan.

Claims

1. A constant pressure water supply system for semiconductor packaging process water, characterized in that: include: An external supply pump, an external supply pressure sensor, a PLC, a host computer, a nitrogen-sealed water tank and an external supply pump frequency adjustment system, wherein the external supply pressure sensor is arranged on the water supply pipeline of the semiconductor packaging equipment; The external supply pump frequency regulation system comprises the following steps: Step 1: Obtain historical data of the semiconductor packaging process, record it once every 1 minute, including the pressure data P of the external pressure sensor, the total water consumption Q, the operating frequency f of the external pump, the current time t and the power consumption C, and perform data preprocessing; Step 2: Using the data from step 1 to train a multivariate linear regression model; Step 3: Calculate the external supply pump speed predicted by the multivariate linear regression model; Step 4: Use a closed-loop PID controller to stabilize the speed of the external supply pump; Step 5: Use the linear weighting method to set the final external supply pump speed.

2. A semiconductor packaging process water constant pressure supply system according to claim 1, characterized in that: The data preprocessing in step 1 is specifically as follows: Step 1.1: Query the historical data to obtain the minimum value P of the pressure data min , the maximum value of the pressure data P max , the minimum total water consumption Q total,min , the maximum total water consumption Q total,max , the minimum operating frequency of the external supply pump is f min , the maximum operating frequency f of the external supply pump max ; Step 1.2: The data in step 1 is normalized using Min-Max. The normalization formula is: X min is the minimum value, X max is the maximum value and X is the current value.

3. A semiconductor packaging process water constant pressure supply system according to claim 2, characterized in that: The training of the multivariate linear regression model in step 2 is specifically as follows: Step 2.1: The historical data pre-processed in step 1 is divided into a training set and a test set by a random number generator at a ratio of 70%:30%; Step 2.2: Construct the multivariate linear regression model C = β0 + β1 × Q total +β2×f+β3×P+β4×t+∈, train the model with the training set and calculate the residual sum of squares i is the observed value, C i is the actual power consumption of the training set; Step 2.3: Take partial derivatives of RSS with respect to β0, β1, β2, β3 and β4 to obtain estimated values. The equation group is as follows: Wherein, β0 is the intercept term, β1 is the influence coefficient of the total water consumption, β2 is the influence coefficient of the external supply pump frequency, β3 is the influence coefficient of the pressure data, β4 is the influence coefficient of the time, ∈ is the error term, β0 is the intercept term, β1 is the influence coefficient of the total water consumption, β2 is the influence coefficient of the external supply pump frequency, β3 is the influence coefficient of the pressure data, β4 is the influence coefficient of the time, ∈ is the error term; Step 2.4: Use the test set to evaluate the trained multivariate regression model and calculate the mean square error (MSE) and the coefficient of determination (R). 2 , if MSE < 0.3 and R 2 >0.75, the training is successful. C i is the actual power consumption of the training set, The multivariate linear regression model predicts electricity consumption, is the average power consumption of the test set.

4. A semiconductor packaging process water constant pressure supply system according to claim 3, characterized in that: The external supply pump speed predicted in step 3 is specifically: Step 3.1: collecting the pressure data P of the external supply pipeline, the total water consumption Q, the current time t and the operating frequency f of the external supply pump every 1 minute; Step 3.2: normalize the data in step 3.1; Step 3.3: input the data preprocessed in step 3.2 into the trained multivariate linear regression model; Step 3.4: The multivariate linear model predicts electricity consumption Historical highest electricity consumption C high , and the water consumption during this period is Q total,real , the historical maximum water consumption Q hist,max If satisfied And Q total,real <0.9×Q hist,max , then reduce the operating frequency of the external supply pump to reduce energy consumption; if If the load is at a low point, the operating frequency of the external supply pump will be increased to avoid peak loads and save energy.

5. A semiconductor packaging process water constant pressure supply system according to claim 4, characterized in that: The step 5 is specifically the following steps: Step 5.1: In combination with the standards of the local power company, the multivariate linear regression model and the PID controller for the peak period, the flat period and the valley period are assigned different weight values ​​w and |1-w| respectively; Step 5.2: Obtain the external supply pump frequency adjustment value Δf output by the multivariate linear regression model model The frequency adjustment value Δf of the external supply pump output by the PID controller PID ; Step 5.3: Output the final external supply pump adjustment value Δf Pcombined =w×Δf model +(|1-w|)×Δf PID .

6. A semiconductor packaging process water constant pressure supply system according to claim 5, characterized in that: The multivariate linear regression model is C = β0 + β1 × Q total +β2×f+β3×P+β4×t+∈, where C is the power consumption.

7. A semiconductor packaging process water constant pressure supply system according to claim 6, characterized in that: There are three external pressure sensors, and the pressure data is obtained by selecting the median.

8. A semiconductor packaging process water constant pressure supply system according to claim 7, characterized in that: A PID controller is used to stabilize the frequency of the external supply pump; a linear weighting method is used to assign corresponding weight values ​​to the multivariate model and the PID controller according to different time zone allocations, so as to adjust the frequency of the external supply pump.

9. A semiconductor packaging process water constant pressure supply system according to claim 8, characterized in that: In the step 5.1, during the peak period, the weight value w of the multivariate model ranges from 0.1 to 0.3, and the weight value |1-w| of the PID controller ranges from 0.7 to 0.9; during the flat period, the weight value w of the multivariate model ranges from 0.4 to 0.6, and the weight value |1-w| of the PID controller ranges from 0.4 to 0.6; during the trough period, the weight value w of the multivariate model ranges from 0.7 to 0.9, and the weight value |1-w| of the PID controller ranges from 0.1 to 0.3.

Images