Systems and methods for optimizing cross-flow membrane filtration devices

Abstract

The present application provides a system for optimizing a filtration process performed on a cross-flow membrane filtration device. The system includes an input interface configured to obtain a plurality of values and a plurality of interferences associated with the CFMF device. The system also includes a control model that estimates the plurality of controlled output values, and an optimizer that updates a state of the control model. The optimizer includes an MPC layer that obtains a target value and determines a new optimal input value for the CFMF device. An output interface causes the optimum value to be output to the CFMF device, thereby optimizing the CFMF process.

Description

Technical Field

[0001] This application relates to the automation of filtration processes such as ultrafiltration. In particular but not exclusively, this application relates to the automation of an optimized crossflow membrane filtration process. In particular but not exclusively, this application relates to a system for optimizing a filtration process performed on a crossflow membrane filtration (CFMF) device that divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. Background Art

[0002] Crossflow membrane filtration (CFMF) is widely used in chemical engineering, biochemical engineering, and protein purification to separate feed components. Depending on the pore size of the filtration material, the membrane filtration process is referred to as microfiltration (MF), ultrafiltration (UF), or nanofiltration (NF). The pore size determines the rejection of certain components in the feed and thus the composition of the final separated streams.

[0003] The goal of the CFMF process is to divide the feed stream into a concentrate stream (also referred to as a retentate stream) and a permeate stream. Controlling the CFMF process to keep the percentage value and concentration of the target component in the final product within an ideal range, such as 80.5% ± 0.5%, is typically challenging.

[0004] In the dairy industry, crossflow UF and crossflow MF or a combination of both are commonly used filtration techniques when processing whey into high-value whey protein concentrate (WPC) and whey protein isolate (WPI). Similarly, UF and MF are well-known techniques for processing skim milk into milk protein concentrate (MPC) and milk protein isolate (MPI). The UF process separates lactose and minerals (permeate) from proteins (retentate) and uses a membrane to concentrate the proteins, which rejects the proteins and is permeable to lactose, minerals, water, etc. For the MF process, the proteins will be in the permeate stream. For example, when concentrating the proteins in whey, the dry matter ratio increases as well as the total solids component. It is necessary to control the process while optimizing the operating efficiency to account for changes in feed components and other disturbances. Examples of optimizing the process include: maximizing productivity; minimizing water consumption; maintaining the quality / concentration of proteins in the outlet above a minimum; and avoiding excessive damage to the (multiple) membranes.

[0005] The production amount of the target component depends on various factors, including the concentration factor (feed stream divided by the retentate stream), the water stream factor (the water stream, also known as the direct water stream, divided by the feed stream), and the quality of the feedstock. Due to the variation of the feedstock components and problems such as membrane fouling over time, it is difficult to control the quality of the feedstock and it may change. Therefore, corrective measures must be taken to maintain the target quality. Such corrective measures are usually carried out by operators who monitor the output and change the concentration and water stream factors. For operators, this is a time-consuming task, usually requiring the operators to collect samples, analyze these samples in the laboratory and take subsequent corrective measures. This leads to operators following a "safe" operating strategy, which means that the target content must be on average significantly higher than the specified quality limit so that such variations do not cause the target content to fall below the quality limit (as Figure 2 shown). Usually, when analyzing the protein content produced by an unoptimized CFMF process, the protein content is far higher than the specification, which may lead to an increase in production costs. The latest development in sensor technology has made it possible to install on-line component sensors, thus eliminating the need for manual sampling and laboratory work. However, operators still need to take manual actions to correct the process to meet the quality specifications.

[0006] Overall, this operator-based standard correction process is time-consuming, inefficient, and may have more environmental impacts, such as a higher carbon footprint. One of the reasons for the greater environmental impact is that, to meet the quality-related requirements of the final product, the water and energy used are often higher than necessary.

[0007] Generally, a more effective control is needed to allow for rapid changes in the process to compensate for different disturbances and components of the feedstock. Summary of the Invention

[0008] This application aims to establish a system for more effectively controlling the CFMF process. The optimization of the CFMF process includes using CFMF-related data, such as data obtained from sensors installed in the components of the CFMF device, to predict these same values at future time points, thereby adjusting the input values of the process (e.g., the concentration factor and the water factor) to achieve the desired protein content.

[0009] One aspect of the present application relates to a system for optimizing a filtration process performed on a CFMF device, where the device divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. The system includes: an input interface communicatively coupled to the CFMF device, the input interface being configured to: obtain a plurality of controlled output values related to the filtration process performed on the CFMF device, and obtain a plurality of disturbances related to the CFMF device. The plurality of controlled output values includes: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of solid components in the outlet product stream. The product stream may include permeate and / or retentate streams. For example, (i) may be a retentate concentration value representing the protein concentration in the outlet retentate stream, and (ii) may be a total concentration value representing the concentration of all solid components in the outlet permeate stream.

[0010] The system further includes a control model of the CFMF device, wherein the control model estimates the controlled output values from the input values based on the state of the control model and the plurality of disturbances. The system further includes an optimizer configured to update the state of the control model based on the plurality of controlled output values. The optimizer includes a model predictive control (MPC) layer, wherein the MPC layer is configured to obtain a plurality of target controlled output values and determine a plurality of input values. The plurality of input values optimize a first difference between the plurality of target controlled output values and the plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from the control model using the plurality of input values. The system further includes an output interface communicatively coupled to the CFMF device, the output interface being configured to output the plurality of input values to the CFMF device, thereby reducing a second difference between the plurality of target controlled output values and the plurality of future controlled output values.

[0011] Advantageously, the optimization system includes a control model and an MPC layer, which allows for continuous automated monitoring and correction of the CFMF process. This can then result in a significant optimization of the concentration of the target component in the output product. For example, applying the optimization system to dairy cross-flow UF can result in a more consistent product with reduced variation in protein content over time. Thus, compared to an unoptimized human CFMF system, the average protein content is closer to the target content without affecting the quality of the final product. For example, when the target protein content is 80%, an unoptimized CFMF device for separating whey protein may produce a product with an average protein content greater than 85%, while an optimized CFMF device may average closer to 80% but not less than 80%. This results in a more efficient CFMF process, less wasted feed, water, energy, and time.

[0012] According to another aspect of the present application, there is disclosed a system for optimizing water usage on a CFMF device, which includes a plurality of membranes, each of which performs a filtration process, and the device divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. Optimizing the total water usage can be achieved by obtaining an optimal separation of water to each membrane stack, where a plurality of membrane stacks are provided. The stack can include any one of any form of filtration membrane, such as UF membrane and MF membrane, or a combination thereof. This optimal separation process allows the CFMF to better utilize water, which can also reduce the total amount of water required. In addition, by reducing the water usage, the resulting retentate has a lower water content, thereby reducing the energy required for post-treatment processes such as drying.

[0013] The system includes a membrane control model associated with the plurality of membranes, where the membrane control model is configured to estimate concentration values based on the input water flow to each of the plurality of membranes, the permeate flow, the retentate flow, and / or the retentate components per stack. Thus, the concentration value is related to the estimated concentration of the solid components in the retentate flow per stack. The system further includes an optimization unit, which is configured to obtain an optimal concentration value associated with the solid components in the retentate flow per stack. The optimization is further configured to determine the optimal input water flow to each of the plurality of membranes such that the difference between the optimal concentration value and the estimated concentration value is minimized. The estimated concentration value is determined by the membrane control model based on the optimal input water flow to each of the plurality of membranes. The system also includes an output interface communicatively coupled to the CFMF device, which is configured to cause the CFMF device to adjust the water flow at each of the plurality of membranes of the CFMF device according to the optimal input water flow determined for each of the plurality of membranes.

[0014] Advantageously, the optimization unit and the membrane control model allow the automated system to continuously monitor and respond to changes in water flow and important concentration values. This is an improvement over non-optimized systems because it can generate optimal water flow and concentration values, thereby improving the performance of the CFMF process while optimizing the membranes. The water usage may be significantly reduced, thereby improving the efficiency of the CFMF process and the environmental impact.

[0015] According to another aspect of the present application, there is disclosed a system for optimizing a filtration process performed on a CFMF device. The CFMF device includes a plurality of membranes, each of which performs a filtration process, and the device divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. The system includes an input interface communicatively coupled to the CFMF device, which is configured to obtain a plurality of controlled output values related to the filtration process performed on the CFMF device, and to obtain a plurality of disturbances related to the CFMF device. The plurality of controlled output values can include: (i) a product concentration value indicating the concentration of the target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of components in the outlet product stream.

[0016] The product stream can include a permeate stream and / or a retentate stream. For example, (i) can be a retentate concentration value representing the protein concentration in the outlet retentate stream, and (ii) can be a total concentration value representing the concentration of all solid components in the outlet retentate stream. The system further includes a device control model associated with the CFMF device, where the device control model estimates a controlled output value from input values based on the state of the device control model and a plurality of disturbances, and further includes a plurality of membrane-related membrane control models, where the membrane models are configured to estimate concentration values based on at least one input water flow to each of the plurality of membranes. The concentration value is related to the estimated concentration of the solid components in the retentate stream.

[0017] The system further includes an optimizer configured to update the state of the device control model based on a plurality of controlled output values. The optimizer further includes an MPC layer and a water flow optimization (WFO) layer. The MPC layer is configured to obtain a plurality of target controlled output values and determine a plurality of input values that optimize a first difference between the plurality of target controlled output values and a plurality of estimated controlled output values. The plurality of estimated controlled output values are determined from the control model by the plurality of input values. The WFO layer is configured to: obtain an optimal concentration value related to the solid components in the outlet retentate stream; determine an optimal water factor related to the input water flow to each of the plurality of membranes to minimize the difference between the optimal concentration value and the estimated concentration value, where the estimated concentration value is determined from the membrane control model based on the optimal input water flow to each of the plurality of membranes; and adjust the plurality of input values determined by the MPC layer based on the optimal input water flow to each of the plurality of membranes. The system further includes an output interface communicatively coupled to the CFMF device, the output interface being configured to output the plurality of input values to the CFMF device. This reduces a second difference between the plurality of target controlled output values and a plurality of future controlled output values.

[0018] Advantageously, the optimization system includes: a control model and an MPC layer to allow continuous automation for monitoring and correcting the CFMF process, an optimization unit and membrane control models to allow the automated system to monitor and respond to changes in water flow and other factors, such as changes in feed components. This can significantly optimize the concentration of the target components in the output product, while also saving water, optimizing the service life of the membranes, and preventing early fouling. This greatly improves the efficiency of the CFMF process, with less wasted feed, water, energy, and time, while reducing the environmental impact of the CFMF process.

[0019] Additional aspects and embodiments of the system are disclosed, and the above aspects and embodiments should not be construed as limiting the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To facilitate a more complete understanding of the present application, reference is now made to the accompanying drawings, in which like numerals represent like elements. These drawings should not be construed as limiting the disclosure, but are merely illustrative.

[0021] Figure 1 Examples of reverse osmosis membranes, UF membranes, and MF membranes are shown.

[0022] Figure 2 An example of an optimized CFMF system is shown.

[0023] Figure 3 A representative trajectory of the target component concentration over time is shown, where the CFMF process is traditional (non-optimized), stable, and then optimized in different time zones.

[0024] Figure 4 A high-level system architecture diagram for optimizing the filtration process performed on a CFMF device according to the present disclosure is shown.

[0025] Figure 5 A high-level system architecture diagram for optimizing the water consumption of a CFMF device according to the present disclosure is shown.

[0026] Figure 6 A high-level system architecture diagram for optimizing the filtration process and water consumption associated with a CFMF device according to the present disclosure is shown.

[0027] Figure 7 A flowchart of a method for optimizing the filtration process performed on a CFMF device is shown.

[0028] Figure 8 A flowchart of a method for optimizing the water consumption of a CFMF device is shown.

[0029] Figure 9A and 9B A flowchart of a method for optimizing the filtration process and water consumption of a CFMF device is shown.

[0030] Figure 10 An example computing system optimized according to an embodiment of the present disclosure is shown. Detailed Description

[0031] The following description discloses an automatic control algorithm that is configured to run at the start of a crossflow membrane filtration process. The algorithm is further configured to continuously (e.g., at 30-second time intervals) acquire numerical values according to the CFMF process. The algorithm includes a model predictive control (MPC) algorithm that is used to calculate a new set of optimal values to compensate for any disturbances in the CFMF process, such as the composition of the feedstock. The purpose of the algorithm is to reduce the variation in the target variables (such as proteins) and the total solid concentration of the concentrate produced.

[0032] Figure 1 Three example types of CFMF processes are shown, where the feedstock 100 moves Figure 1 from left to right.

[0033] Figure 1 Three membranes are shown, including a reverse osmosis membrane 110, a UF membrane 120, and an MF membrane 130. Figure 1 Component flows are further shown, including: an outlet retentate flow 102, an outlet permeate flow 104, and water 106 associated with each membrane. Figure 1 Also included are: salts, sugars, and low molecular width compounds 112, fats and proteins 114, larger molecules or microorganisms 116, lactose and minerals 132, casein and whey proteins 134, and bacterial cells 136.

[0034] The feed 100 encounters the reverse osmosis membrane 110, the UF membrane 120, or the MF membrane 130. The feed 100 is divided into an outlet retentate flow 102 and an outlet permeate flow 104 by a cross-flow membrane filtration process. The outlet retentate flow 102 includes the components that have not permeated the membrane. The outlet permeate flow 104 includes the components that have permeated the membrane.

[0035] Water 106 can permeate the reverse osmosis membrane 110. However, salts, sugars, and low molecular width compounds 112, fats and proteins 114, and larger molecules or microorganisms 116 cannot permeate the reverse osmosis membrane 110. Thus, the outlet permeate flow 104 can include only water 106. The outlet retentate flow 102 can include any components of the feed 100.

[0036] Water 106 can permeate the UF membrane 120, and so can salts, sugars, and low molecular width compounds 112. However, fats and proteins 114 and larger molecules or microorganisms 116 cannot permeate the UF membrane 120. Thus, the outlet permeate flow 104 can include water 106 and any one of salts, sugars, and low molecular width compounds 112.

[0037] Water 106 can permeate the MF membrane 130, and so can salts, sugars, and low molecular width compounds 112 and fats and proteins 114. However, larger molecules or microorganisms 116 cannot permeate the MF membrane 130. Thus, the outlet permeate flow 104 can include water 106, any one of salts, sugars, and low molecular width compounds 112, and fats and proteins 114. By adjusting the pore size of the MF membrane, fats and proteins 114 can be separated from each other, with fats not permeating the MF membrane 130 while proteins permeate the membrane 130 together with water 106 and salts, sugars, and low molecular width compounds 112.

[0038] In one example, the feedstock 100 includes dairy products, such as skim milk. Thus, the feedstock 100 includes the following components: water 106, lactose and minerals 132, casein and whey proteins 134, and bacterial cells 136. The water 106 can permeate the microfiltration membrane 130, and the lactose and minerals 132 and the casein and whey proteins 134 can also permeate the microfiltration membrane 130. The bacterial cells 136 cannot permeate the ultrafiltration membrane. Thus, the outlet permeate stream 104 can include any of the water 106, lactose and minerals 132, and casein and whey proteins 134. This process will be described in more detail below.

[0039] The MF membrane 130, UF membrane 120, reverse osmosis membrane 11, and / or NF membrane (not described) can be used alone or in combination. They can also be part of a membrane stack or used in combination with other filtration techniques or processes, such as ion exchange. Each membrane can be customized to a specific porosity, and the membrane stack can be customized for a specific process and have a specific porosity range. In processing whey into high-value whey protein concentrate (WPC), the UF process is commonly used to treat whey, while for processing high-value whey protein isolate (WPI), a pure, almost fat-free WPC, UF and MF are used in combination. The purpose of MF is to further separate fat from the protein. Similarly, UF is used to extract milk protein concentrate (MPC) from skim milk, and a combination of UF and MF is used to extract milk protein isolate (MPI) from skim milk. Thus, examples include: obtaining WPC from whey using UF; extracting WPI from whey using a combined process of UF, MF, and UF; extracting MPC from skim milk using UF; extracting MPI from skim milk using a combined process of UF and MF; separating casein and whey from skim milk using MF; and extracting lactoferrin from whey using ion exchange and nanofiltration processes.

[0040] More specifically, pasteurized skim milk can be used with MF / diafiltration to form a retentate of casein, residual whey proteins, lactose, and minerals. Micellar casein can be obtained from this retentate. The permeate can contain whey proteins, lactose, and minerals. The permeate can be treated with UF / diafiltration to form a further retentate of whey proteins, residual lactose, and minerals, which can be used to obtain whey proteins. The final permeate can contain lactose and minerals.

[0041] In another example, the feedstock 100 may include a plant protein solution, where the protein is derived from feedstocks such as oats, rice, soybeans, nuts, peas, or legumes. In this example, water 106 may permeate through the UF membrane 120, and plant sugars and minerals (such as salts, sugars, and low molecular weight compounds 112) may also permeate through the UF membrane 120. However, plant fats and proteins (such as fats and proteins 114) cannot permeate through the UF membrane. If a protein concentrate with higher purity and almost no fat is to be produced, a further MF stage will be used in combination with the UF membrane to reduce the fat in the protein concentrate.

[0042] In another embodiment, the feedstock 100 may include an animal protein solution, where the protein is derived from, for example, fish or other animals such as chickens, pigs, cows, etc.

[0043] As described below, the present application can be applied to any one of these processes, or to any other application for purification by filtration and diafiltration.

[0044] Figure 2 An optimized CFMF system 200 according to the present application is shown.

[0045] Specifically, Figure 2 A system architecture diagram of the optimized CFMF device is shown. The CFMF system 200 is an optimized CFMF device, where the CFMF system is configured to reduce the number of consumables required to obtain a final product with the set target quality. In addition, the CFMF system 200 is configured to ensure that the final product is as close as possible to the set target quality without falling below the set target quality.

[0046] The CFMF system 200 includes a feedstock 202, a water stream 204, and a pump 206. The CFMF system 200 also includes a sensor 210, a backpressure valve 220, one or more membrane stacks 230, and a sensor 240. The CFMF system 200 also includes an input feedstock 212, an input water stream 214, an input pump 216, a retentate stream 232, an outlet retentate 218, a permeate stream 238, a pre-valve outlet retentate 224, a post-valve outlet retentate 226, a flow direction 228, and an outlet permeate 242. The various components of the CFMF system 200 referred to above Figure 2 are known commercially available components, and their functions do not require further description.

[0047] The CFMF system 200 also includes an optimization control unit 244, CFMF-related output values 246, and control unit values 248. The optimization control unit 244 may also include a device optimization unit 250 and / or a water optimization unit 252.

[0048] As shown in the input feedstock 212, a portion of the feedstock 202 (such as Figure 1The raw material 100) in flows towards the membrane stack 230. The raw material 202 can be pumped by the pump 206, and the input raw material 212 can be pumped by the input pump 216. The components and flow of the raw material 202 can be monitored by the sensor 210, such as a flow sensor and / or a component sensor, such as a hall-effect sensor, an optical-based sensor, or a sonar-based sensor. As shown in the input water flow 214, a part of the water flow 204 (e.g., Figure 1 the water 106) in flows towards the membrane stack 230. The membrane stack 230 includes at least one membrane (e.g., Figure 1 the UF membrane 120 and / or the MF membrane 130) in. For example, fats and proteins do not permeate the membrane stack 230 and combine to form the retentate flow 232 (e.g., Figure 1 the outlet retentate 102) in. Water, salts, sugars, and low-molecular-width compounds permeate the membrane stack 230 and combine to form the outlet permeate flow 242 (e.g., Figure 1 the outlet permeate 104). The components and flow of the outlet permeate flow 242 can be monitored by the sensor 240 and / or the sensor 210. The sensor 240 / sensor 210 can be installed anywhere in the CFMF system 200 and can be configured to measure the components and / or flow of the relevant components (e.g., measuring the percentage of protein or total solids in the product and the retentate flow 232 in the raw material 202). The retentate flow 232 can form the input raw material 212, such that there is a circulating flow direction 228 between the outlet retentate 218 and the membrane stack 230.

[0049] The backpressure valve 220 is configured to reduce the outlet retentate 224 before the valve to a defined value by the concentration factor (CF). The CF is the ratio of the outlet retentate 224 before the valve to the raw material flow 202. Then, the outlet retentate 224 before the valve becomes the outlet retentate 226 after the valve, also known as the outlet concentrate flow. For example, the backpressure valve 220 can be configured such that the outlet retentate 224 before the valve flowing out of the CFMF system and the part of the outlet retentate 226 after the valve are 50% of the total raw material 202, so the total permeate flow and the outlet permeate flow 242 are 50% of the raw material 202 plus the water flow 204 flowing through the membrane stack 230. There may also be further configurations such that the outlet retentate 226 after the valve can include any percentage between 0% and 100% of the raw material flow 202. Optionally, the backpressure valve 220 is a plurality of backpressure valves.

[0050] The sensor 240 can be installed anywhere in the retentate flow 232, the outlet retentate 224 before the valve, the outlet retentate 226 after the valve, or the outlet permeate flow 242 of the last stack or any previous stack. The sensor 240 is configured to measure the components of the relevant components at a specific point in the retentate flow 232 (e.g., measuring the percentage of protein, lactose, fat, ash, and / or total solids in the product).

[0051] The optimization control unit 244 may include a device optimization unit 250 and / or a water optimization unit 252. The device optimization unit 250 may optimize the CFMF system 200m, including the water flow 204, while the water optimization unit 252 may optimize the input water volume 214, thereby optimizing the relative water supply 230 to each membrane stack. Therefore, the device optimization unit 250 and the water optimization unit 252 may be used in combination or separately.

[0052] The optimization control unit 244 is configured to obtain CFMF-related output values 246, where the CFMF-related output values 246 are a plurality of values related to the CFMF system 200. For example, the values may include controlled output values, such as any one of the following values: a retention concentration value, a permeate concentration value, and a total concentration value, where the retention concentration value indicates the concentration of the target component of the feedstock 202 in the post-valve retention flow 226, the permeate concentration value indicates the concentration of the target component in the outlet permeate flow 242, and the total concentration value indicates the concentration of the solid component in the post-valve retention flow 226. The values may also include interference values, which are measured by the sensor 210 and / or the sensor 240, and may include any one of air temperature, feedstock temperature, or feedstock solid concentration. The values related to concentration may be based on a plurality of filter specifications related to the membrane stack 230, where each of the plurality of filter specifications includes the water permeability, solute concentration, solute permeability, reflection factor, and membrane area available for calculating pressure (e.g., hydrodynamic pressure difference and / or osmotic pressure difference) of the associated membrane. The CFMF-related output values 246 may be obtained from the sensor 210, the sensor 240, or Figure 2 sensors not described herein, such as soft sensors, which will be described in more detail in Figure 4 below.

[0053] The optimization control unit 244 is further configured to output control unit values 248. The control unit values 248 may include the optimal input values of the CFMF system 200, where the optimal input values of the CFMF system 200 are the input values of the CFMF system 200 that will improve the efficiency of the CFMF system 200. For example, the CFMF system 200 may use the control unit values 248 to cause a change in the configuration of the backpressure valve 220 or a change in the water flow 204 relative to the feedstock 202 being input to the membrane stack 230. Advantageously, the optimization control unit 244 will thus adjust the levels of the input water 214 and the outlet product flow to the target quality in the resulting product. This may reduce the energy required to obtain a free-flowing powder in a subsequent process, as less water needs to be evaporated. The function of the optimization control unit 244 will be described in detail below with reference to Figure 6 Specifically, the optimization control unit 244 may be the Figure 6 system 600. An example of the impact of the optimization control unit 244 is shown in the Figure 3 graphic 300.

[0054] The optimization control unit 244 may further include at least one of a device optimization unit 250 or a water optimization unit 252. In one example, the optimization control unit 244 includes a device optimization unit 250 and a water optimization unit 252. The device optimization unit 250 is configured to predict these same values at a future time point using the CFMF-related output value 246, thereby adjusting the input value of the process, the control unit value 248, to achieve a desired target component content. This can result in a rapid, automated, and efficient adjustment of CFMF-related variables such as backpressure value configuration. The following refers to Figure 4 a detailed description of the function of the device optimization unit 250. Specifically, the device optimization unit 250 may be Figure 4 the optimization system 400 in. The water optimization unit 252 is configured to determine the optimal input water flow for each membrane stack 230, which minimizes the difference between the target concentration value (such as the optimal concentration value) and the predicted concentration value at a future time point, and subsequently outputs the optimal input water flow included in the control unit value 248. This can achieve a rapid, automatic, and efficient adjustment of, for example, the water flow. The following refers to Figure 5 a detailed description of the function of the water optimization unit 252. Specifically, the water optimization unit 252 may be Figure 5 the water optimization system 500 in.

[0055] Preferably, the CFMF system 200 may include a plurality of membrane stacks 230 (not described herein). The plurality of membrane stacks 230 may be configured in series or in parallel. For example, the feedstock 202 may flow through a plurality of input feedstocks 212 to the plurality of membrane stacks 230, or the feedstock 202 may flow through a single input feedstock 212 to the plurality of membrane stacks 230. The plurality of membrane stacks 230 may be configured in a combination of parallel and series. In addition, the CFMF system 200 may include a plurality of filtration stages, wherein the feedstock 202 is circulated through at least one of the plurality of membrane stacks 230. In one stage, the feedstock 202 and the water flow 204 flowing out of the CFMF system 200 form at least a part of the feedstock 202 and the water flow 204 flowing to the membrane stack, so that the CFMF process is repeated. The outlet permeate flow 242 may be a plurality of outlet permeate flows from a plurality of membrane stacks (not described herein). The outlet permeate flow 242 may become the feedstock 202 of another CFMF process, and the post-valve outlet retentate flow 226 may become the feedstock 202 of another CFMF process.

[0056] In another example, the CFMF system 200 may further include: one or more pumps, for example, the pump may pump the raw material along the direction of the raw material 202 and / or the input raw material 212; one or more staged coolers, for example, the staged cooler may cool the retained stream 232; one or more heaters, for example, the heater may heat the raw material 202; and a raw material tank, for example, the raw material 202 may flow out from the raw material tank. In another example, any one of the pre-valve outlet retained stream 224, the post-valve retained stream 226, or the outlet permeate stream 242 may flow into the raw material tank.

[0057] Optionally, the raw material 202 and the water stream 204 are combined, and / or the input raw material 212 and the input water stream 214 are combined. The sensor 210 may be combined with the sensor 240, and the pump 206 may be combined with the input pump 216. Alternatively, there may be multiple sensors 210, sensors 240, pumps 206, and input pumps 216.

[0058] One example usage of the CFMF system 200 is to process whey into high-value WPC, where the raw material 202 includes dairy whey and the target component includes whey protein. Another example is that the CFMF system 200 is used to produce milk protein concentrate, where the raw material 202 includes dairy products, such as skim milk, and the target component includes milk protein. Other examples of target components may include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin. Alternatively, the raw material 202 may include a plant protein solution, where the protein is from raw materials such as oats, rice, soybeans, nuts, peas, or legumes. The subsequent target components may include the relevant proteins or fats.

[0059] Alternatively, the raw material 202 may include an animal protein solution, where the protein is from, for example, fish or other animals, such as chickens, pigs, cows, etc. The subsequent target components may include the relevant proteins or fats.

[0060] Figure 3 A graph 300 showing the change in the concentration of the target component over time is shown.

[0061] The graph 300 includes three time periods, including the traditional CFMF process period 302, the stable CFMF process period 304, and the optimized CFMF process period 306. They are separated by two dotted lines, the first timeline 308 and the second timeline 310. The graph 300 also includes the optimal target line 312, the target component concentration trajectory 314, and the average target component concentration line 318.

[0062] The time when the conventional CFMF process period 302 changes to the stable CFMF process period 304 is shown by a first time line 308. The time when the stable CFMF process period 304 changes to the optimized CFMF process period 306 is shown by a second time line 310.

[0063] The range of the concentration trajectory 314 of the target component before the first timeline 308 is greater than that after the first timeline 308 .

[0064] The optimal target line 312 shows the optimal target component concentration required for the CFMF process. The average target component concentration line 318 shows the average value of the target component concentration trajectory over a time zone.

[0065] In the graph 300, the average target component concentration line 318 is the same during the conventional CFMF process period 302 and the stabilized CFMF process period 304, indicating that the stabilized CFMF process 304 does not necessarily bring the target component closer to the optimal amount. This also illustrates that conventional methods (such as using an operator to make adjustments) and prior art stabilization methods (such as using sensor technology) may result in a "safe" strategy, where the average target concentration is significantly higher than the specified mass limit, resulting in increased required water and energy levels. Conversely, during the optimized CFMF process period 306, the output is closer to the optimal target line 312. In addition, optimizing the CFMF process period 306 can produce further efficiency benefits. For example, optimizing the CFMF process period 306 can result in a reduction in the level of required water, and this improvement in water optimization can result in a reduction in required water evaporation, thereby reducing the required energy.

[0066] The optimized CFMF process period 306 may be as follows Figure 4 , Figure 5 or Figure 6 The time period during which the optimization system is effective, or Figure 2 Optimization control unit 244. In the scenario shown in graph 300, in the conventional CFMF process period 302 and the stable CFMF process period 304, the average target component concentration line 318 is substantially higher than the optimal target line 312. Therefore, these time periods are less efficient than during the optimized CFMF process period 306, where the average target component concentration line 318 is closer to the optimal target line 312 and the optimal component concentration, indicating that Figure 2 , Figure 4 , Figure 5 or Figure 6 How to improve efficiency of the optimization system.

[0067] In addition, you can use Figure 2 The CFMF system 200 performs Figure 3the CFMF process in. For example, the target component concentration is Figure 2 the concentration of whey protein detected by sensor 240 in, and the target component concentration trajectory 314 is Figure 2 the concentration of whey protein obtained from the outlet permeate stream 242 in, and the optimization process is performed using the optimization control unit 244.

[0068] One aspect of the present application relates to Figure 4 an optimization system.

[0069] Figure 4 An optimization system 400 and a CFMF device 402 are shown. The optimization system 400 includes an interface unit 404 and a control unit 406. The interface unit 404 further includes an input interface 408, a connection 410, an output interface 412, and a connection 414. The input interface 408 may further include interfaces for inputting a disturbance value 416 and a controlled output value 418. The control unit 406 further includes a control model 420 and an optimizer 422. The optimizer 422 further includes a model predictive control (MPC) layer 424. The optimization system 400 further includes a plurality of disturbance values 426, a plurality of controlled output values 428, the state of the control model 430, a target controlled output value 432, and an optimal input value 434.

[0070] The optimization system 400 is a system for optimizing a filtration process performed on a CFMF device 402, and the CFMF device 402 divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. The optimization system 400 for the CFMF device 402 includes an input interface 408 communicatively coupled to the CFMF device 402. The input interface 408 is configured to obtain a plurality of controlled output values 418 related to the filtration process performed on the CFMF device, wherein the plurality of controlled output values include: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of solid components in the outlet product stream. The product stream may include a permeate stream and / or a retentate stream. For example, (i) may be a retentate concentration value representing the protein concentration in the outlet retentate stream, and (ii) may be a total concentration value representing the concentration of all solid components in the outlet permeate stream. The input interface 408 is further configured to obtain a plurality of disturbance values 416 related to the CFMF device 402.

[0071] The controlled output value 418 is a value associated with the CFMF device 402 at a given time, which can be controlled through manual or automatic intervention in the CFMF process. For example, the controlled output value 418 is a value associated with the components of the outlet product stream (retentate stream or permeate stream). The controlled output value 418 can include any of the following: a product concentration value indicating the target component concentration of the raw material in the outlet product stream, and a total concentration value indicating the concentration of multiple solid components in the outlet product stream. The raw material can be the raw material 202, and the outlet product stream can be Figure 2 the retentate stream 232 or the post-valve outlet retentate stream 226 in Figure 2 the permeate stream 238 or the outlet permeate stream 242 of

[0072] The disturbance value 416 is generally a value associated with the CFMF device 402 at a given time that is not easily controlled. For example, the disturbance value 416 can be a value associated with the conditions at any point inside or around the CFMF device 402, or a value associated with the quality of the consumables entering the CFMF device 402 as part of the CFMF process. For example, the disturbance value 416 can include any of air temperature, raw material temperature, or raw material solid concentration. Other suitable disturbance values 416 can be identified and used according to the operating conditions of the system.

[0073] The optimization system 400 for the CFMF device 402 further includes a control model 420 of the CFMF device 402, where the control model 420 estimates the controlled output value from a plurality of controlled output values 428 and a plurality of input disturbance values. The state of the control model 430 is an augmented version of the CFMF device 402 at a given time. For example, the augmented version of the CFMF device 402 can be determined by the controlled output values 428 and the plurality of input disturbance values. The controlled output value estimated according to the state of the control model 430 is a subsequent enhanced controlled output value, which corresponds to the augmented version of the CFMF device at a given time point.

[0074] The optimization system 400 for the CFMF device 402 further includes an optimizer 422, which includes a model predictive control (MPC) layer 424, where the optimizer 422 is configured to update the state of the control model 430 according to a plurality of controlled output values 428. The MPC layer 424 is configured to: obtain a plurality of target controlled output values 432; and determine a plurality of input values 434 that optimize a first difference between the plurality of target controlled output values 432 and the plurality of controlled output values estimated according to the state of the control model 430, where the plurality of controlled output values estimated according to the state of the control model 430 are determined from the control model 420 using the plurality of controlled output values 428.

[0075] Updating the state of the control model 430 is a multi-step process (described in more detail below) and includes predicting a future state based on a previous state. This results in a predicted controlled output value for the next (predicted) state of the CFMF device 402. Updating the state of the control model 430 may also include correcting a past state of the control model. For example, a controlled output value estimated based on a past state of the control model may be compared to a plurality of controlled output values 428 corresponding to the past state. The state of the control model 430 may subsequently be adjusted based on the plurality of controlled output values 428 obtained by the CFMF device 402.

[0076] The target controlled output value 432 is the controlled output value corresponding to the optimal state of the CFMF system. For example, the optimal conditions are the conditions produced by an optimal profit function, which will result in the most efficient state of the CFMF system based on various constraints. Such constraints are customized for the CFMF device 402 and represent the limitations of the CFMF device and process, such as the maximum water flow into the CFMF device 402, the maximum permeate flow through the associated membrane, and the safe operating limits of the components of the CFMF device 402. For example, the constraints may include minimum and maximum concentration factors, minimum and maximum water flows, minimum and maximum target component concentrations (such as proteins), and minimum and maximum total solid concentrations. The plurality of input values 434 are the values subsequently input to the CFMF device 402 from the optimization system 400, which will cause the process of the CFMF device 402 to be adjusted to the optimal state. Thus, the plurality of input values 434 are input values for the CFMF device 402 that help minimize the difference between the controlled output value of the optimal state of the CFMF device 402 and the (predicted) controlled output value of the next state. In other words, the plurality of input values 434 are input values that help minimize the difference between the plurality of target controlled output values 432 and the plurality of estimated controlled output values based on the state of the control model 430.

[0077] The output interface 412 is communicatively coupled to the CFMF device 402 and is configured to output the plurality of input values 434 to the CFMF device 402, thereby reducing a second difference between the plurality of target controlled output values 432 and future plurality of controlled output values. The plurality of input values 434 may include a concentration factor and a water factor, where the concentration factor includes a first ratio of the feed stream to the outlet retentate stream, and the water factor includes a second ratio of the water flow to the feed stream. The output interface 412 is then configured to output the concentration factor and / or the water factor to the CFMF device 402.

[0078] The interface unit 404 includes an input interface 408 and an output interface 412. The input interface 408 is communicatively coupled to the CFMF device 402 via a connection 410, and the output interface 412 is communicatively coupled to the CFMF device 402 via a connection 414. For example, the interface unit 404 can be communicatively coupled to the CFMF device 402 via a wired connection, an Ethernet connection, or a wireless network connection.

[0079] The optimization system 400 can continue or otherwise repeat the optimization process of the difference between the target controlled output value 432 and the estimated controlled output value according to the state of the control model 430, wherein the target controlled output value 432 and the estimated controlled output value according to the state of the control model 430 are based on at least a plurality of controlled output values 428, and wherein the controlled output value 418 can be a future controlled output value.

[0080] The optimization system 400 can perform an iterative process that repeats at a fixed interval (e.g., 20 or 30 seconds) until a condition (e.g., an exit condition related to the best target or the shutdown of the optimization system 400 or the CFMF device 402) is met. For example, the optimization system 400 iteratively determines a new set of optimal input values at a fixed rate until the system 400 is terminated by an exit command, which may be a UF unit that stops production. Additionally, the control unit 406 can iteratively determine the optimal value, which can terminate at a certain optimality criterion.

[0081] During the iterative process, future controlled output values are obtained through the system 400 using the connection 410. For example, the future controlled output value can become the controlled output value 418. As Figure 3 shown, when the variables in the CFMF device 402 change over time, the iterative process allows for continuous optimization adjustments and may result in a continuous reduction in the difference between the target controlled output value 432 and the future controlled output value, even if there are fluctuations and variabilities in the plurality of controlled outputs 428 and the plurality of disturbance values 426. These fluctuations include changes in raw material component concentration, membrane permeability, and flow rate.

[0082] The control model 420 is configured to obtain a plurality of disturbance values 426 and a plurality of controlled output values 428 from the input interface 408, and is also configured to estimate the controlled output value according to the state of the control model 430, as described above. The control model 420 can be a physics-based, statistics-based, machine learning-based, or other artificial intelligence (AI)-based model. In this example, the control model 420 includes a state space model, where the state space model is a discrete-time model. For example, the control model 420 is a linear control model, and the state of the control model 420 is updated using a time-varying Kalman filter. Additionally, the control model 420 can be any other suitable model.

[0083] Since the control model 420 of this example is a state - space model, which is a discrete - time model, the subscripts k, k - 1, and k + 1 are used to represent the current state, the previous state, and the next concurrent state respectively. In addition, augmented variables use bar notation, such as the variables generated by the control model 420 and the optimizer 422. For example, corresponds to the current state vector x k of the augmented current state vector. Estimated variables use hat notation, such as as the estimated current state vector.

[0084] To obtain an estimated controlled output value based on the state of the control model 430 (where the state is a future state), the control model 420 is updated according to the current state of the control model. When using a linear control model, the updated state is a linearly adjusted current state (combined with other linearly adjusted variables obtained by the control model 420). The linear adjustment of the current state is achieved by multiplying the current state by a predetermined variable to implement. is a state - space matrix specific to the CFMF device 402, which includes constants for adjusting state parameters within the discrete - time interval between the current state and the next state . For example, multiple controlled outputs 428 vary over time, so the current controlled output is multiplied by to facilitate the formation of a predicted future controlled output. Therefore, can be considered as a linearly adjusted current state, as well as a prediction of the next state. However, to improve this prediction, the effects of other variables are considered, including: the input (u k ) of the CFMF device 402, such as multiple input values 434; the disturbance value (d k ), such as multiple disturbance values 426; the noise value The augmented noise value can account for any disturbances in the CFMF process that are not included in the multiple disturbance values 426, such as process variations that cannot be accurately measured. Each variable is linearly adjusted through a separate state - space matrix customized for the variable and the CFMF device. In addition, constants related to the linearization of the control model 420 are included, which include several constants related to the linearization of the non - linear system equations. In summary, these variables help to compensate for the linearization of the non - linear system, thereby improving the prediction accuracy. Customizing the state - space matrix and determining the linearization constants can be achieved through standard methods such as mathematical approximation, parameter tuning experiments, and trial - and - error.

[0085] Therefore, an example of the control model 420 is

[0086] where, is the augmented state vector that controls the next (k+1) state of the model. For example, the controlled output value estimated according to the state of the control model 430, where the current (k) state is is the input, such as multiple input values 434; is the measured disturbance value, such as the disturbance value 426; is the measurement vector, such as the controlled output value 28; is the current state of the controlled output; the augmented noise is the measurement noise is v k ~N iid (0, R v ); is the noise ratio state matrix. The noise variance matrix and R v are estimated by the maximum likelihood (ML) method or any other suitable method.

[0087] The augmented state space matrix is

[0088] where A, B, and E are variables related to the CFMF device 402. σ y and σ z include constants related to the linearization of the control model. For example and The augmented state vector is where, σ z and C z are formed by the row selection of C y and σ y .

[0089] The above process requires the current state of the model The control model 420 can also be configured to estimate the state of multiple estimation models based on multiple controlled output values 428 (y k ). The linear time-varying (LTV) Kalman filter can be used to estimate the state of the model Beneficially, the time variance of the Kalman filter enables the control model 420 to handle changes in the magnitude of y k due to any missing measurements. This scenario will be described in more detail below.

[0090] For example, as described above, after the control model 420 estimates the controlled output value according to the state of the control model 430 by using the previous state (k-1), the optimizer 422 is configured to optimize according to multiple controlled output values 428 (y ). k)Update the state of the control model. This involves calculating the filtered state, where multiple controlled output values 428 (y k ) are used to improve the state of the control model 430 to produce an updated (improved) current state Take this updated state and combine it with multiple input values 434 (u representing the optimal input values k ), the control model 420 can predict a new state through the aforementioned method

[0091] The part of calculating the filtered state can include using the state covariance matrix of the augmented system where the covariance is a measure of the joint variability of the state variables, evaluating the degree to which the variables of the partitioned system co-vary. The noise covariance matrix is estimated according to the covariance of the noise related to the CFMF device 402, R e and calculated iteratively by the Kalman filter. The state covariance of the augmented system improves the computational accuracy of the updated current state

[0092] The online calculation of the control unit 406 according to the above disclosure is as follows:

[0093] Required: y k , d k , u k-1

[0094] Filter: Calculate the one-step-ahead measurement prediction

[0095] Calculate the filtered state

[0096] Obtain (1):

[0097] Obtain (2):

[0098] Predicted value: Calculate the one-step-ahead state Via:

[0099] Return: u k ,

[0100] The filtering step corrects the estimated current state y according to the past state using the latest measurement value k ​。The filtering step can also handle missing observations by constructing attributes related to the measurements to provide In summary, this results in a corrected current state The prediction step can use the control model 420 to predict and the prediction step is based on the corrected current state as well as the obtained target value r k and the optimal input u k to predict the future state The example steps for obtaining (1) and the target controlled output value 432r k and the example steps for obtaining (2) and the multiple input values 434u k are as follows.

[0101] The MPC layer 424 is configured to obtain (1) and the multiple target controlled output values 432(r k ), and determine (2) and the multiple input values 434(u k ), which helps to optimize the future CFMF process as described above. The multiple input values 434(u k ) optimize the difference between the multiple target controlled output values 432(r k ) and the multiple controlled output values related to the estimated updated state . This can be achieved by standard mathematical or computational methods, including absolute difference, mean square error, or more complex methods (such as minimizing the formula of the difference while also including the impact of constraints). Those skilled in the art will understand that the choice of the type of equation used may depend on factors related to the system usage.

[0102] Constraints are system and process limitations related to the CFMF system 402 as described above. Constraints on the minimum and maximum safe operating ranges of the input values of the CFMF device 402, such as minimum and maximum concentration factor and water factor constraints, can be expressed as u min and u max . Time-related constraints can also be incorporated, for example, by constraining the system to a specific state, such as There may also be constraints related to the control model 420, as well as limitations related to the selected configuration. For example, a linear control model has limitations related to the linearization of a non-linear system, where input values beyond the limitations will result in prediction results with lower than acceptable accuracy. Therefore, complex formulas can be formed to minimize the difference between the multiple target controlled output values 432(r k ) and the multiple controlled output values related to the estimated updated state , but any potential input values that break the constraints will be penalized.

[0103] For example, to determine (2), the MPC layer 424 can be formed as an output trajectory problem in the form of a convex quadratic problem (QP).

[0104] Including constraints (a) (b) (c) (d) u min ≤ u k+j ≤ u max , j ∈ N u ,

[0105] where Δu k+j = u k+j - u k+j-1 and N z = {1, 2, …, N}, N u = {0, 1, …, N - 1}. N can be the control layer and / or the prediction layer, and j is the iteration number. The control layer and the prediction layer are different, as shown in N z and N u . The estimated current state is allocated to the initial state through constraint (a). Constraints (b) and (c) are augmented state space constraints, such as and the offset constant related to the linearization of the control model 420 and σ z . Constraint (d) represents the input limit, also known as the input constraint, which guarantees the safe operating area of the CFMF device 402. Since there is no available prediction for the target r k and the measured disturbance value d k , the same prediction as the current one is used because it is unknown how r k or d k will change within the control and prediction ranges. The output constraint of z k+j can also use a slack variable (not shown here for simplicity).

[0106] In some embodiments, existing methods such as the active set method, the interior point method, and the first-order gradient method are used to solve the QP. For example, the primal-dual interior point QP solver can be used based on a customized Mehrotra predictor-corrector algorithm. It is best to use a warm start to speed up the solution time. For example, a previous solution can be reused or early termination can be performed. The solution can subsequently be defined by the function

[0107] which defines the online calculation of the MPC layer 424.

[0108] Advantageously, using the QP based on the above-mentioned multiple constraints results in an optimal input value u k , which can be safely output to the CFMF device 402 without manual or automatic intervention because the limitations related to the safety operating procedures and the operating limitations of the CFMF components have been incorporated into the QP.

[0109] Preferably, the optimizer 422 further includes a real-time optimization (RTO) layer 436, where the RTO layer 436 is configured to determine (1), multiple target controlled output values 432 (r k ), which are obtained by the above-mentioned MPC layer 424. Alternatively, the control model 420 can obtain (1) from an alternative source or method, such as from the input interface 408.

[0110] The RTO layer 436 is configured to determine the optimal future input value 434 based on the future efficiency of the CFMF device 402, and further determine the target controlled output value 432 based on the optimal future input value 434. For example, the RTO layer 436 can be configured to determine the optimal future input value 434 and / or the optimal output value of the optimization objective function, where the objective function estimates the future efficiency of the filtration process of the CFMF device 402 according to one or more constraints (such as those previously disclosed). One or more constraints for optimizing the objective function can include the minimum concentration of the target component in the outlet hold-up stream, the maximum concentration of the target component in the outlet hold-up stream, and one or more of the one or more input value limitations associated with the CFMF device 402.

[0111] The control unit 406 can also be configured to obtain the constraints associated with the CFMF device 402 for optimizing the first cost function. An example of the cost function and related constraints is disclosed below. In another example, the constraints can be obtained from the interface unit 404, and the constraints can include the operating constraints of the CFMF device 402, including the constraints associated with any of the following: the water flow rate associated with the CFMF device 402; the controlled output variables associated with the CFMF device 402; the controlled input variables associated with the CFMF device 402; and the manipulated variable constraints associated with the CFMF device 402. The manipulated variable is based on the controlled output variable or the controlled input variable associated with the CFMF device 402 that has been processed within the control unit 406.

[0112] Multiple optimal future input values 434 of the optimization objective function can be determined by using an interior point algorithm for non - linear problems. Alternatively, linear problems can be used to determine the target, optimal, future input values 434. Additionally, multiple input values 434 and / or output values that optimize the difference between multiple target controlled output values 432 and multiple estimated controlled output values can be determined by using a quadratic programming solver. In one example, the quadratic programming solver is one of the active set method, the interior point method, and the first - order gradient method, or any other suitable method.

[0113] For example, an objective function (p) can be formed to include the optimal steady - state controlled output (z ss ), the input of the optimal steady - state (u ss ), and the disturbance (d k ) that affects the current state. The constraints can be formed as penalty functions, and if the constraints are violated, the objective function cannot be optimized. Thus, the optimization problem may include the objective function (p) and the penalty function. Then, multiple target controlled output values 432 (r ss ) can be determined from the resulting controlled output z k of the resulting optimization problem.

[0114] Preferably, the above - disclosed RTO layer 436 solves the optimization problem

[0115] including the constraints (e) (f) (g) (h)u min +δ u ≤u ss ≤u max -δ u , (i)z min -s + δ z ≤z ss ≤z max +s - δ z , and (j)s≥0.

[0116] The problem function is the sum of the objective function p(z ss ,u ss ,d k ) and the penalty function that penalizes violations of the output constraints. When the above problem is solved, the target is set to the optimal controlled output value, i.e., r k =z ss. For simplicity, the integral disturbance state is fixed to its current value by constraint (e). The control model 420 is used in constraints (f) and (g) to determine the input u ss and the steady-state relationship between the controlled output z ss . u min and u max define the aforementioned system and process input constraints. z min and z max define the system and process output constraints. δ u and δ z include a fallback in the manipulated and controlled variables to maintain the controllability of the MPC layer 424. s w is a preselected variable that is typically large enough to avoid constraint violations, S w = diag(s w ). Preferably, s w is customized specifically for the CFMF device 402. The non-linear profit function p can also be customized for the CFMF device 402. The input and output constraints together provide a region that ensures safe operation.

[0117] According to the formulation of the optimization problem, existing software functions can be used to solve the optimization problem. For example, the sequential quadratic programming (SQP) method can be used to solve the above problem function, where the SQP method is a quasi-Newton implementation that uses line searches to select step sizes. The solution can be defined by the function

[0118] where, is the current state estimate, where only the offset state is used for model correction, d k is the disturbance 426 of the current state k. The determined target controlled output value 432 (r k ) can subsequently be obtained by the MPC layer 424 to determine multiple input values 434 (u ss ), as described above.

[0119] A set of example online calculations for the RTO layer 436 are disclosed below. The online calculations of the RTO layer 436 require three inputs, including: the corrected current state obtained from the control model 420 which includes the controlled output value according to the state estimate of the control model 430; multiple disturbance values 426 d k , which can be obtained from the input interface 408 or otherwise from the CFMF device 402; and the number of iterations k of the current state. The optimal solution state x * is composed of the optimal input u ss and the optimal controlled output z ss that optimizes the cost function q.

[0120] To solve the NLP problem which can be the above problem function, the cost function q needs to be calculated and the relevant constraints c eq and c ineq , where x init is the initial state. c eq represents the difference between the controlled output of the cost function and the optimal controlled output z ss while c ineq represents the aforementioned system and process constraints. The cost function q estimates the future efficiency of the filtration process according to the function constraints S w , s T and s w . As described above, the functional constraints are pre-determined, can be customized for the CFMF device 402, and s T is introduced to consider any additional changes introduced in the online calculation method. Generally speaking, these constraints provide a region to ensure safe operation, and consider the potential limitations introduced by the components of the CFMF device 402 and the mathematical and / or computational methods adopted when optimizing the system 400. Then, the controlled output z ss of the solution state is determined using an existing software solver (e.g., solve), so as to obtain the target controlled output value 432r k .

[0121] Required: d k , k

[0122] Solve the NLP: [u ss ,z ss ,s] = unpack(x * ) r k = z ss

[0123] Return r k

[0124] Required: [u,z,s] = unpack(x * )

[0125] Calculate the cost function

[0126] Calculate the equality constraints z ss = C y x ss + C d xd,k +σ z c eq = z - z ss

[0127] Calculate inequality constraints c ineq = [s ≥ 0; u min +δ u ≤ u ≤ u max -δ u ; z min -s + δ z ≤ z ss ≤ z max +s - δ z

[0128] The control unit 406 can also be configured to act as a soft sensor. A soft sensor is a virtual sensor whose output data includes measured values predicted by mathematical or computational methods, rather than using measured values of hardware sensors and actuators. Generally, a soft sensor uses input data to determine predicted measured values and obtains indirect measured values of target variables. In the case where one or more of the multiple controlled output values 428 are unknown, additional assumptions can be made in the control unit 406. Physical-based (or any statistics-based, machine learning-based, or other artificial intelligence-based) models can be used to estimate the unknown values. Examples of soft sensors used in this application are described in detail below.

[0129] Although multiple controlled output values 428, y k are obtained from one or more sensors of the CFMF device 402 (such as Figure 2 sensor 240), at least one of the multiple controlled output values 428 can be obtained from a soft sensor, such as when y k changes in magnitude due to any missing measurements. For example, the optimizer 422 requires current values of target component contents (such as protein and total solids in the final concentrated stream). In the case where these sensor values are not available, the optimizer 422 can predict these missing values through a soft sensor.

[0130] In one example, the soft sensor can include a control model 420 such that the control model 420 estimates multiple controlled output values 428. As previously described, the control model 420 constructs attributes related to the measurement (providing ​) Handle missing observations. Advantageously, this causes the control unit 406 to remain operational when the number of controlled output values 428 among the multiple controlled output values changes. For example, when a particular sample is missing a particular variable associated with the CFMF device 402 at a corresponding time point, the number of controlled output values among the multiple controlled output values 428 changes. Specifically, due to a mechanical failure of the sensor, the controlled output values of the multiple controlled output values 428 from the current state may be unavailable. The lack of previously available data results in a change in the magnitude of the vector for the multiple controlled output values 428.

[0131] In another example, the soft sensor calculations for obtaining the multiple controlled output values 428 may include calculating the inlet components of each of the multiple membrane stacks. An example of soft sensor calculations uses mass balance, also known as the mass balance method, which applies the conservation of mass to the analysis of the filtration process. By considering the materials entering and leaving the system, other unknown mass flows, such as feedstock and water flow, can be determined. Difficult-to-measure and / or other unquantifiable measurements can be determined by applying known measurements to mass balance techniques and solving the conservation equations (mass balance equations) for the known measurements.

[0132] An example application of the mass balance technique according to one aspect of the present disclosure is as follows. The unknown feedstock flow F f and the unknown feedstock component x f of the feedstock (e.g., Figure 2 the input feedstock 212 in Figure 2 ) enter the stack X (e.g., p the membrane stack 230 in p ), with a total of N stacks. The permeate flow (e.g., the permeate flow 238) leaves the permeate flow F f and the unknown permeate component x p in the stack X. x

[0133] Other variables include the input water flow F w and the water component x w (such as Figure 2 the input water flow 214 in Figure 2 ), and the retentate flow (such as r the outlet retentate flow 218 in r ) leaves the retentate flow F

[0134] The component transport through the filter of the stack X can be written as a function J s (x r , F p , x p ).

[0135] The resulting mass balance equation is Fr = F w + F f - F p F r * x r = F w * x w + F f * x f - F p * x p 0 = F p * x p - A * J s (x r , F p , x p )

[0136] where A is the membrane area 230 specific to the CFMF device 402, and J s (x r , F p , x p ) is a function of the solute flux, which depends on the membrane selection and the permeation resistance of the components, and is specific to the CFMF device 402. These values can be obtained through mathematical methods and parameter fitting tasks. For example, the solute flux J s (x r , F p , x p ) can be calculated from the solute concentration the solute permeability ω and the reflection factor σ r The σ r represents the selectivity of the membrane, and its value generally ranges from 0 to 1, where: σ r = 1 represents an ideal membrane with no solute transport σ r < 1 represents a non-ideal membrane, not a perfect semi-permeable membrane, including solute transport, and σ r = 0 represents a membrane with no selectivity.

[0137] The solute flux J s function can be expressed as

[0138] where J v is the volume flux, denoted as J v = L p (ΔP - σ r Δπ). L p is the solvent (water) permeability, ΔP is the hydrodynamic pressure difference / applied pressure across the membrane, and Δπ is the osmotic pressure difference across the membrane. L pIt can be obtained from pure water experiments, where since J v and ΔP have a linear relationship and Δπ = 0. The factors ω and σ r can be obtained from osmotic diffusion experiments. For example, using the CFMF system 400 and performing additional laboratory measurements to subsequently solve the mass balance equations for the factors ω and σ r .

[0139] Solving the three mass balance equations for the unknown variables F f , x f and x p , the variables 428 required to obtain multiple controlled output values can be estimated or determined.

[0140] This soft sensor calculation method can be repeated for each membrane in multiple membrane stacks and can be repeated, for example, N times for each membrane stack.

[0141] The filtration process of the CFMF device 402 disclosed above can be one of a crossflow UF process, an MF process, or an NF process. For example, the filtration process can be a crossflow UF process, where the feed stream includes dairy raw materials and the target component in the raw materials of the outlet retentate stream is protein. The water associated with the CFMF device 402 can be deionized water or water extracted from another CFMF process. Other example target components can include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin.

[0142] Optionally, the optimization system 400 further includes at least one programmable logic controller (PLC) and an edge computing device communicatively coupled to the at least one PLC and including one or more processors and a memory. In one example, the PLC includes an input interface 408 and an output interface 412, and the edge computing device includes a control model 420 and an optimizer 422.

[0143] The optimization system 400 can be used to optimize Figure 3 the CFMF process period 306 in.

[0144] One aspect of the present disclosure relates to Figure 5 a water optimization system.

[0145] Figure 5 A water optimization system 500 and a CFMF device 502 are shown. The water optimization system 500 includes an interface unit 504 and a control unit 506. The interface unit 504 further includes an output interface 508 and a connection 510. The control unit 506 further includes a membrane control model 512, an estimated concentration value 514, an optimization unit 516, an input membrane water flow value 518, and an optimal input water flow 520.

[0146] Figure 5 shows a water optimization system 500 for optimizing the water consumption of a CFMF device 502, which includes a plurality of membranes, and each membrane performs a filtration process of dividing a feed stream and a water stream into a retentate stream and a permeate stream. The water optimization system 500 includes a membrane control model 512 associated with the plurality of membranes. The membrane control model 512 is configured to estimate a concentration value 514 based on an input membrane water flow value 518, and the input membrane water flow value 518 corresponds to the input water flow of each membrane in the plurality of membranes, where the concentration value is related to the estimated concentration of the solid components in the retentate stream.

[0147] The water optimization system 500 further includes an optimization unit 516, and the optimization unit 516 is configured to obtain an optimal concentration value related to the solid components in the retentate stream and determine an optimal input water flow 520 for each membrane in the plurality of membranes (e.g., each membrane stack), which minimizes the difference between the optimal concentration value and the estimated concentration value 514. Wherein, according to the optimal input water flow 520 of each membrane in the plurality of membranes, the estimated concentration value is determined from the membrane control model 512.

[0148] The water optimization system 500 further includes an output interface 508 communicatively coupled to the CFMF device 502 using a connection 510, and the output interface 508 is configured to cause the CFMF device 502 to adjust the water flow at each membrane in the plurality of membranes of the CFMF device 502 according to the optimal input water flow 520 determined for each membrane in the plurality of membranes. This may reduce the difference between the optimal concentration value and the future concentration value. Optionally, the future water flow value is obtained by the water optimization system 500 through a connection 524 and / or the input membrane water flow value 518. For example, the future water flow value may be the input membrane water flow value 518. The water optimization system 500 may further include an input interface 522 and a connection 524. For example, the input interface 522 may be communicatively coupled to the CFMF device 502 through a connection 524 (e.g., through a wired connection, an Ethernet connection, or a wireless network connection).

[0149] The membrane control model 512 includes one of a physics-based model, a statistical model, or a machine learning model. For any of these methods, the membrane control model 512 can estimate a concentration value 514 related to the solid components in the outlet retentate stream based on the obtained controlled output values (e.g., Figure 4 multiple controlled output values 428 in

[0150] In one example, the membrane control model 512 is further configured to estimate the concentration value 514 based on a plurality of filter specifications related to the plurality of membranes, where each of the plurality of filter specifications includes the water permeability L p of the associated membrane, the solute concentration the solute permeability ω, the reflection factor σ r, the membrane area that can be used to calculate pressure (e.g., hydrodynamic pressure difference ΔP and / or osmotic pressure difference Δπ).

[0151] For example, the membrane control model 512 can use the previously disclosed mass balance equations for soft sensor calculations. Specifically, the optimization unit 516 and / or the membrane control model 512 can use the previously disclosed mass balance equations F r *x r =F w x* w +F f *x f -F p *x p

[0152] The aim is to minimize F w to meet the concentration criteria, where the concentration criteria are set by minimizing the difference 514 between the optimal concentration value and the estimated concentration value and / or maximizing the total concentration (total solids) in the retentate flow, without breaking the predefined constraints. The optimization process is configured to distribute the water flow to the membrane stack so that the water is optimally utilized.

[0153] The mass balance equations can be used to form the objective function to be minimized, such as a non-linear function derived from the mass balance including the sum of all water flows, or a linear function (e.g., the linear version of a non-linear equation). The objective function can be used to optimize the CFMF process. The objective function is subject to an upper constraint equal to the total solids and target component content, and a terminal equality constraint equal to the total solids and target component content. This results in the optimal water flow separation for each stack while still meeting the product quality specifications. Alternative methods of the method can include direct search methods or extremum search methods, which search or control the concentration of the target component / total solids in each stack only using the values of the objective function. Or, the best separation search solution can be searched by applying an automatic trial-and-error method.

[0154] More specifically, the mass balance technique according to one aspect of the present disclosure is as follows. The feed stream F f and the feed component x f of the feed (e.g., Figure 2 the input feed 212 in Figure 2 ) enter the stack X (e.g., p the membrane stack 230 in p ), and there are a total of N stacks. The permeate flow (e.g., the permeate flow 238) leaves the permeate flow F f and the unknown permeate component x p in the stack X. x f and x p can include multiple parameters, such as protein, lactose, fat, and ash. Other variables include the input water flow F w and the water component x w (e.g.,Figure 2 The input water flow 214), the retention flow (such as Figure 2 the outlet retention flow 218) leaves the retention flow F in the laminate X r and retains the component x r . The component transport through the filter of the laminate X can be written as a function J s (x r , F p , x p ). The resulting mass balance equation is as described above: F r = F w + F f - F p F r * x r = F w * x w + F f * x f - F p * x p 0 = F p * x p - A * J s (x r , F p , X p )

[0155] where A is the membrane area specific to the CFMF device 402 (e.g., Figure 2 the membrane area of the membrane laminate 230 of s (x r , F p , X p ) is a function of the solute flux, which depends on the membrane selection and the permeation resistance of the component, and is specific to the CFMF device 402.

[0156] These values can be obtained through mathematical methods and parameter fitting tasks. For example, the solute flux J s (x r , F p , X p ) can be calculated from the solute concentration the solute permeability ω and the reflection factor σ r . σ r represents the selectivity of the membrane, and its value generally ranges from 0 to 1, where: σ r = 1 represents an ideal membrane with no solute transport σ r < 1 represents a non-ideal membrane, not a perfect semi-permeable membrane, including solute transport, and σ r= 0 indicates that the membrane has no selectivity.

[0157] Solute flux J s The function can be expressed as

[0158] where J v is the volume flux, denoted as J v = L p (ΔP - σ r Δπ). L p is the solvent (water) permeability, ΔP is the hydrodynamic pressure difference / applied pressure on the membrane, and Δπ is the osmotic pressure difference on the membrane. L p can be obtained from pure water experiments, where since J v and ΔP have a linear relationship, Δπ = 0. The factors ω and σ r can be obtained from permeation diffusion experiments. For example, using the CFMF system 400 and performing additional laboratory measurements to subsequently solve the mass balance equations for the factors ω and σ r .

[0159] By solving three mass balance equations, the variables required to obtain the concentration value 514 can be estimated or determined.

[0160] Optionally, the water optimization system 500 also includes a real-time optimizer (RTO), where the target concentration value is obtained from the RTO. The target concentration value can be the optimal concentration value. For example, the RTO can be Figure 4 the RTO layer 436 in. The RTO is configured to determine multiple optimal future input values that optimize the first cost function, where the first cost function estimates the future efficiency of the filtration process based on one or more constraints. The RTO is also configured to determine the target / optimal concentration value from the device control model associated with the CFMF device 502 based on the multiple optimal future input values, where the membrane control model 512 estimates the concentration value 514 from the input values based on the state of the device control model and / or the input membrane water flow value 518. Alternatively, the target / optimal concentration value can be predetermined or obtained in other ways, such as from an operator or an algorithm.

[0161] One or more constraints for optimizing the first cost function can include the minimum concentration of the target component in the outlet retentate stream, the maximum concentration of the target component in the outlet retentate stream, and one or more of the one or more input value limitations associated with the CFMF device 502.

[0162] The optimal input water flow 520 for each membrane in the multiple membranes can also be determined using the sequential quadratic programming optimization method, or determined using one of the direct search method, the extremum search method, or the trial-and-error method.

[0163] The water optimization system 500 can perform an iterative process that can be repeated until a condition is met. The water optimization system 500 can continue or otherwise repeat the process of optimizing the difference between the target / optimal concentration value and the estimated concentration value.

[0164] The CFMF device 502 can be Figure 2 the CFMF system 200 of Figure 4 and / or the CFMF device 402 of

[0165] In one example, the filtration process is a crossflow UF process, the feed stream includes dairy raw materials, and the target component in the feed of the product stream is protein. Other example target components can include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin. The water optimization system 500 reduces the water consumption of the crossflow UF process, and the membrane control model 512 uses a previously disclosed soft sensor method to predict the concentration of each component (lactose, minerals, protein, and water) in all streams of the process. Knowing the estimated concentration of the components and the membrane filter specifications, the optimal water flow into each membrane stack is estimated. This results in an increase in total solids and protein from the optimization unit 516 to the specified level, while not blocking the membrane and reducing the amount of water required in the process.

[0166] In another example, the water associated with the CFMF device 502 is deionized water.

[0167] Optionally, the water optimization system 500 further includes at least one PLC and an edge computing device communicatively coupled to the at least one PLC and including one or more processors and a memory. For example, the PLC includes an input interface 522 and an output interface 508, and / or the edge computing device includes a membrane control model 512 and an optimization unit 516.

[0168] One aspect of the present disclosure relates to Figure 6 an optimization system of

[0169] Figure 6 The system 600 and the CFMF device 602 are shown, which are configured to optimize the filtration process and water consumption. Figure 6 Also shown are an input interface 604, a controlled output value 606, a disturbance value 608, a device control model 610, a controlled output value estimated according to the state of the device control model 612, a plurality of controlled output values 614, and a plurality of disturbance values 616. The system 600 further includes a membrane control model 618, an estimated concentration value 620, an input water flow 622, and an optimizer 624. Figure 6Also shown are the MPC layer 626, the target controlled output value 628, the optimal input value 630, the WFO layer 632, the target / optimal concentration value 634, the optimal input water flow 636, the output interface 638, the measured value 640, and the RTO layer 642. Figure 6 Also included are an interface unit 644, connections 646, 648, and a control unit 650.

[0170] System 600 optimizes a filtration process performed on a CFMF device 602, which includes a plurality of membranes, each membrane performing a filtration process of separating a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. The plurality of membranes can form one or more membrane stacks. System 600 includes an input interface 604 communicatively coupled to the CFMF device 602, the input interface 601 being configured to obtain a plurality of controlled output values 606 related to the filtration process performed on the CFMF device 602, where the plurality of controlled output values 606 include: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of components in the outlet product stream; and to obtain a plurality of disturbance values 608 related to the CFMF device 602. The product stream can include the permeate stream and / or the retentate stream. For example, (i) can be a retentate concentration value indicating the protein concentration in the outlet retentate stream, and (ii) can be a total concentration value indicating the concentration of all solid components in the outlet retentate stream. System 600 also includes a device control model 610 related to the CFMF device 602, where the device control model 610 estimates a controlled output value from a plurality of controlled output values 614 and a plurality of disturbance values 616 according to the state of the device control model 612.

[0171] System 600 also includes a membrane control model 618 related to the plurality of membranes, where the membrane control model 618 is configured to estimate a concentration value 620 according to the input water flow 622 of each of the plurality of membranes, where the concentration value is related to the estimated concentration of solid components in the outlet retentate stream.

[0172] System 600 further includes an optimizer 624, which includes a Model Predictive Control (MPC) layer 626 and a Water Flow Optimization (WFO) layer 632. The optimizer 624 is configured to update the state of the device control model 610 based on a plurality of controlled output values 614. The MPC layer 626 is configured to: obtain a plurality of target controlled output values 628; and determine a plurality of input values 630 based on the state of the device control model 612, where the input values 630 optimize a first difference between the plurality of target controlled output values 628 and a plurality of estimated controlled output values, and the plurality of estimated controlled output values are determined from the device control model 610 through the plurality of controlled output values 614. The WFO layer 632 is configured to obtain an optimal concentration value 634 (or target concentration value) related to the solid component in the outlet retention flow; determine an optimal input water flow 636 for each of the plurality of membranes that minimizes the difference between the target / optimal concentration value 634 and the estimated concentration value 620, where the estimated concentration value 620 is determined by the membrane control model 618 based on the input water flow 622 for each of the plurality of membranes; and adjust the plurality of input values 630 determined by the MPC layer 626 according to the optimal input water flow 636 for each of the plurality of membranes.

[0173] System 600 further includes an output interface 638 communicatively coupled to the CFMF device 602. The output interface 638 is configured to output the plurality of input values 630 to the CFMF device 602, thereby reducing a second difference between the plurality of target controlled output values 628 and future plurality of controlled output values.

[0174] The input interface 604 and the output interface 638 can be components of the interface unit 644. Optionally, the device control model 610, the membrane control model 618, and the optimizer 624 can be components of the control unit 650. The input interface 604 is communicatively coupled to the CFMF device 602 through a connection 646, and the output interface 638 2 is communicatively coupled to the CFMF device 602 through a connection 648. For example, the interface unit 644 can be communicatively coupled to the CFMF device 602 through a wired connection, an Ethernet connection, or a wireless network connection. Optionally, the output interface 638 can be Figure 4 the output interface 412. The input interface 604 can be Figure 4 the input interface 408.

[0175] The membrane control model 618 can be a physics-based (or optionally, any statistics-based, machine learning-based, or other artificial intelligence-based) model, and can also be configured to estimate the concentration value 620 based on a plurality of filter specifications related to the plurality of membranes. Optionally, each of the plurality of filter specifications includes the water permeability L of the first associated membrane p , solute concentration solute permeability ω, reflection factor σ r, the membrane area that can be used to calculate pressure. For example, the membrane control model 618 can use the previously disclosed mass balance equation. Optionally, the membrane control model 618 can obtain the measurement value 640 from the input interface 604. The measurement value 640 can include values related to concentration and / or water flow, such as the known variables of the previously disclosed mass balance equation.

[0176] Preferably, the optimizer 624 further includes a real-time optimization (RTO) layer 642, which is configured to determine the multiple target controlled output values 628 obtained by the MPC layer 626. In one example, the RTO layer 642 is further configured to determine multiple optimal future input values that optimize the first cost function, where the first cost function estimates the future efficiency of the filtration process according to one or more constraints. The RTO layer 642 further determines the multiple target controlled output values 628 from the device control model 610 according to the multiple optimal future input values.

[0177] The MPC layer 626 and the RTO layer 642 can utilize Figure 4 the disclosed mathematical methods, and the MPC layer 626 and the RTO layer 642 can be the MPC layer 424 and the RTO layer 436 respectively.

[0178] One or more constraints for optimizing the first cost function can include the minimum concentration of the target component in the outlet hold-up stream (such as Figure 2 the hold-up stream 232 in min ), the maximum concentration of the target component in the outlet hold-up stream, and one or more input value limitations related to the CFMF device 602 (such as, u max ).

[0179] The multiple input values 630 can include a concentration factor and a water factor. The concentration factor includes the first ratio of the feed stream to the outlet hold-up stream, and the water factor includes the second ratio of the water flow to the feed stream. Optionally, the output interface 638 is further configured to: output the outlet hold-up stream set point to the CFMF device 602 according to the concentration factor; and / or output the water flow set point to the CFMF device 602 according to the water factor. The WFO layer 632 can also be configured to adjust the water factor of the multiple input values 630 according to the optimal input water flow 636 of each membrane in the multiple membranes.

[0180] The filtration process can be one of a crossflow UF process, an MF process, or an NF process. For example, the filtration process is a crossflow UF process, the feed stream includes dairy raw materials, and the target component in the raw materials of the outlet hold-up stream is protein. Other example target components can include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin.

[0181] Optionally, system 600 further includes at least one PLC and an edge computing device communicatively coupled to the at least one PLC and including one or more processors and a memory. For example, the PLC includes an input interface 604 and an output interface 638, and / or the edge computing device includes any one of a device control model 610, a membrane control model 618, and an optimizer 624.

[0182] Preferably, the water associated with the CFMF device 602 is deionized water.

[0183] According to one aspect of the present disclosure, system 600 is used during an optimized Figure 3 CFMF process period 306.

[0184] One aspect of the present disclosure relates to Figure 7 an optimization process.

[0185] Figure 7 A flowchart showing a method of optimizing a filtration process performed on a CFMF device that divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. The method is performed in a control system communicatively coupled to the CFMF device, such as the control system described in reference Figure 4 or Figure 6 in detail.

[0186] The method includes performing an optimization process 700, and the optimization process 700 includes the following steps.

[0187] Step 702 includes: obtaining a plurality of controlled output values related to the filtration process performed on the CFMF device (e.g., Figure 4 the controlled output value 418 in Figure 4 ), wherein the plurality of controlled output values include: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of components in the outlet product stream. The obtaining of the plurality of controlled values is described with reference to

[0188] Figure 4 Figure 4 The product stream may include a permeate stream and / or a retentate stream. For example, (i) may be a retentate concentration value indicating the protein concentration in the outlet retentate stream, and (ii) may be a total concentration value indicating the concentration of all solid components in the outlet retentate stream.

[0189] Figure 4 Figure 4 ​​​​the state of the control model 430) and a plurality of disturbance values (e.g., Figure 4 a plurality of disturbance values 426) to estimate a controlled output value from the input values.

[0190] Step 708 includes: updating the state of the control model according to a plurality of controlled output values (e.g., Figure 4 a plurality of controlled output values 428 in ).

[0191] The control model and the functions of the control model used in steps 704 and 706 are preferably as described in detail in reference Figure 4 .

[0192] Step 710 includes: obtaining a plurality of target controlled output values (e.g., Figure 4 target controlled output values 432). Reference Figure 4 describes the process of obtaining the target values.

[0193] Steps 712 and 714 are optional steps and include an example for obtaining the plurality of target controlled output values required in step 710. Therefore, the process can directly jump from step 710 to step 716 to ignore this process. In other embodiments, the process can ignore one of the two steps, that is, the process includes steps 710, 712, and 716 (i.e., ignoring step 714) or steps 710, 714, and 716 (i.e., ignoring step 712). Alternatively, the process can include all steps, that is, execute steps 710, 712, 714, and 716.

[0194] Step 712 includes: determining a plurality of optimal future input values (e.g., Figure 4 optimal future input values 434 in ) that optimize a first cost function, where the first cost function estimates the future efficiency of the filtration process according to one or more constraints. Step 714 includes: determining a plurality of target controlled output values from the control model according to the plurality of optimal future input values. The processes of steps 712 and 714 are described in reference 4.

[0195] Step 716 includes: determining a plurality of input values (e.g., Figure 4 optimal future input values 434 in ) that optimize a first difference between the plurality of target controlled output values and the plurality of estimated controlled output values, where the plurality of estimated controlled output values are determined from the control model by the plurality of input values; and

[0196] Step 718 includes: outputting the plurality of input values to the CFMF device, thereby reducing a second difference between the plurality of target controlled output values and the plurality of future controlled output values. The processes of steps 716 and 718 are shown in reference Figure 4 .

[0197] The plurality of input values ​​of step 716 may include a concentration factor including a first ratio of the feedstock flow to the outlet retentate flow and a water factor including a second ratio of the water flow to the feedstock flow. Optionally, step 716 further includes: outputting an outlet retentate flow set point to the CFMF device based on the concentration factor and / or outputting a water flow set point to the CFMF device based on the water factor.

[0198] Optionally, step 710 further includes: obtaining a signal from one or more sensors (eg Figure 2 Alternatively, step 710 further includes obtaining a plurality of controlled output values ​​from the control model of step 706. Figure 4 An example of obtaining a plurality of controlled output values ​​from a control model is described, wherein the controlled output values ​​may optionally be obtained using soft sensors.

[0199] The optimization process 700 may be performed repeatedly, as the optimization process 700 may form part of an iterative process.

[0200] The filtration process of the optimization process 700 can be a cross-flow UF process, wherein the feed stream includes a dairy feed and the target component in the feed of the retentate stream is protein. Other example target components can include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin. Preferably, the water associated with the optimization process 700 is deionized water.

[0201] Preferably, the optimization process 700 is performed using the optimization system 400 .

[0202] One aspect of the present disclosure relates to Figure 8 The water optimization process is preferably used Figure 5 The CFMF device 502.

[0203] Figure 8 A flow chart showing a method for optimizing water usage on a cross-flow CFMF device, the device comprising a plurality of membranes, each membrane performing a filtration process for separating a feed stream and a water stream into an outlet intercept stream and an outlet permeate stream, the method being performed on a control system communicatively coupled to the CFMF device, preferably Figure 5 Water optimization system 500 in.

[0204] The water optimization process 800 includes the following steps.

[0205] Step 802 includes obtaining a target, such as an optimal concentration value associated with the solid component in the outlet retentate flow. Figure 5 shown.

[0206] Optional steps 804 and 806 illustrate an example method for obtaining optimal concentration values. Optional examples include: obtaining the target / optimal concentration value from a control model, obtaining the target / optimal concentration value using a soft sensor, obtaining the target / optimal concentration value from an input interface, or the target / optimal concentration value can be predetermined.

[0207] Steps 804 and 806 are optional, which means that the process includes all steps, namely steps 802, 804, 806. Alternatively, the process may not include any optional steps, that is, the process goes directly from step 802 to step 808. Or, the process includes one of the two steps, that is, the process includes steps 802, 804 and 808 or steps 802, 806 and 808.

[0208] Step 804 includes: determining a plurality of optimal future input values that optimize a first cost function, where the first cost function estimates the future efficiency of a filtration process based on one or more constraints. The form of the cost function is preferably as shown in the reference Figure 5 as shown.

[0209] Step 806 includes: determining the target / optimal concentration value from a device control model associated with the CFMF device (such as Figure 4 the control model 420 therein) based on a plurality of optimal future input values, where the device control model estimates the concentration value from the input value according to the state of the device control model (such as Figure 4 the state of the control model 430 therein). The process is also as described in the reference Figure 5 as described.

[0210] Step 808 includes: obtaining a membrane control model associated with a plurality of membranes, such as Figure 5 the membrane control model 512 of Figure 2 , where the membrane control model estimates the concentration value (such as Figure 5 the estimated concentration value 514 of Figure 5 ) based on the input water flow of each of the plurality of membranes; where the concentration value is related to the estimated concentration of the solid component in the outlet retentate stream. Preferably, the process is described with reference to Figure 5 described.

[0211] Optionally, the process continues to step 810. Alternatively, the process continues from step 808 to step 812.

[0212] In step 810, the process includes: using the membrane control model to estimate the concentration value based on the input water flow of each of the plurality of membranes. The membrane control model is as shown in Figure 4 and Figure 5 . Additionally, the membrane control model can provide the estimated concentration value in a continuous iterative process without usage requirements.

[0213] Step 812 includes: determining an optimal input water flow for each of a plurality of membranes for minimizing the difference between a target / optimal concentration value and an estimated concentration value (e.g., Figure 5 estimated concentration value 514); wherein, an estimated concentration value is determined from a membrane control model (e.g., Figure 5 membrane control model 512) according to the optimal input water flow for each of the plurality of membranes. Figure 5

[0214] Step 814 includes: causing the CFMF device to adjust the water flow at each of the plurality of membranes of the CFMF device according to the optimal input water flow determined for each of the plurality of membranes (e.g., Figure 5 optimal input water flow 520). The processes of steps 812 and 814 are preferably as disclosed in reference Figure 5 .

[0215] One or more constraints for step 804 for optimizing the first cost function may include one or more of a minimum concentration of a target component in the outlet retentate stream, a maximum concentration of the target component in the outlet retentate stream, and one or more input value limitations associated with the CFMF device.

[0216] The membrane control model of step 808 may estimate concentration values according to a plurality of filter specifications associated with a plurality of membranes. For example, each of the plurality of filter specifications includes the water permeability, solute permeability, reflection factor, and membrane area of the associated membrane. In one example, the membrane control model may utilize the mass balance equations previously described in Figure 4 and Figure 5 .

[0217] The filtration process of the water optimization process 800 may be optionally one of a crossflow UF process, an MF process, or an NF process. For example, the filtration process may be a crossflow UF process, the feed stream includes dairy raw materials, and the target component in the raw materials of the outlet retentate stream is protein. Other example target components may include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin. Preferably, the water associated with the water optimization process 800 is deionized water.

[0218] The water optimization process 800 may be repeatedly executed because the water optimization process 800 may form part of an iterative process.

[0219] Preferably, the water optimization process 800 is executed using the water optimization system 500.

[0220] One aspect of the present disclosure relates to the optimization process of FIG. 9, preferably using the Figure 6 CFMF device 602.

[0221] Figures 9A - 9B A flowchart showing a method for optimizing a filtration process on a CFMF device, the device including a plurality of membranes, where each membrane performs a filtration process of dividing a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream. The method is executed on a control system communicatively coupled to the CFMF device, preferably Figure 6 the system 600 in. The method includes executing an optimization process 900, and the optimization process 900 includes the following steps.

[0222] These steps are described with reference to Figure 4 、 5 and equivalent features in 6.

[0223] Step 902 includes: obtaining a plurality of controlled output values related to the filtration process performed on the CFMF device (e.g., Figure 6 the controlled output value 614 in), where the plurality of controlled output values includes: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of components in the outlet product stream. The product stream may include the permeate stream and / or the retentate stream. For example, (i) may be a retentate concentration value indicating the protein concentration in the outlet retentate stream, and (ii) may be a total concentration value indicating the concentration of all solid components in the outlet retentate stream.

[0224] Step 904 includes: obtaining a plurality of disturbance values related to the CFMF device (e.g., Figure 6 the disturbance value 616 in).

[0225] Step 906 includes: obtaining a device control model of the CFMF device (e.g., Figure 6 the device control model 610 of), where the device control model estimates the controlled output value from the input value according to the state of the device control model (e.g., Figure 6 the state of the device control model 612 of) and a plurality of disturbance values (e.g., Figure 6 the disturbance value 616 of); and updating the state of the device control model according to a plurality of controlled output values (e.g., Figure 6 the controlled output value 614 in).

[0226] Step 908 includes: obtaining a plurality of target controlled output values (e.g., Figure 6 the target controlled output value 628 of).

[0227] Steps 910 and 912 show an example method for obtaining a plurality of target controlled output values. Optional examples include: obtaining the optimal concentration value from the control model, obtaining the optimal concentration value using a soft sensor, obtaining the optimal concentration value from the input interface, or the optimal concentration value may be predetermined. In other embodiments, the optimal concentration value may be the target concentration value.

[0228] Steps 910 and 912 are optional steps. Thus, the process can jump directly from step 908 to step 914. In other embodiments, the process can skip one of the two steps, i.e., the process includes steps 908, 910, and 914 (i.e., skipping step 912) or steps 908, 912, and 914 (i.e., skipping step 910). Alternatively, the process can include all steps, i.e., it performs steps 908, 910, 912, and 914. Step 910 includes determining a plurality of optimal future input values that optimize a first cost function, where the first cost function estimates the future efficiency of a filtration process based on one or more constraints.

[0229] Step 912 includes: determining a plurality of target controlled output values (e.g., Figure 6 target controlled output values 628 of Figure 6 device control model 610) from the device control model (e.g.,

[0230] based on the plurality of optimal future input values. Figure 6 Step 914 includes: determining a plurality of input values (e.g., Figure 6 input values 630 of Figure 6 that optimize a first difference between the plurality of target controlled output values (e.g., Figure 6 target controlled output values 628 of

[0231] device control model 610) and a plurality of estimated controlled output values; wherein, based on the plurality of input values (e.g., Figure 6 input values 630 of Figure 6 device control model 610), a plurality of estimated controlled output values are determined from the device control model (e.g.,

[0232] Step 918 is optional. Thus, the process can jump directly from step 916 to step 920, or the process includes step 918.

[0233] Step 918 includes: estimating a concentration value (e.g., estimated concentration value 620) using the membrane control model (e.g., Figure 6 membrane control model 618 of Figure 4 ), where the membrane control model can also be configured to estimate the concentration value based on a plurality of filter specifications associated with the plurality of membranes, as p disclosed. In one example, the plurality of filter specifications includes the water permeability l of the associated membrane, the solute concentration r, the membrane area that can be used to calculate the pressure. According to the operation of the membrane control model in step 916, step 918 can be performed in step 916. Therefore, step 918 is optional.

[0234] Step 920 includes determining an optimal water input flow (i.e., Figure 6 , which minimizes the optimal concentration value (e.g. Figure 6 The difference between the target controlled output value 628 in the embodiment and the estimated concentration value (eg, estimated concentration value 620); wherein the concentration value is determined by the membrane control model based on the optimal input water flow for each membrane in the plurality of membranes.

[0235] Step 922 includes: determining the optimal water flow rate (i.e., the optimal water flow rate) of each membrane in the plurality of membranes according to the optimal water flow rate (i.e., the optimal water flow rate) of each membrane in the plurality of membranes. Figure 6 Optimal input water flow 636 in the Figure 6 Optionally, the plurality of input values ​​of step 922 include a concentration factor and a water factor, wherein the concentration factor includes a first ratio of the feed stream to the outlet retentate stream, and the water factor includes a second ratio of the water stream to the feed stream.

[0236] Step 924 includes: determining the optimal water input flow (e.g., Figure 6 The optimal input water flow 636) adjusts multiple input values ​​(e.g., Figure 6 Depending on whether the water factor is included in the multiple input values ​​of step 922, step 924 may be performed in step 922. Therefore, step 924 is optional, that is, in some embodiments, the process jumps directly from step 922 to 926.

[0237] Step 926 includes: making multiple input values ​​(such as Figure 6 The input value 630) is output to the CFMF device, thereby reducing the second difference between the multiple target controlled output values ​​and the future multiple controlled output values.

[0238] Step 928 includes returning to step 902 to repeat the process. Therefore, this process may be iterative. However, it should be understood that, for example, if the exit criteria are met, or the optimization process 900 is not an iterative process in other aspects, the optimization process 900 may not need to be repeated. Therefore, step 928 is optional.

[0239] Optionally, the filtration process associated with the optimization process 900 is one of a crossflow UF process, an MF process, or an NF process. For example, the filtration process is a crossflow UF process, and / or the feed stream includes dairy raw materials, and the target component in the raw materials of the retentate stream is protein. Other example target components may include: whey protein concentrate, whey protein isolate, milk protein concentrate, milk protein isolate, micellar casein concentrate, micellar casein isolate, and lactoferrin. Preferably, the water associated with the optimization process 900 is deionized water.

[0240] Preferably, the optimization process 900 is performed using the system 600.

[0241] Figure 10 An example computing system for optimization is shown. Specifically, Figure 10 A block diagram of an embodiment of a computing system according to an example embodiment of the present disclosure is shown.

[0242] The computing system 1000 may be configured to perform any of the operations disclosed herein, for example, any of the operations discussed with reference to the functional units described in Figure 7 , 8 , 9A, and 9B. The computing system includes one or more computing devices 1002. The computing devices 1002 of the computing system 1000 include one or more processors 1004 and a memory 1006. The one or more processors 1004 may be any general-purpose processor configured to execute a set of instructions. For example, the one or more processors 1004 may be one or more general-purpose processors, one or more field-programmable gate arrays (FPGAs), and / or one or more application-specific integrated circuits (ASICs). In one embodiment, the one or more processors 1004 is a single processor. Alternatively, the one or more processors 1004 include multiple processors operatively connected. The one or more processors 1004 are communicatively coupled to the memory 1006 via an address bus 1008, a control bus 1010, and a data bus 1012. The memory 1006 may be a random-access memory (RAM), a read-only memory (ROM), a persistent storage device such as a hard disk drive, an erasable programmable read-only memory (EPROM), etc. The computing device 1002 also includes an I / O interface 1014 communicatively coupled to the address bus 1008, the control bus 1010, and the data bus 1012.

[0243] The memory 1006 may store information accessible by one or more processors 1004. For example, the memory 1006 (e.g., one or more non-transitory computer-readable storage media, storage devices) may include computer-readable instructions (not shown) executable by one or more processors 1004. The computer-readable instructions may be software written in any suitable programming language or implemented in hardware. Additionally, or alternatively, the computer-readable instructions may be executed in logically and / or physically independent threads on one or more processors 1004. For example, the memory 1006 may store instructions (not shown) that, when executed by one or more processors 1004, cause the one or more processors 1004 to perform operations such as any of the operations and functions configured for the computing system 1000, as described herein. Additionally, or optionally, the memory 1006 may store data (not shown) that can be fetched, received, accessed, written, manipulated, created, and / or stored. For example, the data may include data and / or information described herein related to Figures 1 to 9B associated data and / or information. In some implementations, the computing device 1002 may fetch data from and / or store data in one or more storage devices remote from the computing system 1000.

[0244] The computing system 1000 further includes a storage unit 1016, a network interface 1018, an input controller 1020, and an output controller 1022. The storage unit 1016, the network interface 1018, the input controller 1020, and the output controller 1022 are communicatively coupled to the central control unit via the I / O interface 1014.

[0245] The storage unit 1016 is a computer-readable medium, preferably a non-transitory computer-readable medium, which includes one or more programs, and the one or more programs include instructions that, when executed by one or more processors 1004, cause the computing system 1000 to perform the method steps of the present disclosure. Alternatively, the storage unit 1016 is a transitory computer-readable medium. The storage unit 1016 may be a persistent storage device, such as a hard disk drive, a cloud storage device, or any other suitable storage device.

[0246] The network interface 1018 may be a Wi-Fi module, a network interface card, a Bluetooth module, and / or any other suitable wired or wireless communication device. In one embodiment, the network interface 1018 is configured to connect to a network such as a local area network (LAN) or a wide area network (WAN), the Internet, or an intranet.

[0247] In this regard, it should be noted that the optimizations according to the present disclosure described above may to some extent involve the processing of input data and the generation of output data. Such input data processing and output data generation can be implemented in hardware or software. For example, specific electronic components can be used in a control module or similar or related circuits to implement the functions related to the optimizations according to the present disclosure described above. Alternatively, one or more processors operating according to instructions can implement the functions related to the optimizations according to the present disclosure described above. If this is the case, the instructions can be stored on one or more non-transitory processor-readable storage media (e.g., a magnetic disk or other storage media) or transmitted to one or more processors via one or more signals included in one or more carrier waves, all within the scope of the present disclosure.

[0248] The scope of the present disclosure is not limited by the specific embodiments described herein. In fact, in addition to the embodiments described herein, various other embodiments and modifications of the present disclosure will be apparent to those of ordinary skill in the art from the above description and the accompanying drawings. Accordingly, these other embodiments and modifications also fall within the scope of the present disclosure. In addition, although the present application is described in the context of at least one specific implementation in at least one specific environment for at least one specific purpose, those of ordinary skill in the art will recognize that its usefulness is not limited thereto, and the present application can be beneficially implemented in multiple numbers of environments for any number of purposes. Accordingly, the statements listed below should be interpreted in light of the full breadth and spirit of the present disclosure described herein. Sub-item statements of the present application 1. A system for optimizing a filtration process performed on a cross-flow membrane filtration (CFMF) device, the device separating a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the system comprising: An input interface communicatively coupled to the CFMF device, the input interface being configured to: Obtain a plurality of controlled output values related to the filtration process performed on the CFMF device, wherein the plurality of controlled output values include: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of solid components in the outlet product stream; and Obtain a plurality of disturbances related to the CFMF device; A control model of the CFMF device, wherein the control model estimates the controlled output values from input values based on the state of the control model and the plurality of disturbances; An optimizer including a model predictive control (MPC) layer, wherein the optimizer is configured to update the state of the control model based on the plurality of controlled output values, and wherein the MPC layer is configured to: Obtain a plurality of target controlled output values; and Determine a plurality of input values that optimize a first difference between a plurality of target controlled output values and a plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from a control model by the plurality of input values; and An output interface that is communicatively coupled to the CFMF device and is configured to output the plurality of input values to the CFMF device to thereby reduce a second difference between the plurality of target controlled output values and a plurality of future controlled output values. 2. The system according to statement 1, wherein the filtration process is one of a crossflow ultrafiltration (UF) process, a microfiltration (MF) process, or a nanofiltration (NF) process. 3. The system according to statement 2, wherein the filtration process is a crossflow UF process. 4. The system according to any of the preceding statements, wherein the feed stream comprises a dairy feedstock and the target component of the feedstock of the outlet retentate stream is protein. 5. The system according to any of the preceding statements, wherein the plurality of controlled output values are obtained from one or more sensors of the CFMF device. 6. The system according to any of statements 1 to 4, wherein the plurality of controlled output values are obtained from a soft sensor. 7. The system according to statement 6, wherein the soft sensor comprises a control model of the CFMF device such that the control model estimates the plurality of controlled output values. 8. The system according to any of the preceding statements, wherein a plurality of disturbances are obtained from one or more sensors of the CFMF device. 9. The system according to any of the preceding statements, wherein the control model comprises a state space model. 10. The system according to statement 9, wherein the state space model is a discrete-time model. 11. The system according to any of the preceding statements, wherein the control model is a linear control model. 12. The system according to any of the preceding statements, wherein the state of the control model is updated by a time-varying Kalman filter. 13. The system according to any of the preceding statements, wherein the optimizer further comprises a real-time optimization (RTO) layer that is configured to determine the plurality of target controlled output values obtained by the MPC layer. 14. The system according to statement 13, wherein the RTO layer is configured to: Determine a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determine the plurality of target controlled output values from the control model according to the plurality of optimal future input values. 15. The system according to any one of the preceding statements, wherein one or more constraints for optimizing the first cost function include one or more of a minimum concentration of a target component in the outlet retentate stream, a maximum concentration of the target component in the outlet retentate stream, and one or more input value limitations associated with the CFMF device. 16. The system according to any one of the preceding statements, wherein a plurality of optimal future input values for optimizing the first cost function are determined by an interior point algorithm for a non - linear problem. 17. The system according to any one of the preceding statements, wherein a plurality of input values for optimizing a first difference between a plurality of target controlled output values and a plurality of estimated controlled output values are determined by a quadratic programming solver. 18. The system according to statement 17, wherein the quadratic programming solver is one of an active set method, an interior point method, and a first - order gradient method. 19. The system according to any one of the preceding statements, wherein the plurality of input values include a concentration factor and a water factor, the concentration factor includes a first ratio of the feed stream to the outlet retentate stream, and the water factor includes a second ratio of the water stream to the feed stream. 20. The system according to statement 19, wherein the output interface is further configured to output an outlet retentate setpoint to the CFMF device according to the concentration factor. 21. The system according to statement 19 or 20, wherein the output interface is further configured to output a water flow setpoint to the CFMF device according to the water factor. 22. The system according to any one of the preceding statements, wherein the water is deionized water. 23. The system according to any one of the preceding statements, wherein the system further comprises: at least one programmable logic controller (PLC); and an edge computing device communicatively coupled to the at least one PLC and including one or more processors and a memory. 24. The system according to statement 23, wherein the PLC includes an input interface and an output interface. 25. The system according to statement 23 or 24, wherein the edge computing device includes a control model and an optimizer. 26. A method for optimizing a filtration process performed on a cross - flow membrane filtration (CFMF) device, the device dividing a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the method being performed on a control system communicatively coupled to the CFMF device, the method comprising: performing an optimization process, the optimization process comprising: obtaining a plurality of controlled output values associated with the filtration process performed on the CFMF device, wherein the plurality of controlled output values include: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of components in the outlet product stream; Obtain a plurality of interferences associated with the CFMF device; Obtain a control model of the CFMF device, wherein the control model estimates a controlled output value from an input value based on the state of the control model and a plurality of interferences; Update the state of the control model according to a plurality of controlled output values; Obtain a plurality of target controlled output values; Determine a plurality of input values that optimize a first difference between the plurality of target controlled output values and the plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from the control model by using the plurality of input values; and Output the plurality of input values to the CFMF device, thereby reducing a second difference between the plurality of target controlled output values and the plurality of future controlled output values. 27. The method according to statement 26, wherein the filtration process is one of a crossflow ultrafiltration (UF) process, a microfiltration (MF) process, or a nanofiltration (NF) process. 28. The method according to statement 27, wherein the filtration process is a crossflow UF process. 29. The method according to any one of statements 26 to 28, wherein the feed stream comprises a dairy raw material, and the target component of the raw material of the outlet retentate stream is protein. 30. The method according to any one of statements 26 to 29, wherein the step of obtaining the plurality of target controlled output values further comprises: Determine a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determine the plurality of target controlled output values from the control model according to the plurality of optimal future input values. 31. The method according to any one of statements 26 to 30, wherein the step of obtaining the plurality of controlled output values associated with the filtration process performed on the CFMF device further comprises: Obtain the plurality of controlled output values from one or more sensors of the CFMF device. 32. The method according to any one of statements 26 to 30, wherein the step of obtaining the plurality of controlled output values associated with the filtration process performed on the CFMF device further comprises: Obtain the plurality of controlled output values from the control model. 33. The method according to any one of statements 26 to 32, wherein the step of obtaining the plurality of interferences associated with the CFMF device further comprises: Obtain the plurality of interferences from one or more sensors of the CFMF device. 34. The method according to any one of statements 26 to 33, wherein the plurality of input values include a concentration factor and a water factor, the concentration factor includes a first ratio of the feed stream to the outlet retentate stream, and the water factor includes a second ratio of the water stream to the feed stream. 35. The method according to statement 34, wherein the step of outputting a plurality of input values to the CFMF device further comprises: Outputting an outlet retentate set point to the CFMF device according to a concentration factor. 36. The method according to any one of statements 34 or 35, wherein the step of outputting a plurality of input values to the CFMF device further comprises: Outputting a water flow set value to the CFMF device according to a water coefficient. 37. The method according to any one of statements 26 to 36 includes repeatedly performing an optimization process. 38. A computer-readable medium storing instructions, which when executed by a system composed of one or more processors, causes the system to perform any one of the steps in statements 26 to 37. 39. A crossflow membrane filtration device comprising the system according to any one of statements 1 to 25. 40. A system for optimizing water consumption of a crossflow membrane filtration (CFMF) device, the device comprising a plurality of membranes, each membrane performing a filtration process of dividing a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the system comprising: A membrane control model associated with the plurality of membranes, wherein the membrane control model is configured to: Estimate a concentration value based on an input water flow of each of the plurality of membranes, wherein the concentration value is related to an estimated concentration of a solid component in the outlet retentate stream; An optimization unit configured to: Obtain a target concentration value related to the solid component in the outlet retentate stream; and Determine an optimal input water flow of each of the plurality of membranes that minimizes the difference between the target concentration value and the estimated concentration value, wherein the estimated concentration value is determined from the membrane control model based on the optimal input water flow of each of the plurality of membranes; An output interface communicatively coupled to the CFMF device, the output interface being configured to cause the CFMF device to adjust the water flow of each of the plurality of membranes of the CFMF device according to the optimal input water flow determined for each of the plurality of membranes. 41. The system according to statement 40, wherein the filtration process is one of a crossflow ultrafiltration (UF) process, a microfiltration (MF) process, or a nanofiltration (NF) process. 42. The system according to statement 41, wherein the filtration process is a crossflow UF process. 43. The system according to any one of statements 40 to 42, wherein the feed stream comprises a dairy raw material, and the target component in the raw material of the retentate stream is protein. 44. The system according to any one of statements 40 to 43 further comprises a real-time optimizer (RTO), wherein the target concentration value is obtained from the RTO. 45. The system according to statement 44, wherein the RTO is configured to: Determine a plurality of optimal future input values that optimize a first cost function, where the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determine a target concentration value from a device control model associated with the CFMF device based on the plurality of optimal future input values, where the control model estimates the concentration value from the input values based on the state of the device control model. 46. The system according to statement 45, wherein one or more constraints for optimizing the first cost function include one or more of a minimum concentration of a target component in the outlet retentate stream, a maximum concentration of the target component in the outlet retentate stream, and one or more input value limitations associated with the CFMF device. 47. The system according to any one of statements 40 to 46, wherein the membrane control model is further configured to estimate the concentration value based on a plurality of filter specifications associated with a plurality of membranes. 48. The system according to statement 47, wherein each of the plurality of filter specifications includes the water permeability, solute permeability, reflection coefficient, and membrane area of the associated membrane. 49. The system according to any one of statements 40 to 48, wherein the membrane control model includes one of a physics-based model, a statistical model, or a machine learning model. 50. The system according to any one of statements 40 to 49, wherein the optimal input water flow for each membrane of the plurality of membranes is determined by a sequential quadratic programming optimization method. 51. The system according to any one of statements 40 to 49, wherein the optimal input water flow for each membrane of the plurality of membranes is determined by one of a direct search method, an extremum search method, or a trial-and-error method. 52. The system according to any one of statements 40 to 51, wherein the water is deionized water. 53. A method for optimizing the water consumption of a crossflow membrane filtration (CFMF) device, the device including a plurality of membranes, each membrane performing a filtration process that divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the method being executed on a control system communicatively coupled to the CFMF device, the method including: Performing an optimization process, the optimization process including: Obtaining a target concentration value associated with a solid component in the outlet retentate stream; Obtaining a membrane control model associated with the plurality of membranes, where the membrane control model estimates the concentration value based on the input water flow of each membrane of the plurality of membranes, where the concentration value is related to the estimated concentration of the solid component in the outlet retentate stream; Determining the optimal input water flow for each membrane of the plurality of membranes that minimizes the difference between the target concentration value and the estimated concentration value, where the estimated concentration value is determined from the membrane control model based on the optimal input water flow for each membrane of the plurality of membranes; and Cause the CFMF device to adjust the water flow of each of the multiple membranes of the CFMF device according to the optimal input water flow determined for each of the multiple membranes. 54. The method according to statement 53, wherein the filtration process is one of a crossflow ultrafiltration (UF) process, a microfiltration (MF) process, or a nanofiltration (NF) process. 55. The method according to statement 54, wherein the filtration process is a crossflow UF process. 56. The method according to statement 55, wherein the feed stream comprises a dairy feedstock and the target component in the feed of the outlet retentate stream is protein. 57. The method according to any one of statements 54 to 56, wherein the step of obtaining the target concentration value further comprises: Determining a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determining the target concentration value from a device control model associated with the CFMF device according to the plurality of optimal future input values, wherein the control model estimates the concentration value from the input values according to the state of the device control model. 58. The method according to statement 57, wherein one or more of the constraints for optimizing the first cost function include one or more of a minimum concentration of the target component in the outlet retentate stream, a maximum concentration of the target component in the outlet retentate stream, and one or more input value limitations associated with the CFMF device. 59. The method according to any one of statements 53 to 58, wherein the membrane control model estimates the concentration value based on a plurality of filter specifications associated with the multiple membranes. 60. The method according to statement 59, wherein each of the plurality of filter specifications includes the water permeability, solute permeability, reflection coefficient, and membrane area of the associated membrane. 61. The method according to any one of statements 53 to 60 further comprises repeatedly performing the optimization process. 62. A computer-readable medium storing instructions, wherein when the instructions are executed by a system comprising one or more processors, the system performs any of the steps in statements 53 to 61. 63. A crossflow membrane filtration device comprising the system according to any one of statements 40 to 52. 64. A system for optimizing water usage of a crossflow membrane filtration (CFMF) device, the device comprising a plurality of membranes, each membrane performing a filtration process that divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the system comprising: An input interface communicatively coupled to the CFMF device and configured to: Obtain a plurality of controlled output values related to a filtration process performed on a CFMF device, wherein the plurality of controlled output values includes: (i) a product concentration value indicating the concentration of a target component of the raw material in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of solid components in the outlet product stream; and Obtain a plurality of disturbances related to the CFMF device; A device control model related to the CFMF device, wherein the device control model estimates the controlled output values from the input values based on the state of the device control model and the plurality of disturbances; A membrane control model related to a plurality of membranes, wherein the membrane model is configured to estimate a concentration value based on the input water flow of each of the plurality of membranes, and the concentration value is related to the estimated concentration of the solid component in the outlet retentate stream; An optimizer including a model predictive control (MPC) layer and a water flow optimization (WFO) layer, the optimizer being configured to update the state of the device control model based on the plurality of controlled output values, and the MPC layer being configured to: Obtain a plurality of target controlled output values; and Determine a plurality of input values that optimize a first difference between the plurality of target controlled output values and the plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from the device control model by the plurality of input values; and Wherein the WFO layer is configured to: Obtain a target concentration value related to the solid component in the outlet retentate stream; Determine the optimal input water flow of each of the plurality of membranes that minimizes the difference between the target concentration value and the estimated concentration value, wherein the estimated concentration value is determined from the membrane control model based on the optimal input water flow of each of the plurality of membranes; and Adjust the plurality of input values determined by the MPC layer according to the optimal input water flow of each of the plurality of membranes; An output interface communicatively coupled to the CFMF device, the output interface being configured to output the plurality of input values to the CFMF device, thereby reducing a second difference between the plurality of target controlled output values and the plurality of future controlled output values. 65. The system according to statement 64, wherein the filtration process is one of a crossflow ultrafiltration (UF) process, a microfiltration (MF) process, or a nanofiltration (NF) process. 66. The system according to statement 65, wherein the filtration process is a crossflow UF process. 67. The system according to any one of statements 64 to 66, wherein the raw material stream includes dairy raw material, and the target component of the raw material in the retentate stream is protein. 68. The system according to any one of statements 64 to 67, wherein the optimizer further includes a real-time optimization (RTO) layer, the real-time optimization (RTO) layer being configured to determine the plurality of target controlled output values obtained by the MPC layer. 69. The system according to statement 68, wherein the RTO layer is configured to: Determine a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determine a plurality of target controlled output values from the device control model according to the plurality of optimal future input values. 70. The system according to statement 69, wherein the RTO layer is further configured to: Determine a target concentration value from the device control model according to the plurality of optimal future input values. 71. The system according to any one of statements 69 or 70, wherein one or more constraints for optimizing the first cost function include one or more of a minimum concentration of a target component in the outlet retentate stream, a maximum concentration of the target component in the outlet retentate stream, and one or more input value limitations associated with the CFMF device. 72. The system according to any one of statements 64 to 71, wherein the plurality of input values include a concentration factor and a water factor, the concentration factor includes a first ratio of the feed stream to the outlet retentate stream, and the water factor includes a second ratio of the water stream to the feed stream. 73. The system according to statement 72, wherein the output interface is further configured to output the outlet retentate setpoint to the CFMF device according to the concentration factor. 74. The system according to any one of statements 72 or 73, wherein the output interface is further configured to output the water stream setpoint to the CFMF device according to the water factor. 75. The system according to any one of statements 72 to 74, wherein the water stream optimization unit is further configured to adjust the water factor of the plurality of input values according to the optimal input water stream of each of the plurality of membranes. 76. The system according to any one of statements 64 to 75, wherein the membrane control model is further configured to estimate the concentration value according to a plurality of filter specifications associated with the plurality of membranes. 77. The system according to statement 76, wherein each of the plurality of filter specifications includes the water permeability, solute permeability, reflection coefficient, and membrane area of the associated membrane. 78. The system according to any one of statements 64 to 77, wherein the water is deionized water. 79. A method for optimizing a filtration process on a crossflow membrane filtration (CFMF) device, the device including a plurality of membranes, each membrane performing a filtration process of dividing a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the method being executed on a control system communicatively coupled to the CFMF device, the method including: Performing an optimization process, the optimization process including: Obtain a plurality of controlled output values associated with a filtration process performed on a CFMF device, wherein the plurality of controlled output values includes: (i) a product concentration value indicating the concentration of a target component of the feedstock in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of solid components in the outlet product stream; Obtain a plurality of disturbances associated with the CFMF device; Obtain a device control model of the CFMF device, wherein the device control model estimates the controlled output values from input values based on the state of the device control model and the plurality of disturbances; Update the state of the device control model according to the plurality of controlled output values; Obtain a plurality of target controlled output values; Determine a plurality of input values that optimize a first difference between the plurality of target controlled output values and the plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from the device control model by the plurality of input values; Obtain a membrane control model associated with the plurality of membranes, wherein the membrane model estimates a concentration value based on the input water flow of each of the plurality of membranes, and wherein the concentration value is related to the estimated concentration of the solid components in the outlet retentate stream; Determine the optimal input water flow of each of the plurality of membranes that minimizes the difference between the target concentration value and the estimated concentration value, wherein the estimated concentration value is determined from the membrane control model based on the optimal input water flow of each of the plurality of membranes; Adjust the plurality of input values according to the optimal input water flow of each of the plurality of membranes; and Output the plurality of input values to the CFMF device, thereby reducing a second difference between the plurality of target controlled output values and the plurality of future controlled output values. 80. The method according to statement 79, wherein the filtration process is one of a crossflow ultrafiltration (UF) process, a microfiltration (MF) process, or a nanofiltration (NF) process. 81. The method according to statement 81, wherein the filtration process is a crossflow UF process. 82. The method according to any one of statements 79 to 81, wherein the feedstock stream includes a dairy feedstock, and the target component of the feedstock in the retentate stream is protein. 83. The method according to any one of statements 79 to 82, wherein the step of obtaining the plurality of target controlled output values further includes: Determine a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determine the plurality of target controlled output values from the device control model according to the plurality of optimal future input values. 84. The method according to statement 83, wherein the step of obtaining the target concentration value further includes: Determine a target concentration value from a device control model based on multiple optimal future input values. 85. The method according to any one of claims 79 to 84, wherein the multiple input values include a concentration factor and a water factor, the concentration factor includes a first ratio of a feed stream to an outlet retention stream, and the water factor includes a second ratio of a water stream to the feed stream. 86. The method according to claim 85, wherein the step of adjusting the multiple input values includes: Adjusting the water factor of the multiple input values according to the optimal input water flow of each of the multiple membranes. 87. The method according to any one of claims 79 to 86, wherein the membrane control model is further configured to estimate a concentration value based on multiple filter specifications associated with the multiple membranes. 88. The method according to claim 87, wherein each of the multiple filter specifications includes the water permeability, solute permeability, reflection coefficient, and membrane area of the associated membrane. 89. The method according to any one of claims 79 to 88, wherein the water is deionized water. 90. The method according to any one of claims 79 to 89 further includes repeatedly performing an optimization process. 91. A computer-readable medium storing instructions, which when executed by a system composed of one or more processors, causes the system to perform any one of the steps in claims 79 to 90. 92. A crossflow membrane filtration device, comprising a system according to any one of claims 64 to 78.

Claims

1. A system for optimizing a filtration process performed on a cross-flow membrane filtration (CFMF) device, the device separating a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the system comprising: An input interface, wherein the input interface is communicatively coupled to the CFMF device and is configured to: Obtain a plurality of controlled output values related to the filtration process performed on the CFMF device, wherein the plurality of controlled output values includes: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentration of a plurality of solid components in the outlet product stream; and Obtain a plurality of disturbances related to the CFMF device; A control model of the CFMF device, wherein the control model estimates a controlled output value from an input value based on a state of the control model and the plurality of disturbances; An optimizer including a model predictive control (MPC) layer, wherein the optimizer is configured to update the state of the control model based on the plurality of controlled output values, and wherein the MPC layer is configured to: Obtain a plurality of target controlled output values; and Determine a plurality of input values that optimize a first difference between the plurality of target controlled output values and a plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from the control model using the plurality of input values; and An output interface, wherein the output interface is communicatively coupled to the CFMF device and is configured to output the plurality of input values to the CFMF device, thereby reducing a second difference between the plurality of target controlled output values and a plurality of future controlled output values.

2. The system according to claim 1, wherein, The filtration process is a cross-flow ultrafiltration process.

3. The system according to claim 1 or 2, wherein, The feed stream includes dairy products, particularly whey, and the target component in the feed of the outlet retentate stream is protein.

4. The system according to any one of the preceding claims, wherein, The plurality of controlled output values are obtained from one or more sensors of the CFMF device.

5. The system according to any one of the preceding claims, wherein, The optimizer further includes a real-time optimization (RTO) layer, the RTO layer being configured to determine the plurality of target controlled output values obtained by the MPC layer.

6. The system according to claim 5, wherein, The RTO layer is configured to: Determine a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates a future efficiency of the filtration process based on one or more constraints; and Determine the plurality of target controlled output values from the control model based on the plurality of optimal future input values.

7. The system according to any one of the preceding claims, wherein, The plurality of input values include a concentration factor and a water factor, the concentration factor including a first ratio of the feed stream to the outlet retentate stream, and the water factor including a second ratio of the water stream to the feed stream.

8. The system according to any one of the preceding claims, further comprising: At least one programmable logic controller (PLC); And An edge computing device communicatively coupled to the at least one PLC and including one or more processors and a memory; Wherein the PLC includes the input interface and the output interface, and the edge computing device includes the control model and the optimizer.

9. A cross-flow membrane filtration device comprising the system according to any one of claims 1 to 8.

10. A method for optimizing a filtration process performed on a cross-flow membrane filtration (CFMF) device that divides a feed stream and a water stream into an outlet retentate stream and an outlet permeate stream, the method being performed on a control system communicatively coupled to the CFMF device, the method comprising: Obtaining a plurality of controlled output values associated with the filtration process performed on the CFMF device, wherein the plurality of controlled output values includes: (i) a product concentration value indicating the concentration of a target component of the feed in the outlet product stream, and (ii) a total concentration value indicating the concentrations of a plurality of components in the outlet product stream; Obtaining a plurality of disturbances associated with the CFMF device; Obtaining a control model of the CFMF device, wherein the control model estimates the controlled output values from input values based on the state of the control model and the plurality of disturbances; Updating the state of the control model based on the plurality of controlled output values; Obtaining a plurality of target controlled output values; Determining a plurality of input values that optimize a first difference between the plurality of target controlled output values and a plurality of estimated controlled output values, wherein the plurality of estimated controlled output values are determined from the control model using the plurality of input values; and Causing the plurality of input values to be output to the CFMF device so as to reduce a second difference between the plurality of target controlled output values and a plurality of future controlled output values.

11. The method according to claim 10, wherein, The filtration process is a cross-flow ultrafiltration process.

12. The method according to claim 10 or 11, wherein The feed stream includes dairy products, particularly whey, and the target component in the feed of the retentate stream is protein.

13. According to the method as claimed in any one of claims 10 to 12, wherein, The step of obtaining the plurality of target controlled output values further includes: Determining a plurality of optimal future input values that optimize a first cost function, wherein the first cost function estimates the future efficiency of the filtration process based on one or more constraints; and Determining the plurality of target controlled output values from the control model based on the plurality of optimal future input values.

14. The method according to any one of claims 10 to 13, wherein The plurality of input values includes a concentration factor and a water factor, wherein the concentration factor includes a first ratio of the feed stream to the outlet retentate stream, and the water factor includes a second ratio of the water stream to the feed stream.

15. A computer-readable medium, wherein, The medium stores a plurality of instructions that, when executed by a system including one or more processors, cause the system to perform the steps of any one of claims 10 to 14.

Images