Method for performing tangential flow filtration on a biomolecular solution

Abstract

The present invention relates to a method for performing tangential flow filtration on a biomolecular solution as part of a biomolecule manufacturing process, more particularly a method for performing single-pass tangential flow filtration on a biomolecular solution, wherein a tangential flow filtration apparatus (1) is provided, the tangential flow filtration apparatus (1) comprises a tangential flow filtration module (2), tangential flow filtration is performed within the tangential flow filtration module (2), the tangential flow filtration module (2) comprises a filter (3), a supply fluid input line (4) for the biomolecular solution as a supply medium, and a retaining fluid output line (6) for the retaining fluid. The present invention relates to a method comprising a tangential flow filter (1) comprising a filtrate output fluid line (7) for the filtrate and preferably a buffer input fluid line (5) for the buffer, wherein the tangential flow filter (1) comprises at least two actuators (8) that affect the tangential flow filter, the tangential flow filter (1) comprises a control module (10), the control module (10) controls the tangential flow filter by providing control signals to the actuators (8) and by measuring at least two measurement variables (11) that describe the tangential flow filter, and the control module (10) controls the tangential flow filter using a process control unit. It is proposed that the process control unit is a model prediction control unit (13), the control module (10) inputs the measurement variables (11) into a model as input variables (14), and derives at least two operation variables (15) as control signals for the actuators (8).

Description

Technical Field

[0001] The present invention relates to a method for performing tangential flow filtration on a biomolecule solution described in the preamble (general part) of claim 1, a control module described in claim 15, and a tangential flow filtration device described in claim 16.

[0002] The present teachings relate to filtration in a biotechnology environment, particularly single-pass filtration. Single-pass tangential flow filtration (SPTFF) is a technique used in bioprocesses for separating and purifying biomolecules such as proteins from a mixture. It is increasingly being used in the manufacture of biopharmaceuticals and other high-value biological products.

[0003] In SPTFF, a feed medium containing the biomolecule to be purified is passed under pressure through a filtration membrane. SPTFF can be carried out by a tangential flow filtration module equipped with a filter, usually a membrane. On the first side of the membrane and the first side of the module, the feed medium and optionally a buffer are supplied to the module. On the second side of the module, passing tangentially along the membrane and partially through the membrane, the feed medium is separated into a retentate and a filtrate. The biomolecule can be part of either the retentate or the filtrate depending on the particular process. SPTFF is well-suited for the concentration and purification of various components including proteins such as monoclonal antibodies. In applications that require buffer exchange or desalting of the solution, a buffer, sometimes called a diafiltration medium, is added to the feed medium or the retentate. This process is typically called diafiltration.

[0004] One of the main factors in the efficiency of SPTFF is the transmembrane pressure difference. Other related process variables are the concentration factor and the diafiltration factor.

[0005] Known single-pass tangential flow filtration systems may be controlled via PID controllers, for example, three PID controllers. PID controllers are widely used for many types of control and usually work well in relatively simple systems. Tangential flow filtration of biomolecules often has constant setpoints that are set at the start of filtration and are usually not changed during filtration. They are usually not affected by sudden changes or unexpected disturbances. The control parameters of PID controllers can be adapted by a skilled user depending on the specific application, especially the specific feed medium.

[0006] However, due to variability between different customers and batches, as well as slow system changes due to film contamination and / or degradation, PID controllers often require at least some user supervision and skilled operators for PID adaptation. While such controls are not sufficient to resolve the dynamic dependencies between variables from a model theory perspective, they have so far been used as a good approximation to support operators.

[0007] A model prediction control unit or model prediction controller (MPC) is a control unit that uses a model to predict at least one future state and the corresponding output of the controlled system, and takes this state and output into account when deriving control signals to affect the operational variables of the controlled system.

[0008] Model predictive control units are used in advanced control projects in refineries, petrochemicals, chemicals, and several other industries. They are typically used in complex control systems with high volatility of control variables and unknown disturbance variables. Model predictive control units offer high flexibility to highly dynamic systems with coupled process variables, and can accommodate economic and technological changes, and therefore various operating points.

[0009] None of these are typical for filtration in the bioprocess field. Therefore, control variables and process parameters are generally kept relatively constant.

[0010] Improving control methods in known prior art is a challenge.

[0011] This invention is based on the problem of improving known methods to achieve further optimization with respect to a specified problem.

[0012] The above-mentioned objective is achieved by the features of the characterization section of claim 1.

[0013] The primary realization of this invention is that a model predictive control unit can be used in biotechnological filtration to address inter-batch and inter-customer variations. As a result, less user intervention is required to adapt the control unit to new supply media or other changes in the system. Furthermore, filter, particularly membrane, fouling and other degradation effects are well addressed by the model predictive control unit without user intervention. Moreover, when one process variable is intentionally changed, the filtration process can be controlled such that only the value of this desired process variable changes, with little effect on the values ​​of other process variables. As an added benefit, the model can be used to determine the state of the filter and plan maintenance.

[0014] The proposed teaching relates to a method for performing tangential flow filtration on a biomolecular solution as part of a biomolecule manufacturing process, particularly a method for performing single-pass tangential flow filtration on a biomolecular solution, wherein a tangential flow filtration apparatus is provided, the tangential flow filtration apparatus comprises a tangential flow filtration module, tangential flow filtration is performed within the tangential flow filtration module, the tangential flow filtration module comprises a filter, a feed fluid input line for a biomolecular solution as a feed medium, a retain fluid output line for a retain fluid, a filtrate output fluid line for a filtrate, and preferably a buffer input fluid line for a buffer, the tangential flow filtration apparatus comprises at least two actuators that affect the tangential flow filtration, the tangential flow filtration apparatus comprises a control module, the control module controls the tangential flow filtration by providing control signals to the actuators and measuring at least two metric variables that describe the tangential flow filtration, and the control module controls the tangential flow filtration using a process control unit.

[0015] In detail, it is proposed that the process control unit is a model prediction control unit, and that the control module inputs measured variables into the model as input variables and derives at least two operational variables as control signals for the actuators.

[0016] Claim 2 describes a preferred actuator for a tangential flow filtration device, and Claim 3 describes a preferred operating variable.

[0017] In the embodiment described in claim 4, the control variable is set by the user and then preferably kept constant. This may apply to all control variables and highlights the simplicity of the process control unit with respect to changes during the process. Claim 5 specifies preferred control variables.

[0018] Embodiments of claims 6 and 7 relate to a model of the model predictive control unit, in this case a state-space model. The number of states in the model may be in a smaller range (claim 7). The selected states describe the physical properties of the filtration selected to achieve a less complex representation of the tangential flow filtration module, based on the external inputs and outputs of the tangential flow filtration module as well as possible measurements.

[0019] According to claim 8, the moving horizon principle can be applied to ensure a rapid response in the event of unexpected changes, and according to claim 9, the model predictive control unit may include limit values ​​for the operational and / or predictive variables. By including the hard and soft limits of the tangential flow filtration module in the model predictive control unit, fitting to those limits is ensured.

[0020] Claim 10 relates to an embodiment of a cost function that has been found to be suitable for stable control and good adaptation to new supply media.

[0021] Tangential flow filtration modules typically do not include sensors within the module. Furthermore, even if sensors are present, they may not be able to measure all relevant parameters. Also, flow sensors in particular are susceptible to noise in their signals. Some of these problems can be mitigated by using an estimator (Claim 11). An estimator can be used to estimate any process variables inside or outside the tangential flow filtration module. The estimator can also estimate similarly measured process variables and, in particular, derive filter degradation from changes in those variables.

[0022] In the embodiment described in claim 12, the control module may select one of a plurality of state-space models at the start of filtration in order to further adapt the model predictive control unit to the current supply medium and system configuration.

[0023] The state space model may also change over time to adapt to filter degradation (Claim 13).

[0024] Claim 14 relates to predictive maintenance based on a model predictive control unit. By analyzing how the model predictive control unit changes over time, a suitable time for maintenance can be derived.

[0025] Another teaching according to Claim 15, which is equally important, relates to a method for performing tangential flow filtration on a biomolecule solution, particularly single-pass tangential flow filtration on a biomolecule solution, as part of a biomolecule production process, and particularly to a control module configured to use a state space model predictive control unit in a particularly proposed method.

[0026] All explanations given regarding the proposed method are fully applicable.

[0027] Another teaching according to Claim 16, which is equally important, relates to a tangential flow filtration device configured to perform the proposed method.

[0028] All explanations given regarding the proposed method and the proposed control module are fully applicable.

[0029] Hereinafter, embodiments of the present invention will be described in relation to the drawings.

Brief Description of the Drawings

[0030] [Figure 1] It is a schematic diagram of a tangential flow filtration device. [Figure 2] It is a diagram showing a simple overview of a process control unit. [Figure 3] It is a diagram showing a more complex overview of a model predictive control unit and its use.

[0031] Figure 1 shows a simplified tangential flow filtration apparatus 1 having a tangential flow filtration module 2. Here, and preferably, tangential flow filtration is single-pass tangential flow filtration, meaning there is no recirculation of the retaining fluid flow back to the feed medium. Filtration may be continuous and / or concentrated and / or diafiltration. The feed medium and buffer are preferably added continuously.

[0032] The tangential flow filtration module 2 may be a disposable module or may comprise multiple components, such as a disposable filter 3 and a tube set. The filter 3 may comprise a membrane or may be a membrane. As shown in Figure 1, the tangential flow filtration module 2 may have one or two inlets, one connected to the supply fluid input fluid line 4 and optionally the other connected to the buffer input fluid line 5. The tangential flow filtration module 2 may further comprise two outlets connected to the retained fluid output fluid line 6 and the filtrate output fluid line 7. Actuators 8, in particular pumps or valves, and / or sensors 9 may be present in any one or any combination of the fluid lines.

[0033] Buffer exchange can be performed using the illustrated tangential flow filtration module 2. Here, there are two membranes that divide the tangential flow filtration module 2 into three chambers or channels. In the illustrated embodiment, the upper chamber for the buffer does not have an output fluid line so that the buffer can only exit the module through filter 3. The illustrated lower chamber for the filtrate does not have an input fluid line. Naturally, the terms upper and lower are related to the drawing only and do not necessarily relate to the actual orientation.

[0034] As part of the biomolecule manufacturing process, methods for performing tangential flow filtration on biomolecular solutions, particularly methods for performing single-pass tangential flow filtration on biomolecular solutions, have been proposed.

[0035] Biomolecules can be small molecules, particularly proteins such as antibodies, or other molecules produced by bioprocesses, especially by living cells. Filtration during biomolecule production typically needs to meet regulatory standards during execution and control. To ensure the quality of biomolecules, they must be exposed to a controlled environment without leaving established boundaries.

[0036] Here, as shown in Figure 1, a tangential flow filtration apparatus 1 is provided. The tangential flow filtration apparatus 1 comprises a tangential flow filtration module 2. Tangential flow filtration is performed inside the tangential flow filtration module 2. Typically, the feed medium is pressurized and pumped into the tangential flow filtration module 2. The resulting intermembrane pressure difference allows the liquid and small particles to pass through the membrane, while larger particles cannot. This separates the feed medium into a retaining liquid and a filtrate. Preferably, if a buffer solution is applied, the liquid in the feed medium can be replaced with the buffer solution to adjust the salt concentration for subsequent process steps, such as chromatography.

[0037] The tangential flow filtration module 2 comprises a filter 3, a feed fluid input line 4 for a biomolecular solution as a feed medium, a retain fluid output line 6 for a retain fluid, a filtrate output fluid line 7 for a filtrate, and here, and preferably, a buffer input fluid line 5 for a buffer. Here, the filter 3 is a membrane inside the tangential flow filtration module 2.

[0038] The tangential flow filtration device 1 includes at least two actuators 8 that affect tangential flow filtration, in this case, three pumps.

[0039] The tangential flow filtration apparatus 1 comprises a control module 10. The control module 10 controls the tangential flow filtration by providing control signals to the actuator 8 and by measuring at least two measurement variables 11 that describe the tangential flow filtration. The control signals may be direct control signals, such as a PWM voltage applied to the pump. The control signals may also be indirect control signals, such as setpoints 12 of a further control system. The same applies to the measurement variables 11. In particular, a general process control system may be provided, and the control module 10 may communicate with that general process control system to receive the measurement variables 11 and transmit control signals. The control module 10 controls the tangential flow filtration using a process control unit.

[0040] It is essential that the process control unit is a model prediction control unit 13. The control module 10 inputs the measured variable 11 into the model as an input variable 14 and derives at least two operational variables 15 as control signals for the actuator 8. Here, and preferably, the process control operates in real time.

[0041] Figure 2 shows an overall overview of the process control unit. Figure 2 is intended to provide an overview of the terminology used. On the left side, the input variables 14 and setpoints 12 are shown. The input variables 14 are everything supplied to the model, and in particular, the setpoints 12 are also setpoints 12 for some of the input variables 14. The current value of the manipulated variable 16 is fed back to the input variables 14 to derive the error 17, as is known in the art. Note that in Figure 3, the error 17 is derived internally within the model predictive control unit 13, if necessary. The error 17 is used as the input value for the input variable 14. The control module 10 derives the value of the manipulated variable 15 and applies the value of the manipulated variable 15 to the tangential flow filtration module 2 by controlling the actuator 8. The current value of the manipulated variable 16 is shown as an output in Figure 2. In a preferred example, it should be understood that the control variable itself is not measured but estimated or derived from the measured variable 11.

[0042] Figure 3 shows a more detailed description of the preferred model predictive control unit 13. The model receives the setpoint 12, feedback of the last output control variable 15 if the model does not save the output control variable 15, the measured variable 11, and / or an unexplained estimate 18 from the estimator 19.

[0043] According to one embodiment, as already partially described, the tangential flow filter 1 is proposed to include at least one actuator 8 acting on the supply fluid input fluid line 4, in particular a pump, and / or at least one actuator 8 acting on the retaining fluid output fluid line 6, and / or at least one actuator 8 acting on the buffer input fluid line 5. Additionally or alternatively, the tangential flow filter 1 includes at least one sensor 9 in the supply fluid input fluid line 4, in particular at least one pressure sensor 9 and / or at least one flow sensor 9, and / or at least one sensor 9 in the retaining fluid output fluid line 6, in particular at least one pressure sensor 9 and / or at least one flow sensor 9, and / or at least one sensor 9 in the buffer input fluid line 5, in particular at least one pressure sensor 9 and / or at least one flow sensor 9. Additionally or alternatively, the actuator 8, in particular a pump, and / or the flow sensor 9 may be located in the filtrate output fluid line 7.

[0044] The control variable 15 may include the speed of the supply fluid pump and / or the retention fluid pump and / or the buffer pump. The control variable 15 may also include the pressure of the supply fluid and / or retention fluid and / or buffer, and / or the volumetric flow rate of the supply fluid and / or retention fluid and / or buffer. The pressure and / or volumetric flow rate can be indirectly controlled, for example, via a high-speed PID controller that receives signals from the set pressure and pressure sensor 9 and controls the speed of the respective pumps.

[0045] Here, and preferably, the operating variable 15 is the speed of the supply fluid pump, the retention fluid pump, and the buffer pump. A filtrate pump may be present instead of, or in addition to, the retention fluid pump. Everything described for the retention fluid pump may also apply to the filtrate pump, either instead or in addition.

[0046] The number of control variables may rather be small. Here, and preferably, the control variables are the intermembrane pressure difference, the concentration factor, and the diafiltration volume. The term “intermembrane pressure difference” also generally includes “interfilter pressure difference.” Generally, the process control unit preferably receives, in particular from the user, setpoints 12 for at least one, preferably at least two, more preferably at least three, and / or up to five, more preferably up to four, control variables controlled by the process control unit. The user can input the setpoints 12 into the general control system described above. The specified amount of control variables is a preferred embodiment independent of the source of the setpoints 12. Additionally or alternatively, the control variables may include conductivity and / or pH value and / or any concentration.

[0047] The control variable may remain constant during filtration. Here, and preferably, the control variable is kept constant until the control module 10 receives a new setpoint 12 from the user.

[0048] There may be cases where at least one, preferably all, control variables do not linearly depend on the input variable 14 and / or the manipulated variable 15, and / or the number of control variables is equal to the number of actuators 8.

[0049] According to one embodiment, it is proposed that the control variables include, or consist of, an intermembrane pressure difference, particularly the average intermembrane pressure difference, and / or a concentration factor, particularly the ratio of the supply liquid volume flow rate to the retained liquid volume flow rate, and / or a diafiltration ratio, particularly the ratio of the buffer volume flow rate to the supply liquid volume flow rate.

[0050] Turning to the model, it is preferable that the model prediction control unit 13 includes a state-space model 20 having state variables. Preferably, the state-space model 20 describes the effect of the actuator 8 on the state of the tangential flow filtration module 2.

[0051] An exemplary state-space model 20 is: Δx / Δt = Ax + Bu and y = C(x) The formula may also be in the form shown, where Δ represents the derivative of the prediction of the model prediction control unit 13 in a continuous or numerically approximate manner, or the difference between cycles. x is a vector of state variables. A is a matrix representing the behavior and interdependence of the state variables. B represents the influence of the instrument variable 15 named u on the state variables.

[0052] y represents a control variable, and C is a transformation matrix that transforms state variables into control variables. It is preferable that at least one, and in particular each, control variable is derived from at least two state variables. This usually means that the control variable does not linearly depend on any one of the state variables. The state-space model 20 may also be time-invariant.

[0053] According to one embodiment, the state-space model 20 has at least three, preferably at least six, more preferably at least nine, more preferably at least twelve, and / or up to 30, preferably up to 24, more preferably up to 15, and more preferably up to 12 state variables, and / or the state-space model 20 includes, in particular, one first-order differential equation for each state variable, preferably exactly one for each state variable, and / or the state-space model 20 does not include any differential equations of order higher than 1, preferably the state variables include the pump speed, in particular the rotational speed, and / or the flow rate of the feed fluid pump and / or the retention fluid pump and / or the buffer pump, and / or the volumetric flow rate of the feed fluid and / or the retention fluid and / or the buffer, and / or the influence of one or more, preferably all, pressures on the feed fluid pressure and retention fluid pressure, in particular the average feed fluid pressure and the average retention fluid pressure.

[0054] Here, the 12 state variables may be three pump speeds, three volumetric flow rates derived from the pump speeds, and six pressure variables, each derived from the volumetric flow rates, describing the effect of the pressure in each channel on the feed fluid pressure and retain fluid pressure within the tangential flow filtration module 2. Although the pumps are located outside the tangential flow filtration module 2, the pressure is derived here, and preferably, by modeling the effect of the pumps on the pressure inside the tangential flow filtration module 2, since the intermembrane pressure difference is preferably the average pressure difference between the feed fluid side and the retain fluid side. The operational variable 15 is the pump speed. This state-space model 20 is a low-complexity representation. By superimposing the pressure effects of different pumps, the model does not require high computational effort for numerical resolution. The control variables can then be derived from the state variables.

[0055] In the prediction, the model prediction control unit 13 may, in the current cycle, predict the control variable 15 for several cycles of the prediction horizon based on the current state; the first predicted control variable 15 may be used as the control signal for the current cycle; a new prediction cycle may be started for the next cycle; and preferably, during the prediction, the control variable 15 may be changed only during a control horizon of fewer cycles than the prediction horizon. This embodiment is an embodiment of the moving horizon principle.

[0056] The model prediction control unit 13 may also preferably include limit values ​​for the predictor variables and / or manipulated variables 15. These limit values ​​may include physical laws and physical limitations that the tangential flow filtration module 2 faces. The limit values ​​may also include soft limits defined by the user, for example, for regulatory or economic purposes.

[0057] To derive the value of the manipulated variable 15, the model prediction control unit 13 can use the optimizer 21 to derive the manipulated variable 15.

[0058] Preferably, the optimizer 21 uses a cost function 22. The cost function 22 may include a term that depends on the deviation between the setpoint 12 and the current value in one or more forecast cycles, preferably the squared deviation of the calculated difference between the forecast cycles, and in particular, between the setpoint 12 and the current value of the instrumental variable 16 for each forecast cycle. This term ensures that the instrumental variable 15 quickly reaches the setpoint 12.

[0059] The cost function 22 may further include a term that depends on the change in the instrumental variable 15 during one or more forecast cycles, preferably the squared deviation of the change in the instrumental variable 15 over the forecast cycles, particularly across all forecast cycles. This term ensures a low overshoot.

[0060] The cost function 22 may further include a term that depends on the deviation of the state variable between at least two prediction cycles, preferably the squared deviation of the change in the system state during the prediction cycle. This term ensures that no abrupt changes that are physically impossible or that would stress the tangential flow filter 1 or the biomolecular solution are implemented.

[0061] The cost function 22 does not have to depend on the timing of the prediction cycle. The time-invariant cost function 22 is well suited to models that have good predictions over longer periods and have little to no unexpected deviations. This is preferred here.

[0062] As already mentioned above, the estimator 19 can be used to estimate at least one input variable 14. Preferably, the estimated variable is input into the state-space model 20.

[0063] Furthermore, by comparing the predictor variables with the measured variables 11, changes in the filter 3 can be detected. This embodiment is of interest for predictive maintenance. By monitoring the behavior of the state-space model 20 and changes in this behavior as a whole, conclusions can be drawn regarding the state of the filter 3, particularly the degradation of the film.

[0064] The control module 10 can also modify the state-space model 20 using the detected changes. By automatically adapting the state-space model 20, especially the time-invariant state-space model 20, the state-space model 20 can be kept up-to-date with the slow changes in the filter 3. A use case for the estimator 19 may be to estimate the measured variable 11 using a model specifically designed to resemble the state-space model 20. If the estimate begins to deviate from the measured value, this may indicate a change in the filter 3.

[0065] A further use of estimator 19 is to reduce noise in the measured variable 11. Flow data, in particular, can contain a lot of noise. As an alternative to a filter, estimator 19 can derive good estimates of the correct values.

[0066] According to one embodiment, it is proposed that the control module 10, in particular using an estimator 19, selects one of a plurality of predetermined state-space models 20 as a state-space model 20 suitable for the current tangential flow filtration, particularly the current feed medium, preferably at the start of tangential flow filtration. To this end, a plurality of standard models can be stored in the control module 10. The standard models may include criteria for selecting one of them.

[0067] As described in the preferred embodiment, the state-space model 20 is time-invariant. Alternatively, the state-space model 20 may be particularly slowly time-varying and may adapt to changes in the filter 3 over time. Further alternatively, the control module 10 can adapt the state-space model 20 over time based on the measured variables 11, as described above. Preferably, the control module 10 compares the predictive operating variables 15 of the previous cycle with the measured variables 11 and adapts the state-space model 20 based on the deviation.

[0068] According to one embodiment, it is proposed to use changes in the state-space model 20 to plan and / or predict maintenance, in particular, automatically.

[0069] Another equally important teaching relates to a control module 10 configured to use a model predictive control unit 13 in a method for performing tangential flow filtration on a biomolecular solution, particularly for performing single-pass tangential flow filtration on a biomolecular solution, as part of a biomolecule manufacturing process, by the proposed method. All descriptions given with respect to the proposed method are fully applicable. In particular, the control module 10 can be adapted to perform any of the method steps described with respect to the control module 10.

[0070] Another equally important teaching relates to a tangential flow filtration apparatus 1 configured to perform the proposed method. All previously given descriptions are fully applicable.

Claims

1. A method for performing tangential flow filtration on a biomolecular solution as part of a biomolecule manufacturing process, particularly a method for performing single-pass tangential flow filtration on a biomolecular solution, A tangential flow filtration device (1) is provided, the tangential flow filtration device (1) comprises a tangential flow filtration module (2), the tangential flow filtration is performed within the tangential flow filtration module (2), the tangential flow filtration module (2) comprises a filter (3), a supply fluid input line (4) for the biomolecular solution as a supply medium, a retaining fluid output line (6) for the retaining fluid, a filtrate output fluid line (7) for the filtrate, and preferably a buffer input fluid line (5) for the buffer, and the tangential flow filtration device (1) comprises at least two actuators (8) that affect the tangential flow filtration, The tangential flow filtration apparatus (1) comprises a control module (10), and the control module (10) controls the tangential flow filtration by providing a control signal to the actuator (8) and by measuring at least two measurement variables (11) that describe the tangential flow filtration. The control module (10) controls the tangential flow filtration using the process control unit, The process control unit is a model prediction control unit (13), and the control module (10) inputs the measurement variable (11) as an input variable (14) to the model and derives at least two control variables (15) as control signals for the actuator (8). A method characterized by the following.

2. The method according to claim 1, characterized in that the tangential flow filter (1) comprises at least one actuator (8), particularly a pump, that acts on the supply fluid input fluid line (4) and / or the retaining fluid output fluid line (6) and / or the buffer input fluid line (5), and / or the tangential flow filter (1) comprises at least one sensor (9), particularly a pressure sensor (9) and / or a flow sensor (9) on the supply fluid input fluid line (4) and / or the retaining fluid output fluid line (6) and / or the buffer input fluid line (5).

3. The method according to claim 2, characterized in that the operating variable (15) includes the speed of the supply fluid pump and / or the retention fluid pump and / or the buffer pump, and / or the pressure and / or volumetric flow rate of the supply fluid and / or the retention fluid and / or the buffer, preferably the speed of the supply fluid pump, the retention fluid pump and the buffer pump.

4. The method according to any one of claims 1 to 3, characterized in that the process control unit receives, in particular from a user, set values ​​(12) for at least one, preferably at least two, more preferably at least three, and / or up to five, more preferably up to four, control variables controlled by the process control unit, and in particular holds them constant until a new set value (12) is received from the user, at least one, preferably all, control variables are not linearly dependent on the input variable (14) and / or the manipulated variable (15), and / or the number of control variables is equal to the number of actuators (8).

5. The method according to any one of claims 1 to 4, characterized in that the control variables include, or consist of, an intermembrane pressure difference, in particular an average intermembrane pressure difference, and / or a concentration factor, in particular a ratio of the supply liquid volume flow rate to the retained liquid volume flow rate, and / or a diafiltration ratio, in particular a ratio of the buffer volume flow rate to the supply liquid volume flow rate.

6. The method according to any one of claims 1 to 5, wherein the model predictive control unit (13) includes a state-space model (20) having state variables, preferably the state-space model (20) describes the effect of the actuator (8) on the state of the tangential flow filtration module (2), and / or at least one, in particular each control variable, is derived from at least two state variables.

7. The method according to any one of claims 1 to 6, characterized in that the state-space model (20) has at least three, preferably at least six, more preferably at least nine, more preferably at least twelve, and / or up to 30, preferably up to 24, more preferably up to 15, more preferably up to 12 state variables, and / or the state-space model (20) includes, in particular, one first-order differential equation for each state variable, preferably exactly one for each state variable, and / or the state-space model (20) does not include a differential equation of order higher than 1, preferably the state variables include the pump speed, in particular rotational speed, and / or the flow rate of the feed fluid pump and / or the retention fluid pump and / or the buffer pump, and / or the volumetric flow rate of the feed fluid and / or the retention fluid and / or the buffer, and / or the influence of one or more, preferably all, pressures on the feed fluid pressure and retention fluid pressure, in particular the average feed fluid pressure and the average retention fluid pressure.

8. The method according to any one of claims 1 to 7, characterized in that, in the current cycle, the model prediction control unit (13) predicts an operational variable (15) for several cycles of the prediction horizon based on the current state, the first predicted operational variable (15) is used as the control signal for the current cycle, a new prediction cycle is started for the next cycle, and preferably, during the prediction, the operational variable (15) is changed only during a control horizon of fewer cycles than the prediction horizon.

9. The method according to any one of claims 1 to 8, characterized in that the model prediction control unit (13) includes limit values ​​for the prediction variable and / or the manipulated variable (15).

10. The method according to any one of claims 1 to 9, wherein the model prediction control unit (13) uses an optimizer (21) to derive the manipulated variable (15), preferably the optimizer (21) uses a cost function (22) which includes the squared deviation of the calculated difference between the set value (12) for the prediction cycle and the current value of the manipulated variable (16), and / or the squared deviation of the change of the manipulated variable (15) over the prediction cycle, and / or the squared deviation of the change of the state variable during the prediction cycle, and / or a time-independent cost function (22) for the prediction cycle.

11. The method according to any one of claims 1 to 10, wherein an estimator (19) is used to estimate at least one input variable (14), preferably the estimated variable is input to the state-space model (20), and / or the predictor variable is compared to the measured variable (11) to detect a change in the filter (3), more preferably the control module (10) uses the detected change to modify the state-space model (20).

12. The method according to any one of claims 6 to 11, characterized in that the control module (10) selects, in particular, the estimator (19), one of a plurality of predetermined state-space models (20) as the state-space model (20) that is suitable for the current tangential flow filtration, in particular the current feed medium, preferably at the start of the tangential flow filtration.

13. The method according to any one of claims 6 to 12, characterized in that the state-space model (20) is time-invariant, or the state-space model (20) is time-varying and adapted to changes in the filter (3) over time, or the control module (10) adapts the state-space model (20) over time based on the measured variable (11), preferably the control module (10) adapts the state-space model (20) based on the deviation by comparing the predicted operating variable (15) of the previous cycle with the measured variable (11).

14. The method according to any one of claims 11 to 13, characterized in that the changes in the state-space model (20) are used, in particular, automatically to plan and / or predict maintenance.

15. A control module configured to use a model predictive control unit (13) in a method for performing tangential flow filtration on a biomolecular solution as part of a biomolecule manufacturing process, particularly for performing single-pass tangential flow filtration on a biomolecular solution, as described in any one of claims 1 to 14.

16. A tangential flow filtration apparatus configured to perform the method according to any one of claims 1 to 14.

Images