Method for controlling a particle-forming fluidization process proceeding in a fluidizer - Patent Application 20070122997

Abstract

The present invention relates to a method for controlling a particle-forming fluidization process proceeding in a fluidizer with respect to at least one product characteristic (w) of the process product.

Description

[Technical Field]

[0001] The present invention relates to a method for controlling a particle-forming fluidization process occurring in a fluidizer with respect to at least one product characteristic of the process product. [Background technology]

[0002] According to the undocumented prior art, particle-forming fluidization processes are typically driven by a set of process parameters, a so-called "recipe", which contain process parameters that lead to the desired product properties when a predefined sequence of process parameters is observed, so that the fluidization process always follows the same time flow.

[0003] Furthermore, typical product properties, such as the absolute moisture content of the particles, can be adjusted by controlling process parameters, such as the drying gas temperature or the volumetric flow rate of the drying gas. A corresponding method for treating particulate process products in a fluidizer and the associated fluidizer are disclosed in German patent application DE 10 02 04 199 A1.

[0004] Although known methods approach the desired target value for the product characteristic being controlled, they still have deviations therefrom that have a negative impact on the product quality of the process. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] German Patent Invention No. 102020208204 Summary of the Invention [Problem to be solved by the invention]

[0006] The object of the present invention is therefore to develop an improved method for controlling a particle-forming fluidization process taking place in a fluidizer, which further optimizes the product properties of the process product with respect to target values. [Means for solving the problem]

[0007] The object of the present invention is achieved by the method comprising the steps of: determining a number of process parameters of a fluidization process at a first time point; transmitting the process parameters as actual process parameter values ​​to a control device having a control function; calculating in the control device a process model product characteristic value for a second time point subsequent to the first time point using the actual process parameter values ​​based on a process model stored for at least one product characteristic; forming in the control device a number of optimized variable sets from the provided number of optimized process parameter values; calculating a number of optimized preview values ​​for a third time point using the optimized variable sets by the optimization model using the optimized variable sets; applying a correction value to each of the optimized preview values ​​to form a corrected optimized preview value; calculating an optimized difference value for the third time point by comparing each corrected optimized preview value with a preview target value for at least one product characteristic determined at the third time point from a target value function stored in the control device; and outputting the optimized process parameter value of the optimized variable set having the smallest absolute value as a reference variable for the second time point subsequent to the first time point. Advantageously, the method for controlling a particle-forming fluidization process occurring in a fluidizer determines suitable reference variables for achieving at least one product characteristic of the process product to be controlled without user presettings, thereby clearly improving the product quality of the process product while providing highly reproducible results.

[0008] In this regard, multiple method cycles are performed one after the other, with the second point in time of each method cycle forming the first point in time of the following method cycle, so that, for example, the value at the second point in time of a first method cycle forms the value at the first point in time of a second method cycle.

[0009] The process parameters are determined expediently by measuring the process parameters or by simulation, whereby it is advantageous that the process parameters can be provided in various ways, which can lead to savings in measurement technology, for example, when simulating the process parameters.

[0010] According to another development of the method, the measurement of the process parameters is carried out as an in-line measurement and / or an at-line measurement and / or an online measurement, and the process parameters are suitably measured at a process parameter sampling frequency. In the method for controlling the particle-forming fluidization process occurring in a fluidizer with respect to at least one product property of the process product, the current process parameters are always provided for control.

[0011] In another development of the method, the actual process parameter values ​​determined at the first time point form a process parameter variable set, which facilitates transferring the actual process parameter values ​​to the control device. In this regard, each optimization variable set is formed from a number of optimized process parameter values ​​corresponding to a number of actual process parameter values ​​of the process parameter variable set, and in the optimization variable set, at least one optimized process parameter value replaces the corresponding actual process parameter value.

[0012] Furthermore, each process parameter optimization value can take any optimization value, which is preferably selectable from a preset number of optimization values. This limits the number of optimization variable sets resulting from any combination of the process parameter and the preset number of optimization values ​​for each process parameter. When the preset number of optimization values ​​for each process parameter optimization value is purposefully the same, the maximum number of optimization variable sets is the number of process parameter optimization values ​​multiplied by the number of the preset number of optimization values. In principle, the number of preset number of optimization values ​​is arbitrary, and the number of optimization variable sets, and thus the number of optimization values, is purposefully determined by the calculation time required for the optimization model in the control device. The higher the performance of the control device, the more optimization values ​​and process parameter optimization values ​​can be used. Furthermore, a high number of optimization values ​​and process parameter optimization values ​​also affects the accuracy of control of the fluidization process.

[0013] Preferably, the preset optimization values ​​are based on the actual values ​​of the respective process parameters. The optimization values ​​then map as large a value range as possible that is within the technical boundaries of the fluidization process. The optimization values ​​that lead to instability in the fluidization process then no longer lie within the technical boundaries of the fluidization process. The technical and product-specific boundaries of the fluidization process for each process parameter are usually determined in a preliminary experiment associated with the fluidization process.

[0014] In an additional development of the method, a first time span having at least one time step is located between the first and second time points, and a second time span having at least one time step is located between the second and third time points. In this regard, the first time span and the second time span each have a different number of time steps, with the first time span advantageously having a single time step. In preliminary experiments, it has been found to be advantageous for the first time span to have one time step and the second time span to have 19 time steps. This results in 20 time steps between the first and third time points. However, the time span between the first and third time points may also be directed further into the future, for example, 30, 40, 50, or more time steps. The second time span should be adjusted accordingly. Expediently, the number of time steps of the first and second time spans is adapted to one another, but the number of time steps of the first and second time spans remains freely selectable for each individual fluidization process.

[0015] In one development of the method, a setpoint function is stored in the control device for each product characteristic to be controlled. The setpoint function stored in the control device for each product characteristic to be controlled advantageously maps the desired progression of the respective product characteristic to be controlled over time. If the selected product characteristic to be controlled is particle size, the setpoint function maps, for example, particle growth over time.

[0016] In another development of the method, a target value function for at least one product characteristic to be controlled is generated from experimental data or from a target value process model, whereby the target value function is adapted to the physicochemical basis applied to the particle-forming fluidization process. Furthermore, the method allows for the target value function to be any function preset by an operator and storable in the control device.

[0017] In this regard, the target value process model is suitably based on a kinetic model of at least one product characteristic. A kinetic model here refers to a mathematical representation of the progression of at least one product characteristic among all product characteristics to be controlled in the fluidization process in relation to various process parameters, such as the growth kinetics of particle size. Preferably, in this method, the at least one product characteristic is particle size and / or particle moisture and / or particle composition. The use of a kinetic model also allows the target value function to be adapted to the physicochemical basis applied to the particle formation fluidization process.

[0018] According to another development of the method, at least one product characteristic to be controlled is determined as a product characteristic measured at a first time point and transmitted to the control device as a product characteristic actual value. In this regard, the product characteristic actual value is smoothed by a mathematical smoothing method, preferably by the Whittaker-Henderson method. Furthermore, the product characteristic actual value forms a product characteristic variable set. This facilitates the transfer of the product characteristic actual value to the control device.

[0019] Preferably, the correction value at the first time point is calculated by subtracting the process model product characteristic value of at least one product characteristic calculated for the at least one product characteristic at the first time point from the at least one product characteristic actual value detected at the first time point. By calculating the correction value in this manner, an error calculated by the process model at each time step and accumulating over time is corrected. Preferably, the correction value at the first time point is "0" in the first method cycle, since no product characteristic actual value at the first time point has been detected.

[0020] Preferably, the detection of the product characteristic is carried out as an in-line measurement and / or an at-line measurement and / or an online measurement, and the product characteristic is detected at a product characteristic sampling frequency. In the method for controlling the particle-forming fluidization process proceeding in the fluidization device with respect to at least one product characteristic of the process product, the current product characteristic is always provided for control.

[0021] In a further preferred development of the method, the process parameter sampling frequency and the product characteristic sampling frequency have the same value, thereby ensuring that, during a method cycle or at a given point in time, the current process parameter actual value and, at the same time, the current product characteristic actual value are provided.

[0022] Furthermore, a product characteristic variable set is formed from two product characteristics to be controlled, and the product characteristics to be controlled are prioritized with respect to preferential control of one of the product characteristics. The product characteristics to be controlled (also referred to as weighting the product characteristics to be controlled) are prioritized according to the importance of each product characteristic in the fluidization process and / or with respect to product quality and / or user preference. For example, in a fluidization process, if the particle humidity to be achieved in the process product is more important than the particle size of the process product, the particle humidity should be prioritized or weighted accordingly. The prioritization causes the optimized difference value of the prioritized product characteristic to be preferentially achieved compared to lower-prioritized product characteristics.

[0023] For this purpose, for example, a plurality of optimization difference values ​​of the product characteristic to be controlled of the optimization variable set are added, and when added, each optimization difference value is multiplied by a weighting factor according to its priority, i.e., weighted. The sum of the optimization difference values ​​can then be divided by the number of optimization difference values. In one example, the lowest sum or lowest average process parameter optimization value of the optimization variable set belonging to the optimization difference value is then output as the reference variable.

[0024] According to one development of the method, the optimization model is based on the process model, in particular the optimization model corresponds to the process model, which ensures that the multiple optimized forecast values ​​calculated by the optimization model at the third time point have the same basis as the process model product characteristic values ​​calculated by the process model at the first time point, thereby achieving improved control of at least one product characteristic to be controlled.

[0025] In a further development of the method, the process model for calculating the process model product characteristic value is based on a linear or nonlinear process model of the fluidization process to be controlled, and the nonlinear process model is suitably an artificial neural network. In this regard, the artificial neural network is particularly configured as a multilayer perceptron, or as a simple recurrent network, such as an ELMAN network, or as a nonlinear autoregressive exogenous network, such as a NARX network. The artificial neural network is suitably trained by experiments carried out in a fluidizer prior to carrying out the method for controlling a particle-forming fluidization process occurring in the fluidizer with respect to at least one product characteristic of the process product, during which various process parameters are changed in each case taking into account at least one product characteristic to be controlled.

[0026] As process parameters, preferably one or more process parameters are used selected from the group of atomizing gas pressure and / or atomizing rate and / or atomizing amount and / or particle temperature and / or drying gas temperature at the inlet of the fluidizing device and / or relative humidity of the drying gas at the outlet and / or drying gas volumetric flow rate.

[0027] The present invention will be described in detail below with reference to the accompanying drawings. [Brief explanation of the drawings]

[0028] [Figure 1] 1 is a schematic diagram of a method for controlling a particle-forming fluidization process occurring within a fluidizer with respect to at least one product characteristic of the process product. [Figure 2] 1 is a graph including product properties plotted over time in a first process cycle and a display of detailed extracts A and B. [Figure 3] This is an enlarged view of detailed extract A. [Figure 4] This is an enlarged view of detailed extract B. [Figure 5]1 is a graph including product properties plotted over time in a second process cycle and a display of detailed extracts C and D. [Figure 6] This is an enlarged view of detailed extract C. [Figure 7] This is an enlarged view of detailed extract D. DETAILED DESCRIPTION OF THE INVENTION

[0029] Unless otherwise specified, the following description relates to all embodiments shown in the drawings of a method for controlling a particle-forming fluidization process occurring in a fluidizer with respect to at least one product characteristic w of the process product.

[0030] The subscript n used in this description is any natural number, where n can take on different values ​​for different alphabets. For example, for on, n=3, and for o"n, n=27.

[0031] In a method cycle z, the flow of which is described below on the basis of the embodiment shown in Fig. 1, a number of process parameters p of the fluidization process are determined at a first time t1, where the first time t1 is to be understood as the current time t of the fluidization process. The process parameters p can be, inter alia, the atomizing gas pressure, the atomizing rate, the atomizing amount, the particle temperature, the drying gas temperature at the inlet of the fluidizer, the relative humidity of the drying gas at the outlet, and / or the drying gas volumetric flow rate.

[0032] The process parameter p is determined by simulation or by measurement. The measurement of the process parameter p is optionally carried out as in-line, at-line, or online measurement using corresponding measurement techniques known to those skilled in the art. Expediently, the process parameter p is measured at a process parameter sampling frequency fp. The determined process parameter p is transmitted as an actual process parameter value p' to a control device 1 having a control function. The actual process parameter value p' determined at a first time point t1 then preferably forms a process parameter variable set p". In an embodiment not shown, some of the process parameters p are simulated, and other parts of the process parameters p are determined by measurement technology.

[0033] In parallel with the measurement and / or simulation of the process parameter p, at a first time point t1, at least one product characteristic wm is also measured with a product characteristic sampling frequency fw. The measurement is preferably performed as an in-line, at-line, or online measurement. The at least one product characteristic w to be controlled is determined as a product characteristic actual value w'm measured at the first time point t1 and transmitted to the control device 1. The product characteristic actual values ​​w'm preferably form a product characteristic variable set w"m. Particle size and / or particle moisture and / or particle composition, among others, are used as product characteristics.

[0034] Particularly preferably, the process parameter sampling frequency fp and the product characteristic sampling frequency fw have the same value, so that the process parameter actual value p′ and the product characteristic actual value w′m are present in the control device 1 at the same time.

[0035] In the control device 1, the delivered actual product characteristic values ​​w'm are smoothed by a mathematical smoothing method in a smoothing module 2 assigned to the control device 1. Expediently, this is carried out by a mathematical smoothing method, for example the Whittaker-Henderson method. The mathematically smoothed product characteristic values ​​w's then form, inter alia, a product characteristic variable set w"s.

[0036] The control device 1 further comprises a process model module 3 in which, based on a stored process model for each at least one product characteristic w, a process model product characteristic value w'c is calculated for a second time t2 subsequent to the first time t1 using the detected actual process parameter values ​​p', which preferably form a process parameter variable set p" on the basis of the detected process model. The process model product characteristic values ​​w'c suitably form a process model product characteristic variable set w"c.

[0037] The process model for calculating the corresponding process model product characteristic value w'c is based on a linear or nonlinear process model of the fluidization process to be controlled, and as the nonlinear process model, an artificial neural network is suitably used. In this case, the artificial neural network is preferably configured as a multilayer perceptron, a simple recurrent network, or a nonlinear autoregressive exogenous network. Prior to carrying out the method for controlling a particle-forming fluidization process occurring in a fluidizer with respect to at least one product characteristic of the process product, the artificial neural network is suitably trained by experiments carried out in the fluidizer, during which various process parameters p are respectively changed taking into account at least one product characteristic w to be controlled.

[0038] The control device 1 further includes a correction module 4. In the correction module 4, a correction value k at the first time t1 is calculated. The correction value k is calculated for each product characteristic w to be controlled by subtracting a process model product characteristic value w'c calculated for the at least one product characteristic w at the first time t1 from at least one product characteristic actual value w'm, preferably a mathematically smoothed product characteristic actual value w's, detected at the first time t1. The correction value k may also form a correction value variable set k". In a first method cycle z1, based on the missing process model product characteristic value w'c at the first time t1, each correction value k is set to the value "zero".

[0039] Furthermore, before starting the method for controlling the particle-forming fluidization process occurring in the fluidizer with respect to at least one product characteristic w of the process product, an optimization value v is preset based on a predictable effective range of the process parameter p. The effective range here means, for example, a specific effective range, i.e., for example, 75°C to 100°C, for the particle temperature used as the process parameter optimization value o, and the optimization value v divides the effective range into particularly equal intervals, for example, into portions having six values, here, i.e., v1 = 75°C, v2 = 80°C, v3 = 85°C, v4 = 90°C, v5 = 95°C, and v6 = 100°C.

[0040] Generally speaking, each of the process parameter optimization values ​​o can take any optimization value v, which can be selected from a number of preset optimization values ​​v. The number of optimization values ​​v can suitably range from two optimization values ​​to n optimization values. For this reason, FIG. 1 exemplarily shows an optimization value vo1 for the process parameter optimization value o1 for every optimization value v. In the illustrated embodiment, the process parameter optimization value o1 has six optimization values ​​vo1,1 to vo1,6. Suitably, but not necessarily, all process parameter optimization values ​​o have the same number of optimization values ​​v.

[0041] The control device 1 further comprises an optimization module 5. In a next step, in a combinatorial logic module 6 assigned to the optimization module 5, each optimization variable set o" is formed from a number of optimized process parameter values ​​o corresponding to a number of actual process parameter values ​​p' of the process parameter variable set p", where at least one optimized process parameter value o replaces the corresponding actual process parameter value p' in the optimization variable set o". Therefore, in the combinatorial logic module 6, n6 optimization variable sets o" are formed, for example, when there are n process parameter optimized values ​​o and six optimized values ​​vo1,1 to vo1,6. In particular, for three process parameter optimized values ​​o, each with six optimized values ​​v, 36 = 729 optimization variable sets o" will result. The number of optimized values ​​v and the number of process parameter optimized values ​​o depend, inter alia, on the performance of the control device 1.

[0042] Additionally, the optimization module 5 has an optimization model module 7. Using the determined optimization variable set o″, a number of optimization forecast values ​​x for the third time point t3 is calculated by means of an optimization model stored in the control device 1, preferably in the optimization model module 7. The number of optimization forecast values ​​x in this case corresponds to the number of optimization variable sets o″. The optimization model in this case is preferably based on a process model, in particular the optimization model corresponds to the process model.

[0043] In the next step, a correction value k is applied to each of the optimized preview values ​​x to form a modified optimized preview value xk. In the illustrated embodiment of Figure 1, the correction value k is subtracted from the optimized preview value x to form the modified optimized preview value xk.

[0044] Subsequently, in a comparison module 8 assigned to the control device 1, an optimized difference value Δ at the third time t3 is calculated by comparing each of the modified optimized preview values ​​xk with a preview setpoint xS for at least one product characteristic w, determined at the third time t3 from at least one setpoint function S stored in the control device 1. Preferably, one setpoint function S is stored in the control device 1 for each product characteristic w to be controlled. In this case, the setpoint function S for the at least one product characteristic w to be controlled is formed, in particular, from experimental data or from a setpoint process model. Appropriately, the kinetics of the product characteristic, for example the growth kinetics of particle size, can be used to pre-set the setpoint xS.

[0045] In the illustrated embodiment of Figure 1, the optimization difference value Δ is formed as the absolute value of the subtraction of the modified optimized preview value xk and the target value xS. In the illustrated embodiment, a set of variables is considered respectively.

[0046] The control device 1 further comprises an evaluation module 9 in which the absolute values ​​of the optimized difference values ​​Δ are evaluated and compared with each other. The optimized process parameter value o of the optimized variable set o″ belonging to the smallest absolute value of the optimized difference values ​​Δ is then output as a reference variable r for a second time point t2 subsequent to the first time point t1. The respective reference variable r is then provided to and regulated by further control (P, PI, PID control) of the process parameter p in the fluidization process.

[0047] From the two product characteristics w to be controlled, a product characteristic variable set w" is formed, and the product characteristics w to be controlled are prioritized in terms of priority control. The product characteristics w to be controlled are prioritized according to the importance of each product characteristic w in the fluidization process. For example, in a fluidization process, the product characteristics of the particle size of the process product and the particle humidity of the process product are to be controlled, and the particle humidity to be achieved in the process product is more important than the particle size of the process product, then the particle humidity should be prioritized or weighted accordingly. For this purpose, the control device 1 has a weighting module 10, and the prioritization is performed in the weighting module 10.

[0048] The prioritization causes prioritized product characteristics w to be preferentially achieved compared to lower prioritized product characteristics w. For this purpose, for example, multiple numerical values ​​of the optimization difference values ​​Δ of the product characteristics w to be controlled are added, and each absolute value of the optimization difference values ​​Δ is weighted during the addition according to its priority, for example multiplied by its weighting coefficient g. The sum of the optimization difference values ​​can then be divided by the number of optimization difference values. In one example, the process parameter optimization value o with the lowest sum of numerical values ​​of the optimization variable set o belonging to the optimization difference values ​​Δ is output as the reference variable.

[0049] Between the first time point t1 and the second time point t2 lies a first time span Δt1 having at least one time step d, and between the second time point t2 and the third time point t3 lies a second time span Δt2 having at least one time step d. The first time span Δt1 and the second time span Δt2 preferably have different numbers of time steps d, with the first time span Δt1 expediently having a single time step d.

[0050] In the embodiment shown in Figure 1, a time span Δt1 having one time step d is located between a first time point t1 and a second time point t2, and a time span Δt2 having 19 time steps d is located between the first time point t2 and a third time point t3, resulting in 20 time steps d between the first time point t1 and the third time point t3. The first and second time spans Δt1 and Δt2 may have other value combinations, and the listed value combinations represent preferred combinations.

[0051] A number of method cycles z may proceed one after the other, with the second time point t2 of the preceding method cycle z respectively forming the first time point t1 of the following method cycle z+1.

[0052] FIG. 2 shows a graph with a product characteristic w plotted over time t in a first method cycle z1 and with indications of detailed extracts A and B.

[0053] In this embodiment, control is performed according to one product characteristic w, which is selected as the product characteristic w.

[0054] For the control itself, in the illustrated embodiment, three process parameters p were used: atomizing gas pressure as process parameter p1, atomization rate as process parameter p2, and particle temperature as process parameter p3, with each process parameter p being allowed to take only two optimized values ​​v.

[0055] Between the first time point t1 and the second time point t2, there is a time span Δt1 having one time step d, and between the first time point t2 and the third time point t3, there is a time span Δt2 having 19 time steps d, resulting in 20 time steps d between the first time point t1 and the third time point t3.

[0056] Additionally, a target value function S for particle size as a product characteristic w is shown.

[0057] In the combinatorial theory module 6, when there are three process parameter optimization values ​​o and two optimization values ​​v, 2 Optimization variable sets o" are formed. Specifically, within the combinatorial theory module 6, nine optimization variable sets o" are therefore formed. The nine functions resulting from the optimization model with the respective optimization variable sets o" are indicated in the graph as F1 to F9.

[0058] The optimized preview values ​​x1 to x9 are generated as function values ​​of the nine functions F1 to F9 at the third time point t3.

[0059] 3, the profile of the measured actual product characteristic value w'm and the smoothed actual product characteristic value w's are shown. The smoothed actual product characteristic value w's is additionally shown at time t2. Furthermore, the actual product characteristic value w'c calculated by the process model at a second time t2 is shown.

[0060] 4 shows an enlarged view of functions F5 and F6 and the optimized preview values ​​x5-x6 at time t3. Both functions F5 and F6 are located closest to the target value xS of the target value function S at time t3.

[0061] The correction value k is calculated for the product characteristic w to be controlled by subtracting the process model product characteristic value w'c calculated for the product characteristic w at the first time point t1 from the mathematically smoothed product characteristic actual value w's at the first time point t1. Since the respective correction value k cannot be calculated in the first method cycle z1 due to the missing process model product characteristic value w'c at the first time point t1, the correction value k is set to the value "zero" in the first method cycle z1.

[0062] In the illustrated embodiment of FIG. 4, the optimized preview value x5 or x6 is subtracted from the target value xS to form optimized difference values ​​Δ5 and Δ6. A comparison of the absolute values ​​of the optimized difference values ​​Δ5 and Δ6 shows that the absolute value of the optimized difference value Δ6 is smaller than the absolute value of the optimized difference value Δ5. As a result, the optimized value o of the process parameter of the optimized variable set o6 belonging to the optimized difference value Δ6 is output as the reference variable r for the next time point t2.

[0063] 5 to 7 show the same as FIGS. 2 to 4, but for the second method cycle z2.

[0064] FIG. 5 shows a schematic diagram of a second method cycle z2 of an exemplary embodiment, including product characteristics w plotted over time t and excerpts C and D.

[0065] The product characteristic actual values ​​w's and the process model product characteristic values ​​w'c available at the second time point t2 of the method cycle z form the product characteristic actual values ​​w's and the process model product characteristic values ​​w'c at the first time point t1 in the following method cycle z+1.

[0066] Between the first instant t1 and the second instant t2 there is still a time span Δt1 with one time step d, and between the second instant t2 and the third instant t3 there is a time span Δt2 with 19 time steps d, so that, as in the first method cycle z1, there are 20 time steps d between the first instant t1 and the third instant t3.

[0067] Additionally, a target value function S for particle size as a product characteristic w is shown.

[0068] Within the combinatorial theory module 6, for three process parameter optimization values ​​o and two optimization values ​​v respectively, 2Optimization variable sets o" are formed. Specifically, within the combinatorial theory module 6, nine optimization variable sets o" are therefore formed. The nine functions resulting from the optimization model with the respective optimization variable sets o" are indicated in the graph as F1 to F9.

[0069] The optimized preview values ​​x1 to x9 are generated as function values ​​of the nine functions F1 to F9 at the third time point t3.

[0070] 6, the curves of the measured actual product characteristic value w'm and the smoothed actual product characteristic value w's are shown. The smoothed actual product characteristic value w's is additionally shown at time t2. Furthermore, the actual product characteristic value w'c calculated by the process model at a second time t2 is shown.

[0071] 7 shows an enlarged view of functions F6 and F7 and the optimized preview values ​​x6-x7 at time t3. Both functions F6 and F7 are located closest to the target value xS of the target value function S at time t3.

[0072] The correction value k is calculated for the product characteristic w to be controlled by subtracting the process model product characteristic value w'c(t1) calculated for the product characteristic w at the first time point t1 from the mathematically smoothed product characteristic actual value w's(t1) at the first time point t1.

[0073] In the illustrated embodiment of FIG. 7, the correction value k is subtracted from the optimized preview value x6 or x7, respectively, and then the target value xS is subtracted from the corrected optimized preview value x6,k or x7,k to form the optimized difference values ​​Δ6 and Δ7. A comparison of the absolute values ​​of the optimized difference values ​​Δ6 and Δ7 shows that the absolute value of the optimized difference value Δ7 is smaller than the absolute value of the optimized difference value Δ6. As a result, the optimized value o of the optimized variable set o″7 belonging to the optimized difference value Δ7 is output as the reference variable r for the next time point t2. The present application relates to the invention described in the claims, but also includes the following as other aspects. 1. 1. A method for controlling a particle-forming fluidization process occurring in a fluidizer with respect to at least one product characteristic (w) of a process product, comprising: In a method cycle (z), a number of process parameters (p) of a fluidization process are determined at a first time point (t1); The process parameter (p) is transmitted as a process parameter actual value (p') to a control device (1) having a control function; Calculating in the control device (1) a process model product characteristic value (w'c) for a second time point (t2) subsequent to the first time point (t1) based on the process model stored for at least one of the product characteristics (w) and the actual process parameter value (p'); In the control device (1), a number of optimized variable sets (o") are formed from the provided number of optimized process parameter values ​​(o); Using the optimized variable set (o"), an optimization model calculates a number of optimized predicted values ​​(x) at a third time point (t3); applying a correction value (k) to each of the optimized preview values ​​(x) to form a corrected optimized preview value (xk); calculating an optimization difference value (Δ) at the third time point (t3) from a comparison between each of the modified optimization preview values ​​(xk) and a preview target value (xS) for at least one product characteristic (w) determined at the third time point (t3) from a target value function (S) stored in the control device (1); Subsequently, the process parameter optimized value (o) of the optimized variable set (o") that belongs to the smallest absolute value of the optimized difference value (Δ) is output as the reference variable (r) for the second time point (t2) subsequent to the first time point (t1); A method characterized by: 2. 2. The method according to claim 1, characterized in that a number of method cycles (z) proceed one after the other, and the second time point (t2) of a method cycle (z) forms the first time point (t1) of the next succeeding method cycle (z+1). 3. 3. The method according to 1 or 2 above, wherein the process parameter (p) is determined by measurement or simulation. 4. 4. The method according to claim 3, characterized in that the measurement of the process parameter (p) is carried out as an in-line measurement and / or an at-line measurement and / or an online measurement, and suitably the process parameter (p) is measured at a process parameter sampling frequency (fp). 5. 5. The method of any one of claims 1 to 4, wherein the process parameter actual values ​​(p') determined at the first time point (t1) form a process parameter variable set (p"). 6. 6. The method of claim 5, wherein each of the optimization variable sets (o") is formed from a number of optimized process parameter values ​​(o) corresponding to a number of actual process parameter values ​​(p') of the process parameter variable set (p"), and in the optimization variable set (o"), at least one optimized process parameter value (o) replaces a corresponding actual process parameter value (p'). 7. 7. Any one of the methods 1 to 6 above, characterized in that each of the process parameter optimization values ​​(o) can take on any optimization value (v), and the optimization value (v) can be selected from a number of preset optimization values ​​(v). 8. 8. The method of claim 7, wherein the preset optimization value (v) is based on each of the process parameter actual values ​​(p'). 9. 9. Any one of the methods 1 to 8 above, characterized in that a first time span (Δt1) having at least one time step (d) is located between the first point in time (t1) and the second point in time (t2), and a second time span (Δt2) having at least one time step (d) is located between the second point in time (t2) and the third point in time (t3). 10. 10. The method of claim 9, wherein the first time span (Δt1) and the second time span (Δt2) each have a different number of time steps (d), and the first time span (Δt1) expediently has a single time step (d). 11. 11. The method according to any one of claims 1 to 10, characterized in that for each product characteristic (w) to be controlled, a respective setpoint function (S) is stored in the control device (1). 12. 12. The method of any one of 1 to 11 above, characterized in that the target function (S) for at least one product characteristic (w) to be controlled is formed from experimental data or from a target process model. 13. 13. The method of claim 12, wherein the target process model is based on a kinetic model of at least one product characteristic (w). 14. 14. The method according to any one of claims 1 to 13, characterized in that at least one product characteristic (w) to be controlled is detected as a product characteristic (p) measured at a first time point (t1) and transmitted to the control device (1) as a product characteristic actual value (p'). 15. 15. The method according to claim 14, characterized in that the detected actual product characteristic value (p') is smoothed by a smoothing method, expediently by the Whittaker-Henderson method. 16. 16. The method of claim 14 or 15, wherein the product characteristic actual values ​​(p') form a product characteristic variable set (p"). 17. 17. The method according to any one of claims 14 to 16, characterized in that the correction value (k) at the first time point (t1) is calculated by subtracting a process model product characteristic value (w'c) of at least one product characteristic (w) calculated for the at least one product characteristic (w) at the first time point (t1) from at least one product characteristic actual value (w'm) detected at the first time point (t1). 18. 18. The method according to any one of claims 14 to 17, characterized in that the detection of the product characteristic (w) is carried out as an in-line measurement and / or an at-line measurement and / or an online measurement, and suitably the product characteristic (w) is detected at a product characteristic sampling frequency (fw). 19. 19. The method of claim 4 or 18, wherein the process parameter sampling frequency (fp) and the product characteristic sampling frequency (fw) have the same value. 20. 20. The method according to any one of claims 1 to 19, characterized in that a product characteristic variable set (w") is formed from two product characteristics (w) to be controlled, and a prioritization of the product characteristics (w) to be controlled is performed with respect to preferential control of one product characteristic (w) among said product characteristics (w). 21. 21. The method of any one of claims 1 to 20, wherein at least one product characteristic (w) is particle size and / or particle moisture and / or particle composition. 22. 22. The method of any one of claims 1 to 21, wherein the optimization model is based on, and in particular corresponds to, the process model. 23. 23. The method according to any one of claims 1 to 22, characterized in that the process model for calculating the process model product characteristic value (w'c) is based on a linear or non-linear process model of the fluidization process to be controlled, whereby an artificial neural network is suitably used as the non-linear process model. 24. 23. The method of claim 22, wherein the artificial neural network is formed as a multilayer perceptron, as a simple recurrent network, or as a nonlinear autoregressive exogenous network. 25. 25. The method according to any one of claims 1 to 24, characterized in that as process parameter (p) one or more process parameters (p) are used selected from the group consisting of atomizing gas pressure and / or atomizing rate and / or atomizing amount and / or particle temperature and / or dry gas temperature at the inlet of the fluidizing device and / or relative humidity of the dry gas at the outlet and / or dry gas volumetric flow rate.

Claims

1. 1. A method for controlling a particle-forming fluidization process occurring in a fluidizer with respect to at least one product characteristic (w) of a process product, comprising: In a method cycle (z), actual values ​​of a plurality of process parameters (p) of the fluidization process are determined at a first time point (t1); transmitting the actual value of the process parameter (p) as a process parameter actual value (p') to a control device (1) having a control function; In the control device (1), based on the stored process model for at least one of the product characteristics (w), a process model product characteristic value (w'c) of the at least one of the product characteristics (w) is calculated using the process parameter actual value (p') for a second time point (t2) subsequent to the first time point (t1) of the method cycle (z), the second time point (t2) forming the first time point (t1) of a method cycle (z+1) subsequent to the method cycle (z), calculating the correction value (k) at the first time point (t1) by subtracting from the at least one actual product characteristic value (w'm) detected at the first time point (t1) a process model product characteristic value (w'c) previously calculated for the at least one product characteristic (w) at the first time point (t1) of the method cycle (z) for a second time point (t2) of the method cycle preceding the method cycle (z); In the control device (1), a plurality of process parameters of the fluidization process are each set to an arbitrary optimization value (v) for optimizing at least one of the product characteristics (w) with respect to a target value, thereby providing a plurality of process parameter optimization values ​​(o), and a plurality of optimization variable sets (o") are formed by combining the plurality of provided process parameter optimization values ​​(o) for each different optimization value (v); calculating, by an optimization model using each of the plurality of optimized variable sets (o"), a plurality of optimized forecast values ​​(x) for at least one of the product characteristics (w) at a third time point (t3), the number of which corresponds to the number of the optimized variable sets (o"); applying the correction value (k) to each of the optimized preview values ​​(x) to form a corrected optimized preview value (xk); calculating an optimization difference value (Δ) at the third time point (t3) from a comparison between each of the modified optimization preview values ​​(xk) and a preview target value (xS) for at least one product characteristic (w) determined at the third time point (t3) from a target value function (S) stored in the control device (1); Subsequently, the optimized process parameter value (o) of the optimized variable set (o") that belongs to the smallest absolute value of the optimized difference value (Δ) is output as a reference variable (r) for the second time point (t2) subsequent to the first time point (t1). A method characterized by:

2. The method described in claim 1, characterized in that a plurality of method cycles (z) proceed in succession, and the second point in time (t2) of a method cycle (z) each forms the first point in time (t1) of the next subsequent method cycle (z+1).

3. 3. The method according to claim 1, wherein the process parameter (p) is determined by measurement or by simulation.

4. 4. The method according to claim 3, characterized in that the measurement of the process parameter (p) is performed as an in-line measurement and / or an at-line measurement and / or an online measurement, and the process parameter (p) is measured at a process parameter sampling frequency (fp).

5. 3. The method according to claim 1, wherein the process parameter actual values ​​(p') determined at the first time point (t1) form a process parameter variable set (p'').

6. 6. The method of claim 5, wherein each of the optimization variable sets (o") is formed from a plurality of optimized process parameter values ​​(o) corresponding to a plurality of actual process parameter values ​​(p') of the process parameter variable set (p"), and wherein in the optimization variable set (o"), at least one optimized process parameter value (o) replaces a corresponding actual process parameter value (p').

7. 3. The method according to claim 1, wherein each of the process parameter optimization values ​​(o) can take any optimization value (v), and the optimization value (v) can be selected from a plurality of preset optimization values ​​(v).

8. 8. The method of claim 7, wherein the preset optimization values ​​(v) are based on the respective process parameter actual values ​​(p').

9. 3. The method according to claim 1, wherein a first time span (Δt1) having at least one time step (d) is located between the first time point (t1) and the second time point (t2), and a second time span (Δt2) having at least one time step (d) is located between the second time point (t2) and the third time point (t3).

10. 10. The method of claim 9, wherein the first time span (Δt1) and the second time span (Δt2) each have a different number of time steps (d), and the first time span (Δt1) has a single time step (d).

11. 3. The method according to claim 1, wherein a setpoint function (S) is stored in the control device (1) for each product characteristic (w) to be controlled.

12. 3. The method according to claim 1, wherein the setpoint function (S) for at least one product characteristic (w) to be controlled is formed from experimental data or from a setpoint process model.

13. 13. The method of claim 12, wherein the setpoint process model is based on a kinetic model of at least one product characteristic (w).

14. 3. The method according to claim 1, wherein at least one product characteristic (w) to be controlled is determined as a product characteristic (p) measured at a first time point (t1) and transmitted to the control device (1) as a product characteristic actual value (p').

15. 15. The method according to claim 14, characterized in that the detected product characteristic actual value (p') is smoothed by the Whittaker-Henderson method.

16. 15. The method of claim 14, wherein the product characteristic actual values ​​(p') form a product characteristic variable set (p'').

17. 15. The method according to claim 14, characterized in that the detection of the product characteristic (w) is performed as an in-line measurement and / or an at-line measurement and / or an online measurement, and the product characteristic (w) is detected at a product characteristic sampling frequency (fw).

18. At least one product characteristic (w) to be controlled is detected as a product characteristic (p) measured at a first time point (t1) and transmitted to the control device (1) as a product characteristic actual value (p'); The detection of the product characteristic (w) is carried out as an in-line measurement and / or an at-line measurement and / or an online measurement, and the product characteristic (w) is detected at a product characteristic sampling frequency (fw); 5. The method of claim 4, wherein the process parameter sampling frequency (fp) and the product characteristic sampling frequency (fw) have the same value.

19. 3. The method according to claim 1 or 2, characterized in that a product characteristic variable set (w") is formed from two product characteristics (w) to be controlled, and a prioritization of the product characteristics (w) to be controlled is performed with respect to preferential control of one of the product characteristics (w).

20. 3. The method according to claim 1 or 2, characterized in that at least one product characteristic (w) is particle size and / or particle moisture and / or particle composition.

21. 3. The method of claim 1, wherein the optimization model is based on or corresponds to the process model.

22. 3. The method according to claim 1 or 2, characterized in that the process model for calculating the process model product characteristic value (w'c) is based on a linear or non-linear process model of the fluidization process to be controlled, and an artificial neural network is used as the non-linear process model.

23. 23. The method of claim 22, wherein the artificial neural network is formed as a multi-layer perceptron, or as a simple recurrent network, or as a non-linear autoregressive exogenous network.

24. 3. The method according to claim 1 or 2, characterized in that as process parameter (p) a plurality of process parameters (p) are used which are selected from the group consisting of the atomizing gas pressure, and / or the atomizing rate, and / or the atomizing amount, and / or the particle temperature, and / or the drying gas temperature at the inlet of the fluidizing device, and / or the relative humidity of the drying gas at the outlet, and / or the drying gas volumetric flow rate.

Images