CN106056243B - A kind of control system and method for high consistency refining system output fiber fractions distribution - Google Patents
A kind of control system and method for high consistency refining system output fiber fractions distribution Download PDFInfo
- Publication number
- CN106056243B CN106056243B CN201610367551.0A CN201610367551A CN106056243B CN 106056243 B CN106056243 B CN 106056243B CN 201610367551 A CN201610367551 A CN 201610367551A CN 106056243 B CN106056243 B CN 106056243B
- Authority
- CN
- China
- Prior art keywords
- iterative learning
- learning control
- output
- probability density
- density function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000000835 fiber Substances 0.000 title claims abstract description 118
- 238000009826 distribution Methods 0.000 title claims abstract description 98
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000007670 refining Methods 0.000 title claims abstract description 21
- 230000006870 function Effects 0.000 claims abstract description 147
- 230000008859 change Effects 0.000 claims description 26
- 239000013598 vector Substances 0.000 claims description 26
- 230000007246 mechanism Effects 0.000 claims description 12
- CIWBSHSKHKDKBQ-JLAZNSOCSA-N Ascorbic acid Chemical compound OC[C@H](O)[C@H]1OC(=O)C(O)=C1O CIWBSHSKHKDKBQ-JLAZNSOCSA-N 0.000 claims 2
- ZZZCUOFIHGPKAK-UHFFFAOYSA-N D-erythro-ascorbic acid Natural products OCC1OC(=O)C(O)=C1O ZZZCUOFIHGPKAK-UHFFFAOYSA-N 0.000 claims 1
- 229930003268 Vitamin C Natural products 0.000 claims 1
- 235000019154 vitamin C Nutrition 0.000 claims 1
- 239000011718 vitamin C Substances 0.000 claims 1
- 238000004537 pulping Methods 0.000 abstract description 31
- 230000008569 process Effects 0.000 abstract description 22
- 238000003775 Density Functional Theory Methods 0.000 abstract description 3
- 238000005259 measurement Methods 0.000 abstract description 3
- 239000003245 coal Substances 0.000 abstract 2
- 241001269238 Data Species 0.000 abstract 1
- 238000004519 manufacturing process Methods 0.000 description 10
- 239000000123 paper Substances 0.000 description 6
- 238000005070 sampling Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005265 energy consumption Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 3
- 238000010009 beating Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 229920001131 Pulp (paper) Polymers 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 206010061592 cardiac fibrillation Diseases 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000018044 dehydration Effects 0.000 description 1
- 238000006297 dehydration reaction Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000002600 fibrillogenic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000011087 paperboard Substances 0.000 description 1
- 238000004886 process control Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Theoretical Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Feedback Control In General (AREA)
- Educational Administration (AREA)
- Manufacturing & Machinery (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Crushing And Grinding (AREA)
Abstract
The present invention provides the control system and method for a kind of high consistency refining system output fiber fractions distribution, which includes: data acquisition unit and output fiber fractions distribution control unit.This method comprises: highly concentrated abrasive disk space, the feeding coal of measurement high consistency refining system, give the probability density function of target fibers fractions distribution；The control of two close cycles iterative learning is carried out to the probability density function of output fiber fractions distribution, ideal highly concentrated abrasive disk space, feeding coal is obtained, is sent to high consistency refining system actuator.The present invention is based on the high concentration plate mill input and output related datas of sensor measurement on chemical-mechanical pulping process line, probability density function theory is approached in conjunction with random distribution Gaussian bases and establishes the state-space model between high consistency refining system output fiber fractions distribution PDF and the main input variable of disc mill, construct double-closed-loop control device, it realizes effective control to slurrying output fiber fractions distribution PDF, the given PDF of output fiber form PDF tracking is changed.
Description
Technical Field
The invention relates to the technical field of pulping and papermaking process control, in particular to a control system and a control method for output fiber form distribution of a high-consistency pulping system.
Background
The paper industry is closely related to national economic development and social civilization, and the consumption level of paper and paperboard is one of important marks for measuring the modernization and civilization degree of a country. The whole paper making process consists of two major links of pulping and paper making. The main function of the pulping link is to produce fibers with specific forms from plant fiber raw materials; the function of the paper-making link is mainly to produce various paper products by using fibers in specific forms as raw materials. The two major links of pulping and papermaking both need to consume a large amount of energy, and particularly the high-consistency millstone pulping link of the pulping process is an important link in the pulping and papermaking production process. The pulp is treated by the pulping machine to obtain finished pulp, the quality of finished paper is directly influenced, the pulping not only has huge energy consumption, but also influences the dehydration efficiency and power consumption of the paper pulp in the subsequent papermaking process, so that the domestic and foreign great attention is paid to the control of the pulping process to improve the quality of the finished pulp.
The high consistency refining process has characteristics of multivariable, strong coupling and nonlinearity, so that the mechanism analysis, modeling and control of the high consistency refining process have great difficulty. At present, the research of a mechanism model of a pulping process is seriously lagged behind the requirements of production and control practice, and mechanism hypothesis models widely used for pulping control at home and abroad are a brooming theory, a specific edge load theory and a specific surface load, but most of control models are still single-variable models which are not enough for representing the whole pulping process, and a commonly accepted mechanism model is not found yet. The previous research mainly focuses on the low-consistency refining process, the single-disc refiner and the improvement of the grinding disc, the hypothesis of the research is strong, and the obtained high-consistency refining model is lack of universality. The method has an important relation with the fact that direct operation data such as blade spacing, fiber length, wet weight, beating degree and other fiber characteristic data cannot be obtained on site, and is also a big disadvantage and limitation of mechanism modeling. Throughout the research on the operational control of the refining process at home and abroad, the research is mainly carried out by using a mechanism model, using a single disc mill as an object and aiming at controlling the average value of the fiber length.
Recent research shows that the energy-saving and consumption-reducing papermaking pulping optimization firstly needs to solve the operation optimization control problem of fiber form distribution in the pulping process. The energy consumption of the pulping process and the quality of the generated fiber form distribution (such as the probability density function shape of the fiber length) are directly related to the energy consumption and the product quality of the subsequent papermaking link, and the dewatering efficiency and the power consumption in papermaking are further influenced. At present, no research report on the closed-loop control of the fiber morphology distribution by directly using the measured value of the fiber morphology distribution as a feedback signal exists. However, the distribution of the output fiber shape of the disc refiner beating does not accord with the Gaussian distribution, and the distribution probability density function can not be modeled and controlled by the variance and the mean value. Mainly because the fiber bundle is transversely extruded and longitudinally broomed by the millstone, the fibrillation is gradually dissociated into single fibers, the output fiber form has strong randomness and uncertainty, and the output fiber form cannot be characterized by a single variable due to the limitation of a measuring instrument, so that the modeling and the control of the fiber form become extremely difficult. Thus achieving a characterization and measurement of the output fiber morphology will have an important impact on the actual production.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a control system and a control method for the output fiber form distribution of a high-consistency refining system.
The technical scheme of the invention is as follows:
a system for controlling the output fiber profile of a high consistency refining system comprising:
a data acquisition unit: obtaining the measured clearance and feeding quantity of a high-consistency grinding disc of the high-consistency grinding system, and giving a probability density function of the target fiber form distribution;
an output fiber form distribution control unit: and taking the clearance and the feeding amount of the high-consistency grinding disc as input, taking the probability density function of the target fiber form distribution as output, and performing double-closed-loop iterative learning control on the probability density function of the output fiber form distribution to obtain ideal control input of the high-consistency grinding system, namely the ideal clearance and the feeding amount of the high-consistency grinding disc, and sending the ideal control input to an execution mechanism of the high-consistency grinding system.
The output fiber form distribution control unit comprises:
the inner loop iterative learning control module: selecting a group of Gaussian basis functions to approximate the probability density function of the target fiber form distribution, and performing iterative learning to control the performance index through a minimized inner ring to obtain an optimal group of Gaussian basis functions;
the outer loop iterative learning control module: taking the measured high-concentration millstone gap and the measured feeding amount as initial outer ring iterative learning control input, measuring the probability density function of output fiber form distribution at the current moment, decoupling the probability density function of the output fiber form distribution at the current moment by using the optimal set of Gaussian functions obtained by the inner ring iterative learning control module to obtain the output weight vector at the current moment, identifying a subspace to obtain a state space equation of the outer ring iterative learning control input and output, and iteratively learning to obtain the ideal high-concentration millstone gap and feeding amount.
The method for controlling the output fiber form distribution of the high-consistency refining system by adopting the control system comprises the following steps:
measuring the clearance and feeding quantity of a high-consistency grinding disc of a high-consistency grinding system, and giving a probability density function of target fiber form distribution;
taking the clearance and the feeding quantity of the high-concentration grinding disc as input, taking the probability density function of the target fiber form distribution as output, and carrying out double closed-loop iterative learning control on the probability density function of the output fiber form distribution to obtain ideal control input of a high-concentration grinding system, namely the ideal clearance and the feeding quantity of the high-concentration grinding disc;
and sending the ideal gap and feeding amount of the high-consistency grinding disc to an actuating mechanism of the high-consistency grinding system, and realizing the probability density function of the output fiber form distribution tracking target fiber form distribution.
The method for controlling the double closed loop iterative learning comprises the following steps:
inner loop iterative learning control: selecting a group of Gaussian basis functions to approximate the probability density function of the target fiber form distribution, and performing iterative learning to control the performance index through a minimized inner ring to obtain an optimal group of Gaussian basis functions;
and (3) outer loop iterative learning control: taking the measured high-concentration millstone gap and the measured feeding amount as initial outer ring iterative learning control input, measuring the probability density function of output fiber form distribution at the current moment, decoupling the probability density function of the output fiber form distribution at the current moment by using the optimal group of Gaussian basis functions to obtain the output weight vector at the current moment, identifying a subspace to obtain a state space equation of the outer ring iterative learning control input and output, and iteratively learning to obtain the ideal high-concentration millstone gap and the feeding amount.
The specific method for the inner loop iterative learning control is as follows:
selecting a group of initial Gaussian basis functions to decouple the probability density function of the target fiber form distribution, and converting the probability density function into a weight vector corresponding to the initial Gaussian basis functions, namely an initial output weight vector;
multiplying the initial Gaussian function by the corresponding weight vector, namely outputting initial inner loop iterative learning control;
if the inner loop iterative learning control performance index meets the requirement, the current Gaussian base function is the optimal Gaussian base function, the final inner loop iterative learning control output, namely the optimal Gaussian base function, is obtained, and outer loop iterative learning control is carried out; otherwise, determining the inner loop iterative learning rate according to the inner loop iterative learning control performance index to obtain the change rate of the Gaussian function position parameter and the change rate of the shape parameter;
and superposing the change rate of the position parameter and the change rate of the shape parameter of the Gaussian basis function on the position parameter and the shape parameter of the initial Gaussian basis function, updating the initial Gaussian basis function, decoupling the probability density function of the target fiber form distribution again until the inner loop iterative learning control performance index meets the requirement, wherein the current Gaussian basis function is the optimal Gaussian basis function, and the final inner loop iterative learning control output, namely the optimal Gaussian basis function, is obtained.
The specific method for the outer loop iterative learning control is as follows:
taking the measured high-concentration millstone gap and the measured feed amount as initial outer ring iterative learning control input, measuring the control output at the current moment, namely outputting a probability density function of fiber form distribution, decoupling the probability density function of the output fiber form distribution by using an optimal Gaussian basis function obtained by inner ring iterative learning control to obtain an output weight vector, assuming that the output weight vector is n-dimensional, selecting the first n-1 output weight vectors as the output of the outer ring iterative learning control, and then performing subspace identification to obtain a state space equation of the outer ring iterative learning control input output;
if the outer ring iterative learning control performance index meets the requirement, the current high-concentration grinding disc gap and the feeding amount are the ideal high-concentration grinding disc gap and the ideal feeding amount; otherwise, determining the outer loop iterative learning rate according to the outer loop iterative learning control performance index, and further obtaining the outer loop iterative learning control input change rate;
and (4) superposing the change rate of the outer ring iterative learning control input to the initial outer ring iterative learning control input to serve as a new outer ring iterative learning control input, returning to perform subspace identification again until the outer ring iterative learning control performance index meets the requirement, and obtaining the ideal outer ring iterative learning control input, namely the ideal high-concentration grinding disc gap and the ideal feeding amount.
Has the advantages that:
the method is based on the relevant input and output data of the high consistency disc mill measured by a sensor on a production line of the chemical mechanical pulping process, combines a random distribution Gaussian function approximation probability density function theory and applies a subspace modeling method to establish a state space model between the output fiber form distribution PDF of the high consistency pulping system and the main input variable of the disc mill, and then constructs a double closed-loop controller by iterative learning control to realize the effective control of the pulping output fiber form distribution PDF, thereby guiding the actual production operation, and tracking the fiber form PDF of the output pulp to the given PDF change within the error allowable range. This is advantageous in controlling the fluctuation of the quality of the resulting pulp within a desired range, while significantly reducing the energy consumption, and is of great significance in the actual production.
Drawings
FIG. 1 is a structural diagram of a high consistency disc grinder in a pulping process and input and output variables;
FIG. 2 is a flow chart of a dual closed-loop iterative learning control in accordance with an embodiment of the present invention;
FIG. 3 is a system block diagram in accordance with an embodiment of the present invention;
FIG. 4 is a graph of the output fiber shape distribution probability density function according to an embodiment of the present invention;
FIG. 5 is a graph comparing the position parameters of the basis functions after the last iteration control of the inner loop with the initial basis functions in accordance with an embodiment of the present invention;
FIG. 6 is a graph illustrating changes in the center values of basis functions during the 250 inner loop iteration control process in accordance with an embodiment of the present invention;
FIG. 7 is a graph of change in basis function width during 250 inner loop iteration controls in accordance with an embodiment of the present invention;
FIG. 8 is a graph of inner loop iterative learning control performance index variation in accordance with an embodiment of the present invention;
FIG. 9 is a diagram illustrating the change of the weights of Gaussian functions in the last outer loop iteration control cycle in accordance with an embodiment of the present invention;
FIG. 10 is a graph illustrating outer loop iterative learning control performance indicator variation in an embodiment of the present invention;
FIG. 11 is a graph showing the effect of the present invention on controlling the output fiber profile PDF of a high consistency mill.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings.
In the high consistency disc mill configuration of the pulping process as shown in FIG. 1, the input variable is the feed rate u1(t) (t/min), high disc gap u2(t) (mm), the output variable is the probability density function PDF (ProbalanceDensity function) of the output fiber morphology distribution. In the embodiment, 1500 groups of disc mills are selected to control the output fiber shape distribution of the high-consistency pulping system by acquiring the online real-time data of the online fiber shape measuring instrument fiber vision of the APMP pulping production line of a factory.
A system for controlling the output fiber profile of a high consistency refining system, as shown in fig. 3, comprising:
a data acquisition unit: obtaining the measured clearance and feeding quantity of a high-consistency grinding disc of the high-consistency grinding system, and giving a probability density function of the target fiber form distribution;
an output fiber form distribution control unit: and taking the clearance and the feeding amount of the high-consistency grinding disc as input, taking the probability density function of the target fiber form distribution as output, and performing double-closed-loop iterative learning control on the probability density function of the output fiber form distribution to obtain ideal control input of the high-consistency grinding system, namely the ideal clearance and the feeding amount of the high-consistency grinding disc, and sending the ideal control input to an execution mechanism of the high-consistency grinding system.
An output fiber morphology distribution control unit comprising:
the inner loop iterative learning control module: selecting a group of Gaussian basis functions to approximate the probability density function of the target fiber form distribution, and performing iterative learning to control the performance index through a minimized inner ring to obtain an optimal group of Gaussian basis functions;
the outer loop iterative learning control module: taking the measured high-concentration millstone gap and the measured feeding amount as initial outer ring iterative learning control input, measuring the probability density function of output fiber form distribution at the current moment, decoupling the probability density function of the output fiber form distribution at the current moment by using the optimal Gaussian function obtained by the inner ring iterative learning control module to obtain the output weight vector at the current moment, identifying a subspace to obtain a state space equation of the outer ring iterative learning control input and output, and iteratively learning to obtain the ideal high-concentration millstone gap and the feeding amount.
The method for controlling the output fiber form distribution of the high-consistency pulping system by adopting the control system comprises the following steps:
step 1, measuring the high-consistency grinding disc gap u of the high-consistency grinding system at the moment t2(t) feed amount u1(t) giving a probability density function γ (y) of the target fiber morphology distribution as shown in fig. 4; the feeding amount is measured by measuring the rotating speed of the feeding screw, and the high-consistency millstone clearance is indirectly measured by measuring parameters such as pressure, load, temperature, vibration and the like in a refining area, or is realized by using a high-precision distance measuring sensor.
Step 2, setting the high-concentration grinding disc gap u at the moment t2(t) feed amount u1(t) as input U ═ U1(t)；u2(t)]Taking the probability density function gamma (y) of the target fiber form distribution as output, and performing double closed-loop iterative learning control on the probability density function of the output fiber form distribution to obtain the ideal control input of the high-consistency refining system, namelyThe ideal high-concentration grinding disc gap and feeding amount are shown in the figure 2;
the method for controlling the double closed loop iterative learning comprises the following steps:
step 2-1, inner loop iterative learning control: selecting a group of Gaussian basis functions to approximate the probability density function of the target fiber form distribution, and performing iterative learning to control the performance index through a minimized inner ring to obtain an optimal group of Gaussian basis functions;
the specific method for inner loop iterative learning control is as follows:
step 2-1-1, defining intervals [ a, b ] in probability density function of target fiber form distribution](mm) selecting an initial set of Gaussian basis functions R (y) R1(y)，R2(y)，…，Rn(y)]Decoupling the probability density function of the target fiber shape distribution, and converting the probability density function into a weight vector corresponding to the initial Gaussian function, namely an initial output weight vector V [ [ omega ] ]1，ω2，…，ωn]；
In this embodiment, a is 0, b is 2, n is 3, i is 1, 2, … … n, and the basis functions are:wherein, muiCharacterizing a position parameter, δ, for the central value of the basis functioni 2The shape parameters are characterized for the width of the basis function. V ═ ω1，ω2，ω3]＝[0.65，1.05，1.45]
A set of initial gaussian basis functions is selected as:
approximating the probability density function of the output fiber morphology distribution with a gaussian basis function:
wherein y represents the output fiber length variable, Ri(y) denotes the ith Gaussian base function, wiRepresents Ri(y) the corresponding weight, e (y) represents the approximation error.
Ignoring the approximation error e (y) is:
the initial output weight vector is: v ═ 1.0195, 0.7722, 1.5641.
Step 2-1-2, multiplying the initial Gaussian basis function R by the corresponding weight vector V, namely outputting gamma' (y) by initial inner loop iterative learning control;
step 2-1-3, if the inner loop iterates to learn and control the performance index Jk，MMeet the requirement of inner loop iterative learning control error precision E1I.e. J1≤E1If the current Gaussian base function is the optimal Gaussian base function, then performing outer loop iterative learning control; otherwise, the performance index J is controlled according to the inner loop iterative learning1Determining α inner loop iterative learning rate to obtain the position parameter mu of Gaussian base function1，μ2，μ3]Of the change rate Δ μ and the shape parameter δ2＝[δ1 2，δ2 2，δ3 2]Rate of change Δ δ of2；
For the purpose of measuring the approximation effect, an inner ring iterative learning control performance index J is definedk，MTo express the error between the inner loop iteration learning control output of the kth sampling point and the probability density function of the target fiber form distribution after the Mth iteration updating:
to ensure that the approximation error gradually decreases with the number of adjustments, the following inequality holds:
wherein,represents the sum of the M +1 th iterative approximation errors.
To derive the inner loop iterative learning rate α, an inner loop iterative learning control performance index J is calculatedk，MIncrement of (d):
controlling performance index J according to inner loop iterative learning1Determining an inner loop iterative learning rate α of 0.001, a Gaussian function position parameter change rate delta mu of α and a shape parameter change rate delta2＝0.5*α。
From the criterion of Lee's stability, it can be known that if the system is kept stable, the delta J is requiredk，MLess than or equal to 0, the following inequality can be obtained:
then let μ + Δ μ and δ2+Δδ2As a new gaussian basis function parameter, the target PDF continues to be approximated.
Step 2-1-4, superposing the change rate of the Gaussian function position parameter and the change rate of the shape parameter on the position parameter and the shape parameter of the initial Gaussian function, and utilizing mu + delta muAnd delta2+Δδ2Updating the initial Gaussian base function, decoupling the probability density function of the target fiber form distribution again until the inner loop iterative learning control performance index meets the requirement, wherein the current Gaussian base function is the optimal Gaussian base function:
fig. 5 is a comparison graph of the initial gaussian basis function position and shape and the gaussian basis function position and shape after the last inner-loop iterative learning control, and fig. 6(a) to (c) are respectively fiber lengths corresponding to the center values of different basis functions in the 250 inner-loop iterative learning control process. Fig. 7(a) - (c) respectively show the fiber lengths corresponding to the widths of different basis functions in the process of 250 times of inner-loop iterative learning control, and finally, the position parameters of the gaussian basis functions are stabilized at 0.4400, 0.9680 and 1.3540 respectively, and the shape parameters of the gaussian basis functions are stabilized at 0.0900, 0.0530 and 0.1300. Fig. 8 is a graph of the change of the control performance index of inner loop iterative learning for 250 times, and it can be seen that the control performance index value of inner loop iterative learning becomes smaller and smaller, and finally tends to 0.
Step 2-2, outer loop iterative learning control: taking the measured high-concentration millstone gap and the measured feeding amount as initial outer ring iterative learning control input, measuring the probability density function of output fiber form distribution at the current moment, decoupling the probability density function of the output fiber form distribution at the current moment by using the optimal Gaussian function to obtain the output weight vector at the current moment, identifying a subspace to obtain a state space equation of the outer ring iterative learning control input and output, and iteratively learning to obtain the ideal high-concentration millstone gap and the feeding amount.
The specific method of the outer loop iterative learning control is as follows:
step 2-2-1, taking the measured high-concentration grinding disc gap and the measured feeding amount as initial outer ring iterative learning control input U ═ U1(t)；u2(t)]Measuring the control output at the present momentThat is, the probability density function of the fiber form distribution is output, then the optimal gaussian basis function obtained by inner loop iterative learning control is utilized to decouple the probability density function of the fiber form distribution to obtain an output weight vector V', since 3 basis function weights have correlation, and the integral of the probability density function on an output definition interval is 1, only two of the weights need to be controlled, the output weight vector of the embodiment is n-3 dimensions, and the first n-1-2 output weight vectors V are selected0＝[ω1，ω2]＝[1.2291，0.3410]As the output of the outer loop iterative learning control, performing subspace identification to obtain a state space equation of the input and the output of the outer loop iterative learning control;
performing subspace identification on input and output data of outer loop iterative learning control to obtain an input and output state space equation as follows:
xj+1，N＝Axj，N+BUj，N
V′j，N＝Cxj，N+DUj，N
wherein A, B, C, D are all state equation parameter matrixes, xj，NIs a state variable, Uj，NIs an outer loop iterative learning control input, V'i，NAnd outputting a weight vector for the outer loop iteration learning control, wherein j represents the jth sampling moment, and N represents the outer loop iteration times, which is recorded as an outer loop iteration period.
The state space matrix obtained by identifying the state space model obtained by the input and output data is respectively
The number of sampling instants per outer loop iteration period is set to 30, i.e., j is 1, 2, 3, …, 30.
The time domain is divided into W iteration cycles in the whole outer loop iterative learning control process, N represents the Nth iteration cycle, and N belongs to [0, 1, 2, …, W ]. j represents the jth sampling instant of the nth iteration cycle. The output weight value of the jth sampling moment in the nth iteration period can be obtained as follows:
the above equation can be expressed in matrix form:
V′j，N＝V0+GUj，N
wherein, V'j，NIs the output weight, u, of the Nth iteration cyclej，NIs the input of the nth iteration cycle.
Selecting the following outer loop iterative learning control performance indexes:
wherein Q and Z are predefined positive definite matrices.
If each iteration cycle has j sampling moments, the outer-loop iterative learning control performance index is expressed as follows:
in order to ensure the monotonous decrease of the performance index of the outer loop iterative learning control, the derivative of the stability index is only required to be smaller than zero according to the Lyapunov stability criterion. The above equation is thus derived:
therefore, monotonic convergence of the outer loop iterative learning control can be ensured, and the input of the outer loop iterative learning control in the k +1 th iteration cycle can be obtained as follows:
UN+1＜UN-[GTQG+Z]-1GTQeN
the ideal outer loop iterative learning control input satisfies the relation:
UN+1＝UN+ΔU(N)
so as to ensure that:
ΔU(N)＜-[GTQG+Z]-1GTQeN
where Q ═ I, Z ═ 0.002 ×, I, is defined as an identity matrix with dimension 60.
Step 2-2-2, if the outer loop iterative learning control performance index J2＝||ek+1||2 Q+||UN+1-UN||2 ZMeet the requirement of outer loop iterative learning control error precision E2I.e. J2≤E2If the current high-concentration grinding disc gap and the feeding amount are the ideal high-concentration grinding disc gap and the ideal feeding amount, executing the step 3; otherwise, iterative learning controllability according to outer loopEnergy index J2Determining the change rate delta U of the outer loop iterative learning control input, adjusting the input, and continuing to perform outer loop iteration:
and 2-2-3, superposing the change rate of the outer ring iterative learning control input to the initial outer ring iterative learning control input, namely U + delta U, serving as a new outer ring iterative learning control input, returning to perform subspace identification again until the performance index of the outer ring iterative learning control meets the requirement, and obtaining ideal outer ring iterative learning control input, namely ideal high-concentration grinding disc gap and feeding amount.
Fig. 9 is a graph of weight variation in the last outer-loop iteration control period, and it can be seen that as the iteration progresses, the weight gradually tends to be stable, i.e. the target weight is tracked without error, which is also one of the characteristics of the iterative learning control.
Fig. 10 is a graph of the change of the performance index of the outer-loop iterative learning control, and as the number of iterations increases, the performance index becomes smaller, that is, the control error becomes smaller.
FIG. 11 is a control effect diagram of the probability density function of the output fiber shape distribution of the high consistency mill, and the final output fiber shape distribution PDF completely tracks the target PDF.
And 3, sending the ideal high-consistency grinding disc gap and the feeding amount to an actuating mechanism of the high-consistency grinding system, and realizing that the probability density function of the output fiber form distribution tracks the probability density function of the target fiber form distribution.
The method is based on the relevant input and output data of the high consistency disc mill measured by a sensor on a production line of the chemical mechanical pulping process, combines a random distribution Gaussian function approximation probability density function theory and applies a subspace modeling method to establish a state space model between the output fiber form distribution PDF of the high consistency pulping system and the main input variable of the disc mill, and then constructs a double closed-loop controller by using an iterative learning control algorithm to realize the effective control of the pulping output fiber form distribution PDF, thereby guiding the actual production operation, and tracking the fiber form PDF of the output pulp to the given PDF change within the error allowable range. The method is favorable for controlling the fluctuation of the quality of the finished pulp in an expected range, provides key quality indexes for the optimized operation and running of the pulping process, and has great significance in actual production.
It should be understood that the detailed description of the present invention is only for illustrating the present invention and is not limited by the technical solutions described in the embodiments of the present invention, and those skilled in the art should understand that the present invention can be modified or substituted equally to achieve the same technical effects; as long as the use requirements are met, the method is within the protection scope of the invention.
Claims (5)
1. A system for controlling the output fiber profile of a high consistency refining system comprising:
a data acquisition unit: obtaining the measured clearance and feeding quantity of a high-consistency grinding disc of the high-consistency grinding system, and giving a probability density function of the target fiber form distribution;
an output fiber form distribution control unit: taking the gap and the feeding amount of the high-concentration grinding disc as input, taking the probability density function of the target fiber form distribution as output, performing double closed-loop iterative learning control on the probability density function of the output fiber form distribution to obtain ideal control input of the high-concentration grinding system, namely the ideal gap and the feeding amount of the high-concentration grinding disc, and sending the ideal control input to an execution mechanism of the high-concentration grinding system;
the output fiber form distribution control unit comprises:
the inner loop iterative learning control module: selecting a group of Gaussian basis functions to approximate the probability density function of the target fiber form distribution, and performing iterative learning to control the performance index through a minimized inner ring to obtain an optimal group of Gaussian basis functions;
the outer loop iterative learning control module: taking the measured high-concentration millstone gap and the measured feeding amount as initial outer ring iterative learning control input, measuring the probability density function of output fiber form distribution at the current moment, decoupling the probability density function of the output fiber form distribution at the current moment by using the optimal set of Gaussian functions obtained by the inner ring iterative learning control module to obtain the output weight vector at the current moment, identifying a subspace to obtain a state space equation of the outer ring iterative learning control input and output, and iteratively learning to obtain the ideal high-concentration millstone gap and feeding amount.
2. A method of controlling the output fiber morphology distribution of a high consistency refining system employing the control system of claim 1, comprising:
measuring the clearance and feeding quantity of a high-consistency grinding disc of a high-consistency grinding system, and giving a probability density function of target fiber form distribution;
taking the clearance and the feeding quantity of the high-concentration grinding disc as input, taking the probability density function of the target fiber form distribution as output, and carrying out double closed-loop iterative learning control on the probability density function of the output fiber form distribution to obtain ideal control input of a high-concentration grinding system, namely the ideal clearance and the feeding quantity of the high-concentration grinding disc;
and sending the ideal gap and feeding amount of the high-consistency grinding disc to an actuating mechanism of the high-consistency grinding system, and realizing the probability density function of the output fiber form distribution tracking target fiber form distribution.
3. The high consistency refining system output fiber morphology distribution control method of claim 2, wherein the method of the double closed loop iterative learning control is as follows:
inner loop iterative learning control: selecting a group of Gaussian basis functions to approximate the probability density function of the target fiber form distribution, and performing iterative learning to control the performance index through a minimized inner ring to obtain an optimal group of Gaussian basis functions;
and (3) outer loop iterative learning control: taking the measured high-concentration millstone gap and the measured feeding amount as initial outer ring iterative learning control input, measuring the probability density function of output fiber form distribution at the current moment, decoupling the probability density function of the output fiber form distribution at the current moment by using the optimal group of Gaussian basis functions to obtain the output weight vector at the current moment, identifying a subspace to obtain a state space equation of the outer ring iterative learning control input and output, and iteratively learning to obtain the ideal high-concentration millstone gap and the feeding amount.
4. The high consistency refining system output fiber morphology distribution control method according to claim 3, characterized in that the specific method of the inner loop iterative learning control is as follows:
selecting a group of initial Gaussian basis functions to decouple the probability density function of the target fiber form distribution, and converting the probability density function into a weight vector corresponding to the initial Gaussian basis functions, namely an initial output weight vector;
multiplying the initial Gaussian function by the corresponding weight vector, namely outputting initial inner loop iterative learning control;
if the inner loop iterative learning control performance index meets the requirement, the current Gaussian base function is the optimal Gaussian base function, the final inner loop iterative learning control output, namely the optimal Gaussian base function, is obtained, and outer loop iterative learning control is carried out; otherwise, determining the inner loop iterative learning rate according to the inner loop iterative learning control performance index to obtain the change rate of the Gaussian function position parameter and the change rate of the shape parameter;
and superposing the change rate of the position parameter and the change rate of the shape parameter of the Gaussian basis function on the position parameter and the shape parameter of the initial Gaussian basis function, updating the initial Gaussian basis function, decoupling the probability density function of the target fiber form distribution again until the inner loop iterative learning control performance index meets the requirement, wherein the current Gaussian basis function is the optimal Gaussian basis function, and the final inner loop iterative learning control output, namely the optimal Gaussian basis function, is obtained.
5. The high consistency refining system output fiber morphology distribution control method of claim 3, characterized in that the specific method of the outer loop iterative learning control is as follows:
taking the measured high-concentration millstone gap and the measured feed amount as initial outer loop iterative learning control input, measuring the control output at the current moment, namely outputting the probability density function of the fiber form distribution, decoupling the probability density function of the output fiber form distribution by using the optimal Gaussian basis function obtained by inner loop iterative learning control to obtain an output weight vector, and assuming the output weight vector asnBefore vitamin C is selectednTaking the 1 output weight vector as the output of the outer loop iterative learning control, and then performing subspace identification to obtain a state space equation of the input and the output of the outer loop iterative learning control;
if the outer ring iterative learning control performance index meets the requirement, the current high-concentration grinding disc gap and the feeding amount are the ideal high-concentration grinding disc gap and the ideal feeding amount; otherwise, determining the outer loop iterative learning rate according to the outer loop iterative learning control performance index, and further obtaining the outer loop iterative learning control input change rate;
and (4) superposing the change rate of the outer ring iterative learning control input to the initial outer ring iterative learning control input to serve as a new outer ring iterative learning control input, returning to perform subspace identification again until the outer ring iterative learning control performance index meets the requirement, and obtaining the ideal outer ring iterative learning control input, namely the ideal high-concentration grinding disc gap and the ideal feeding amount.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610367551.0A CN106056243B (en) | 2016-05-27 | 2016-05-27 | A kind of control system and method for high consistency refining system output fiber fractions distribution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610367551.0A CN106056243B (en) | 2016-05-27 | 2016-05-27 | A kind of control system and method for high consistency refining system output fiber fractions distribution |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106056243A CN106056243A (en) | 2016-10-26 |
| CN106056243B true CN106056243B (en) | 2019-06-25 |
Family
ID=57175621
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610367551.0A Active CN106056243B (en) | 2016-05-27 | 2016-05-27 | A kind of control system and method for high consistency refining system output fiber fractions distribution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106056243B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446236B (en) * | 2018-10-18 | 2021-12-21 | 太原理工大学 | Cement particle size distribution prediction method based on random distribution |
| CN109829177A (en) * | 2018-10-18 | 2019-05-31 | 太原理工大学 | Ball mill overflow granularity is distributed flexible measurement method under a kind of multi-state environment |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2595551C (en) * | 2005-02-11 | 2009-12-08 | Fpinnovations | Method of refining wood chips or pulp in a high consistency conical disc refiner |
| US20070137806A1 (en) * | 2005-12-15 | 2007-06-21 | Honeywell International Inc. | Rapid sampling assembly for thermo-mechanical pulp control application |
| CN104460319B (en) * | 2014-10-20 | 2019-04-12 | 南京工业大学 | Two-dimensional multi-model prediction iterative learning control method for multi-batch intermittent reaction process |
| CN104459089B (en) * | 2014-12-12 | 2016-05-11 | 东北大学 | A kind of flexible measurement method of high consistency refining system freedom |
| CN104915505B (en) * | 2015-06-18 | 2018-01-16 | 东北大学 | A kind of high consistency refining system output fiber fractions distribution PDF modeling methods |
-
2016
- 2016-05-27 CN CN201610367551.0A patent/CN106056243B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106056243A (en) | 2016-10-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106283806B (en) | A kind of high consistency refining system pulp quality control method and system | |
| CN104915505B (en) | A kind of high consistency refining system output fiber fractions distribution PDF modeling methods | |
| CN106056243B (en) | A kind of control system and method for high consistency refining system output fiber fractions distribution | |
| CN104155958A (en) | Optimal control system for chemimechanical pulp technology | |
| CN104459089A (en) | Soft measurement method of freeness of high-consistency pulp grinding system | |
| CN106676934A (en) | Economical and optimal running control method and system for high-concentration grinding system | |
| Li et al. | Nonlinear multiobjective MPC-based optimal operation of a high consistency refining system in papermaking | |
| Chu et al. | Model predictive control and optimization for papermaking processes | |
| CN109695174B (en) | PDF shape prediction method and system for fiber length distribution in pulping process | |
| CN101145750A (en) | Multi-model integrated intelligent control method of large generator group | |
| CN112686538A (en) | Thermal process regulation quality calculation method and device based on data driving | |
| CN101995845A (en) | Field programmable gate array (FPGA)-based auto-leveling control system and method | |
| Zhang et al. | Soft Sensors for Pulp Freeness and Outlet Consistency Estimation in the Alkaline Peroxide Mechanical Pulping (APMP) High-Consistency Refining Process. | |
| Zhou et al. | Energy consumption prediction model of plane grinder processing system based on BP neural network | |
| Gao et al. | Prediction of fiber quality using refining parameters in medium-density fiberboard production via the support vector machine algorithm | |
| Phatwong et al. | Kappa number prediction of pulp digester using LSTM neural network | |
| Li et al. | Research on fuzzy fractional order PID control of liquid temperature in displacement digester | |
| Xiao et al. | Research on the Control Strategy of Battery Roller Press Deflection Device by Introducing Genetic Algorithm to Optimize Integral Separation PID | |
| CN104866725A (en) | Method for predicting steam consumption of cigarette enterprise on the basis of least squares support vector machine | |
| Li et al. | Supervisory predictive control of weighted least square support vector machine based on Cauchy distribution | |
| Hong et al. | Characterising energy efficiency in maching processes: A milling case | |
| Hu et al. | Differential Evolution Algorithm Based Self-adaptive Control Strategy for Fed-batch Cultivation of Yeast. | |
| Chu et al. | Model Predictive Control and Optimization | |
| Talebjedi et al. | Energy Efficiency Analysis of the Refining Unit in Thermo-Mechanical Pulp Mill. Energies 2021, 14, 1664 | |
| CN103616819B (en) | The PID controller design method of calm single-input single-output Systems with Multiple Time-Delays |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |