CN1651567A - Nerve network reverse controller of biological fermentation device and its construction method - Google Patents
Nerve network reverse controller of biological fermentation device and its construction method Download PDFInfo
- Publication number
- CN1651567A CN1651567A CN 200510038098 CN200510038098A CN1651567A CN 1651567 A CN1651567 A CN 1651567A CN 200510038098 CN200510038098 CN 200510038098 CN 200510038098 A CN200510038098 A CN 200510038098A CN 1651567 A CN1651567 A CN 1651567A
- Authority
- CN
- China
- Prior art keywords
- nerve network
- controller
- input
- network reverse
- biological fermentation
- 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.)
- Pending
Links
Images
Landscapes
- Apparatus Associated With Microorganisms And Enzymes (AREA)
Abstract
A nerve network inverse controller for biologic fermentor is such constructed that the difference of its two inputs is the input of a closed-loop controller, one output of said closed-loop controller is connected to the first input of static nerve network and the input of the first integrator whose output is connected to the second input of static nerve nework, and the another output of closed-loop controller is connected to the third input of static nerve network and the input of the second integrator whose output is connected to the fourth input of static nerve network whose output is connected to input of biologic fermentor.
Description
Technical field
The present invention is a kind of nerve network reverse controller and building method of biological fermentation device, is applicable to the high performance control of biological fermentation device in the biotechnology, belongs to the technical field of biotechnology control.
Background technology
Biological fermentation device is an indispensable visual plant in the biotechnology, is the cradle that various biotechnologys are born.For satisfying the needs of producing green food, adopt biological fermentation device can significantly improve the industrial scale and the quality of agricultural biological goods (biological pesticide, bio-feritlizer, biological fodder, fodder additives etc.), so in biotechnology, be with a wide range of applications, progressively be applied in the process of actual microbial fermentation.The multivariate coupled system such as become, non-linear, uncertain when fermentation process is, relate to the growth and breeding process of life entity, mechanism is very complicated.Conventional control methods such as definite value control, PID control are adopted in existing fermenting process control mostly, and its purpose is a steady operation conditions.By observation to some constants of fermenting process (as fermented liquid PH, dissolved oxygen content DO, temperature and pressure), regulate production process by rule of thumb to realize stable operation, if because of infer or regulate improper, light then influence fermentative production, weight then makes the product of whole fermentor tank scrap.
Because the time variation of Fermentation Process of Parameter, diversity, coupling and uncertainty, adopt traditional linear processing methods obviously improper, the gamma controller that some of present research design based on nonlinear control method depends on the mathematical models of fermenting process mostly by force, and the time variation of the uncertainty of fermenting process model and parameter makes the mathematical models of fermenting process be difficult to obtain.So existing gamma controller method of design mostly can only be studied the naive model of fermenting process, so just further reduced its suitability.Just because of above reason, make that up to the present the various gamma controller method of design of fermenting process also are difficult to be applied to actual fermentation equipment.
For the real high performance control that realizes fermenting process, need to adopt some new control techniques designs.
Technology contents
The purpose of this invention is to provide a kind of fermenting process arbitrarily that is suitable for, and can in the bigger operating range of fermenting process, all obtain the nerve network reverse controller and the building method of the biological fermentation device of satisfied control effect.
The nerve network reverse controller of biological fermentation device of the present invention, the structure that it is characterized in that this nerve network reverse controller is: the input terminus at this nerve network reverse controller is connected to closed loop controller, the input terminus of the output termination integrator of closed loop controller and the input terminus of static neural network, another input terminus of the output termination static neural network of integrator, the output of static neural network is the output of this nerve network reverse controller, the output terminal of this nerve network reverse controller of input termination of biological fermentation system.
The static neural networks with 4 input nodes, 2 output nodes that are configured to use of nerve network reverse add the nerve network reverses that 2 integrators formations have 2 input nodes, 2 output nodes, and wherein: first input terminus of static neural network is first input terminus of nerve network reverse; First input that second input terminus of static neural network is nerve network reverse obtains via integrator; The 3rd input terminus of static neural network is second input terminus of nerve network reverse; The 4th input terminus of static neural network is that second input of nerve network reverse obtains via integrator; Static neural network is formed nerve network reverse with 2 integrators, and the output of static neural network is exactly the output of nerve network reverse.
The building method of the nerve network reverse controller of biological fermentation device of the present invention is: thus at first select the feedback quantity of biological fermentation and controlled volume to constitute controlled biological fermentation system; Then adopt the external integrator of static neural network to constitute the nerve network reverse of biological fermentation system, and make the inverse system function of nerve network reverse realization biological fermentation system by the weight coefficient of adjusting static neural network; Then nerve network reverse is serially connected in before the biological fermentation system, nerve network reverse and biological fermentation system are combined into two single order integral form pseudo-linear systems; On this basis, adopt the integrated approach of linear system that two single order integral form pseudo-linear systems are designed closed loop controller respectively; The common nerve network reverse controller that constitutes that at last nerve network reverse and closed loop controller is in series, nerve network reverse controller is serially connected in the biological fermentation system before, system controls to biological fermentation.
The implementation method of nerve network reverse controller is a singlechip controller for adopting chip microprocessor, realizes by the establishment Single Chip Microcomputer (SCM) program.To obtain the fine control performance index.
Principle of the present invention is contrary by constructing neural network, the control problem of this System with Nonlinear Coupling of biological fermentation system is converted into the control problem of simple two single order integral form pseudo-linear systems, correspondingly just can designs linear closed loop controller easily.The present invention adopts static neural network to add the inverse system function that integrator is realized the biological fermentation system, the controller design method that is proposed does not just rely on the mathematical models of fermentation unit and internal object like this, but the measurement signal that only needs feedback biological fermentation system, the biological fermentation controller changes parameter and the robustness of external disturbance thereby can improve greatly, and realization decoupling zero control, be suitable for biological fermentation process arbitrarily simultaneously, realistic engineering application requiring has very strong using value.
The invention has the advantages that:
A. adopt nerve network reverse to realize control to the biological fermentation system, be completely free of existing analysis formula controller design method for the strong dependency of mathematical model and the limitation that can only study the biological fermentation naive model, reduced the biological fermentation process parameter effectively and changed and the influence of external disturbance, improved the control performance of biological fermentation significantly biological fermentation control effect.
B. by constructing the nerve network reverse of biological fermentation system, this control problem non-linear, coupled system of biological fermentation system is converted into the control problem of two single order integral form pseudo-linear systems, and further appropriate design linear closed-loop controller, overcome traditional less limitation of linear controller efficient working range, can in the bigger operating range of biological fermentation system, obtain satisfied control effect.
But the nerve network reverse controller of c. designed biological fermentation system only adopts the measurement signal of biological fermentation, and the engineering application requiring of realistic biological fermentation system is easy to Project Realization.
The present invention can be used for constructing new bio fermentation controller biological fermentation is carried out high performance control, improves the quality of product.
Description of drawings
Fig. 1 is biological fermentation system 1 block diagram.
Fig. 2 is the synoptic diagram and the isoboles thereof of two single order integral form pseudo-linear systems 3 of nerve network reverse 2 and biological fermentation system 1 compound formation.Static neural network 21, integrator 22,23 are wherein arranged; Cell concentration subsystem 31, substrate concn subsystem 32.
Fig. 3 is added to the input signal that biological fermentation system 1 input terminus is used to obtain static neural network 21 training datas.
Fig. 4 is the synoptic diagram at the closed loop controller 4 of two single order integral form pseudo-linear systems, 3 designs.Cell concentration controller 41, substrate concn controller 42 are wherein arranged.
Fig. 5 is nerve network reverse controller 5 synoptic diagram, and closed loop controller 4, static neural network 21, integrator 22 and 23, biological fermentation system 1 are wherein arranged.
Fig. 6 adopts singlechip controller 6 to form synoptic diagram as apparatus of the present invention of nerve network reverse controller 5.Cell concentration transmitter 7, substrate concn transmitter 8, biological fermentation system 1 are wherein arranged.
Fig. 7 adopts singlechip controller 6 to realize system software block diagram of the present invention as nerve network reverse controller 5.
1, biological fermentation system; 2, nerve network reverse; 21, static neural network; 22, integrator; 23, integrator; 3, pseudo-linear system; 31, cell concentration subsystem; 32, substrate concn subsystem; 4, closed loop controller; 5 nerve network reverse controllers; 6, singlechip controller; 7, cell concentration transmitter; 8, substrate concn transmitter
Embodiment
The structure of this inverse controller is: the given (X that is input as controlled volume of nerve network reverse controller 5
*, S
*) with the work output (X of biological fermentation system 1, S), both differences are the input of closed loop controller 41 and 42, first input terminus of the output termination nerve network reverse 2 of closed loop controller 41, first input terminus I1 of first input termination static neural network 21 of nerve network reverse 2 and the input terminus of integrator 22, second input terminus I2 of the output termination static neural network 21 of integrator 22, second input terminus of the output termination nerve network reverse 2 of closed loop controller 42, the 3rd the input terminus I3 of second of nerve network reverse 2 input termination static neural network 21 and the input terminus of integrator 23, the 4th input terminus I4 of the output termination static neural network 21 of integrator 23, the output terminal of the input termination nerve network reverse controller 5 of biological fermentation system 1 also is the output terminal of static neural network 21 simultaneously.
The static neural networks 21 with 4 input nodes, 2 output nodes that are configured to use of nerve network reverse 2 add the nerve network reverses 2 that 2 integrations, 22,23 formations have 2 input nodes, 2 output nodes, and wherein: first input terminus I1 of static neural network 21 is first input terminus of nerve network reverse 2; First input that second input terminus I2 of static neural network 21 is nerve network reverse 2 obtains via integrator 22; The 3rd input terminus I3 of static neural network 21 is second input terminus of nerve network reverse 2; The 4th input terminus I4 of static neural network 21 is that second input of nerve network reverse 2 obtains via integrator 23; Static neural network 21 is formed nerve network reverse 2 with 2 integrators 22,23, and the output of static neural network 21 is exactly the output of nerve network reverse 2.The implementation method of nerve network reverse controller 5 is a singlechip controller 6 for adopting chip microprocessor, realizes by the establishment Single Chip Microcomputer (SCM) program.
The building method of the nerve network reverse controller of biological fermentation device is: thus at first select the feedback quantity of biological fermentation and controlled volume to constitute controlled biological fermentation system 1; Then adopt static neural network 21 to add the nerve network reverse 2 that integrator 22 and 23 constitutes biological fermentation system 1, and make nerve network reverse 2 realize the inverse system function of biological fermentation system 1 by the weight coefficient of adjusting static neural network 21; Then nerve network reverse 2 is serially connected in before the biological fermentation system 1, nerve network reverse 2 is combined into two single order integral form pseudo-linear systems 3 with biological fermentation system 1; On this basis, adopt the integrated approach of linear system to single order integral form pseudo-linear system 3 design closed loop controllers 4; The common nerve network reverse controller 5 that constitutes that at last nerve network reverse 2 and closed loop controller 4 is in series, nerve network reverse controller 5 is serially connected in biological fermentation system 1 before, biological fermentation system 1 is controlled.
Concrete enforcement divides following 7 steps.
1 determines that thereby the feedback quantity and the controlled volume of biological fermentation constitute controlled biological fermentation system.Determine feed rate F and charging substrate concn S
fAs two manipulated variables independently, controlled volume is the biological fermentation system (its block diagram as shown in Figure 1) of cell concentration X and substrate concn S.
2 by analyzing, derive, for the structure of nerve network reverse and learning training provide basis on the method.At first set up the mathematical model of the biological fermentation system that considers the complete dynamic process of biological fermentation, exist through provable its inverse system of deriving, and can determine that four inputs of its inverse system are respectively the first order derivative of cell concentration
The first order derivative of cell concentration X, substrate concn
Substrate concn S, two control signal u that are output as the biological fermentation system
1, u
2Be respectively feed rate F and charging substrate concn S
fNeed to prove, this step only provides basis on the method for the structure of following nerve network reverse and learning training, in concrete enforcement of the present invention, and this step, comprising that theoretical proof that biological fermentation system inverse system is existed and some are corresponding derives etc., can skip.
It is contrary that 3 employing static neural networks add two integrator constructing neural networks.Wherein static neural network adopts 3 layers MLN network, and the input layer number is 4, and the hidden layer node number is 11, and output layer node number is 2, and the hidden layer neuron activation function uses S type hyperbolic tangent function f (x)=(e
2x-e
-2x)/(e
2x+ e
-2x), the neurone of output layer adopts pure linear function f (x)=x, and x is neuronic input, and the weight coefficient of static neural network will be determined in next step off-line learning.Constitute nerve network reverse (shown in the frame of broken lines of the left figure of Fig. 2) with the static neural network with 4 input nodes, 2 output nodes then, wherein first of static neural network is input as first input of nerve network reverse
Second input X of static neural network is first input of nerve network reverse
Obtain via integrator; The 3rd second input that is input as nerve network reverse of static neural network
The 4th input S of static neural network is second input of nerve network reverse
Obtain via integrator.Static neural network is formed nerve network reverse with two integrators, and the output of static neural network is exactly the output of nerve network reverse.
4 adjust the weight coefficient of static neural network.(a) will control input signal (as shown in Figure 3) and be added to the biological fermentation system input terminus of (as shown in Figure 1), gather the input { u of biological fermentation system with 25 minutes sampling periods
1, u
2, output { X, S}, and preserve data { u
1, u
2, X, S}.(b) X and S off-line are asked its first order derivative respectively, and signal is done standardization processing, form the training sample set of static neural network
(c) adopt the error anti-pass BP algorithm that drives quantifier and learning rate changing that static neural network is trained, through 600 training, static neural network output square error meets the demands less than 0.0001, thereby has determined each weight coefficient of static neural network.
5 form compound pseudo-linear system.The nerve network reverse of the biological fermentation system that constructed is serially connected in (shown in the left figure of Fig. 2) before the biological fermentation system, and nerve network reverse and biological fermentation system are combined into two single order integral form pseudo-linear system (s
-1) (shown in the right figure of Fig. 2), thereby the control of complicated nonlinear systems is converted into the control of two simple single order integral form pseudo-linear systems.
6 make the linear closed-loop controller.To single order integral form pseudo-linear system (s
-1) make closed loop controller (as shown in Figure 4).Closed loop controller adopts proportional plus integral plus derivative controller PID, POLE PLACEMENT USING or the most excellent method of quadratic performance in the lineary system theory to design, in the embodiment that the present invention provides, closed loop controller has been selected ratio P controller for use, and its parameter tuning is P1=10, P2=15.
7 form nerve network reverse controller.Nerve network reverse and closed loop controller are formed nerve network reverse controller (shown in the long and short dash line wire frame among Fig. 5) jointly.Can require to adopt different hardware or software to realize according to different control.
Fig. 8 has provided the synoptic diagram of a kind of specific embodiment of the present invention, and wherein nerve network reverse and closed loop controller are that singlechip controller is realized by software by chip microprocessor, and the system program block diagram as shown in Figure 9.
According to the above, just can realize the present invention.
Claims (4)
1, a kind of nerve network reverse controller of fermentation unit is characterized in that the structure of this nerve network reverse controller (5) is: the given (X that is input as controlled volume of nerve network reverse controller (5)
*, S
*) with the work output (X of biological fermentation system (1), S), both differences are the input of closed loop controller (4), first input terminus of the output termination nerve network reverse (2) of the cell concentration controller (41) in the closed loop controller (4), first input terminus (I1) of first input termination static neural network (21) of nerve network reverse (2) and the input terminus of integrator (22), second input terminus (I2) of the output termination static neural network (21) of integrator (22), second input terminus of the output termination nerve network reverse (2) of the substrate concn controller (42) in the closed loop controller (4), the 3rd input terminus (I3) of second input termination static neural network (21) of nerve network reverse (2) and the input terminus of integrator (23), the 4th input terminus (I4) of the output termination static neural network (21) of integrator (23), the output terminal of the input termination nerve network reverse controller (5) of biological fermentation system (1) also is the output terminal of static neural network (21) simultaneously.
2, the nerve network reverse controller of fermentation unit according to claim 1, the static neural networks (21) with 4 input nodes, 2 output nodes that are configured to use that it is characterized in that nerve network reverse (2) add the nerve network reverses (2) that 2 integrators (22,23) formation has 2 input nodes, 2 output nodes, and wherein: first input terminus (I1) of static neural network (21) is first input terminus of nerve network reverse (2); Second input terminus (I2) of static neural network (21) is that first input of nerve network reverse (2) obtains via integrator (22); The 3rd input terminus (I3) of static neural network (21) is second input terminus of nerve network reverse (2); The 4th input terminus (I4) of static neural network (21) is that second input of nerve network reverse (2) obtains via integrator (23); Static neural network (21) is formed nerve network reverse (2) with 2 integrators (22,23), and the output of static neural network (21) is exactly the output of nerve network reverse (2).
3, the nerve network reverse controller of fermentation unit according to claim 1 is characterized in that the implementation method of nerve network reverse controller (5) is singlechip controller (6) for adopting chip microprocessor, realizes by the establishment Single Chip Microcomputer (SCM) program.
4, a kind of building method of nerve network reverse controller of fermentation unit is characterized in that at first selecting the feedback quantity of biological fermentation and controlled volume to constitute controlled biological fermentation system (1); Then adopt the external integrator of static neural network (21) (22,23) to constitute the nerve network reverse (2) of biological fermentation system (1), and make the inverse system function of nerve network reverse (2) realization biological fermentation system (1) by the weight coefficient of adjusting static neural network (21); Then nerve network reverse (2) is serially connected in biological fermentation system (1) before, nerve network reverse (2) is combined into two single order integral form pseudo-linear systems (3) with biological fermentation system (1); On this basis, adopt the integrated approach of linear system to two single order integral form pseudo-linear systems (31,32) design, two closed loop controllers (41,42); The common nerve network reverse controller (5) that constitutes that at last nerve network reverse (2) and closed loop controller (4) is in series, it is preceding that nerve network reverse controller (5) is serially connected in biological fermentation system (1), and biological fermentation system (1) is controlled.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200510038098 CN1651567A (en) | 2005-03-11 | 2005-03-11 | Nerve network reverse controller of biological fermentation device and its construction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200510038098 CN1651567A (en) | 2005-03-11 | 2005-03-11 | Nerve network reverse controller of biological fermentation device and its construction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1651567A true CN1651567A (en) | 2005-08-10 |
Family
ID=34876200
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200510038098 Pending CN1651567A (en) | 2005-03-11 | 2005-03-11 | Nerve network reverse controller of biological fermentation device and its construction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1651567A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101639902B (en) * | 2009-08-12 | 2012-05-23 | 江苏大学 | Modeling method of support vector machine (SVM)-based software measurement instrument in biological fermentation process |
| CN101630376B (en) * | 2009-08-12 | 2012-06-20 | 江苏大学 | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process |
| CN103064292A (en) * | 2013-01-15 | 2013-04-24 | 镇江市江大科技有限责任公司 | Biological fermentation adaptive control system and control method based on neural network inverse |
-
2005
- 2005-03-11 CN CN 200510038098 patent/CN1651567A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101639902B (en) * | 2009-08-12 | 2012-05-23 | 江苏大学 | Modeling method of support vector machine (SVM)-based software measurement instrument in biological fermentation process |
| CN101630376B (en) * | 2009-08-12 | 2012-06-20 | 江苏大学 | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process |
| CN103064292A (en) * | 2013-01-15 | 2013-04-24 | 镇江市江大科技有限责任公司 | Biological fermentation adaptive control system and control method based on neural network inverse |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | A case study of adaptive nonlinear regulation of fed-batch biological reactors | |
| CN1060915A (en) | The universal process control of using artificial neural networks | |
| CN109634108A (en) | The different factor full format non-model control method of the MIMO of parameter self-tuning | |
| CN111515962B (en) | Transmission error compensation control method for flexible joint with harmonic reducer | |
| CN103792845B (en) | A kind of Ferment of DM process mends the method and system of sugared rate optimized control | |
| CN1651567A (en) | Nerve network reverse controller of biological fermentation device and its construction method | |
| Kaisare et al. | Simulation based strategy for nonlinear optimal control: Application to a microbial cell reactor | |
| Bayen et al. | Time-optimal control of concentration changes in the chemostat with one single species | |
| Teissier et al. | A hybrid recurrent neural network model for yeast production monitoring and control in a wine base medium | |
| Yunfeng et al. | On-line fault detection using SVM-based dynamic MPLS for batch processes | |
| CN1234056C (en) | Self-adaptation nonlinear time varying controller and controlling method thereof | |
| CN114609986A (en) | Cement decomposing furnace denitration regulation and control optimization system and method based on predictive control | |
| van Impe et al. | Optimal control of the penicillin G fed‐batch fermentation: An analysis of the model of heijnen et al. | |
| Rodrigues et al. | A tuned approach of the predictive–adaptive GPC controller applied to a fed-batch bioreactor using complete factorial design | |
| CN1648801A (en) | Method for setting robust PID controller parameter using nervous network | |
| Patnaik | Hybrid neural simulation of a fed-batch bioreactor for a nonideal recombinant fermentation | |
| CN1292525C (en) | Neural network inverse controller for controlled series capacitance compensating device and configuration method | |
| Thatipamala et al. | On-line state estimation and adaptive optimization using state equations for continuous production of bioethanol | |
| CN109814388A (en) | The inclined format non-model control method of the different factor of the MISO of parameter self-tuning | |
| CN1431768A (en) | Method for structuring neural network reversal composite controller of variable frequency speed control system with constant ratio of voltage and frequency | |
| CN103048926A (en) | Online neural network inverse controller in biological fermentation process and construction method of controller | |
| CN1932700A (en) | Distributing internal model control system parameter quantitative setting method | |
| CN1275111C (en) | Error feedback anti-saturation static compensation controller and determining method for compensator | |
| Rodríguez-Miranda et al. | Application of a symmetric-send-on-delta event-based controller for a microalgal raceway reactor | |
| Patnaik | Lyapunov comparison of noise‐filtering methods for oscillating yeast cultures |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |