RU2649791C1  Multiparameter fuzzy processor for automatic regulators and method for synthesis of control signal  Google Patents
Multiparameter fuzzy processor for automatic regulators and method for synthesis of control signal Download PDFInfo
 Publication number
 RU2649791C1 RU2649791C1 RU2017111665A RU2017111665A RU2649791C1 RU 2649791 C1 RU2649791 C1 RU 2649791C1 RU 2017111665 A RU2017111665 A RU 2017111665A RU 2017111665 A RU2017111665 A RU 2017111665A RU 2649791 C1 RU2649791 C1 RU 2649791C1
 Authority
 RU
 Russia
 Prior art keywords
 control
 row
 control signal
 calculating
 input
 Prior art date
Links
Images
Classifications

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06F—ELECTRIC DIGITAL DATA PROCESSING
 G06F1/00—Details not covered by groups G06F3/00 – G06F13/00 and G06F21/00
 G06F1/26—Power supply means, e.g. regulation thereof
 G06F1/32—Means for saving power
 G06F1/3203—Power management, i.e. eventbased initiation of powersaving mode
Abstract
Description
The invention relates to the field of realtime control of complex objects and fast processes that cannot be represented by a mathematical model in the form of differential equations. The technical result is an increase in control performance while reducing hardware redundancy, in the unification of the structure of fuzzy controllers when changing the number of parameters of the control object.
A clear logical controller is known for controlling technological processes [Patent RU No. 2445669, IPC G05B 13/02], in which the controller includes a fuzzifier with seven inputs, a logic output unit with predetermined membership functions of clear terms of variables, the input of which is provided with input and output variables, as well as discrete input and output variables of the control object, defuzzifier, actuator, control object and feedback sensor. The comparing device is implemented as part of the conditional part of the production rules of the fuzzy inference unit. To increase accuracy and speed, the input and output variables of the controller are represented by a set of clear terms, and an additional increase in the speed of the controller is achieved by automatically positioning the ANYTIME algorithm to the beginning of the rule production system with the highest response frequency. The expansion of the control functions of the controller is achieved through the use of discrete and output variables of the control object in the antecedents of production rules.
The disadvantages of this controller are: a sequential method of processing production rules reduces the speed of the controller, the use of a rectangular shape of the membership function reduces the accuracy of regulation in the fuzzy region. Known threeposition controllers with an adaptive middle position [Patent RU No. 2220432, IPC G05B 13/02, G05B 11/18], in which the noncontact connection of the functional link reconfiguration is achieved by connecting it to the output of the threeposition controller with fixed positions in parallel with the amplification link with gain K ≥2 and the subsequent supply of the total signal of these two parallel branches to a saturationtype link with a unity gain and saturation signals equal to the values of the extreme positions Foot threeposition control relative to the average.
The disadvantages of this controller are: the use of the integration circuit reconfiguration function, which reduces the speed of regulation when the controlled variable (parameter) runs out of the dead band (zone), the applicability of the controller with only one parameter of the control object.
A known method of forming a fuzzy control action [Patent No. 2514127, IPC G05B 19/00], according to which a fuzzy adaptive positional method for automatically controlling objects with discrete actuators is proposed, implemented by a logical controller and consisting in the formation of control parameters according to fuzzy rules and the supply of these parameters control on the object, and the formation of control parameters is divided into two levels, on the first of which, using fuzzy logic, it is localized (Allocated) control range, in which will be further adaptation, the localization of the control range is performed by determining the main (base) control parameter values of the range according to the formula
, where U _{ad1} is an adaptable control parameter, U _{o} and U _{k} are control parameters in two extreme (opposite) states of the object, normalized equivalent state of the object ( at B _{o} at B _{k} , where B _{o} is the initial state of the object, B _{k} is the final state of the object), taken as the value of the adaptive average position parameter in this range, and at the second level, the value of the control parameter is determined using an adaptive threeposition control method.The disadvantages of this method is the limitation of the characteristics of the state of the control object by one parameter, the sequential processing of production rules that reduces performance, and the use of only a discrete actuator.
The closest in technical essence to the invention are a method and device for constructing fuzzy logic systems [Patent No. 2417442, IPC G06N 7/02], in which a sequence of fuzzy logic rules is first formulated, then a numerical characteristic is assigned to each of these rules — an indicator of control quality, and the fuzzy logic rules are implemented on the basis of a trained neural network, information signals or signals from the control object are fed to the inputs of the neural network, while at its output a sequence of signals or a sequence of instructions and recommendations, where a trained neural network is a trained large artificial neural network, each of the fuzzy logic rules being implemented as a separate fragment of a trained large artificial neural network (domain), where the number of domains corresponds to the number of fuzzy logic rules and, in addition, contains a certain excess number of backup domains, one of the domains acting as an arbiter and commuting the outputs of the domains with the outputs of the neural network, taking into account the indicator achestva management.
The disadvantages of this method and device is the technical complexity of the domain neural network, which is implemented using computer components. Computer components require software processing, which is also a significant drawback of the device, since they significantly reduce the speed achieved by parallelizing the processing of production rules.
The purpose of the present invention:
 improving the speed of computing the value of the control signal;
 simplification of the structure of fuzzy control devices. The technical result of the claimed method and device for its implementation:
 increased performance due to hardware parallelization of the calculation of the value of the fuzzy control function;
 unification of processing units of fuzzy regulation rules.
The technical result in a patented method for synthesizing a control signal in the region of fuzzy values of the control function is achieved by the fact that the original values of the control signal
at discrete points in the region of variation of the parameters U _{ 1 } , U _{ 2 } , U _{ 3 } , ..., U _{ N } of the control object are known in advance as a result of experimental or theoretical studies of the N dimensional control function, the current values of the parameters of the control object U _{ a } _{ 1 } , U _{ a2 are } determined in the control process , ..., U _{ a } _{ N } , using N phase identifiers determine the belonging of the parameters of the control object U _{ a } _{ 1 } , U _{ a } _{ 2 } , ... U _{ a } _{ N to the } intervals U _{1} _{ min } ≤U _{ a } _{ 1 } , <U _{ 1 } _{ max } , U _{2} _{ min } ≤U _{ a } _{ 2 } < U _{2} _{ max } , ..., U _{ N } _{ min } ≤U _{ a } _{ N } <U _{ N } _{ max } between adjacent discrete points of the area of variation of the parameters of the control object, in the interval ale [ U _{1} _{ min } , U _{1max} ) calculate 2 ^{ N } ^{1} intermediate values of the control signal , and the desired value of the control signal is in the same proportion to the pair of values of the control signal and , as well as the value of U _{ a } _{ 1 } to the values of U _{1} _{ min } and U _{ 1 } _{ max } respectively, where with an even value of the integer part of the expression and when odd, j varies from 2 to N , in the interval [ U _{ j } _{ min } , U _{ j } _{ max } ] calculate 2 ^{ N } ^{j} intermediate values of the control signal , and the desired value of the control signal is in the same proportion to a pair of intermediate values of the control signal and , as well as the value of U _{ aj } to the values of U _{ j } _{ min } and U _{ j } _{ max, } respectively, at the Nth step, the final value of the control signal F _{y is} calculated, and the desired value of the control signal F _{y} is in the same proportion to the pairs of values of the control signal and as well as the value of U _{ a } _{ N } to the values of U _{ N } _{ min } and U _{ N } _{ max } respectively, the control signal F _{y is} transmitted to the object control device.The technical result in the patented multiparameter fuzzy processor is achieved by the fact that it contains N fuzzifiers, a memory block, N rows of control signal computing devices, the first row contains 2 ^{ N } ^{  } ^{ 1 } control signal computing devices, jth row 2 ^{ N } ^{  } ^{ j } control signal computing devices, the first input of the jth fuzzifier is connected to the jth input of the multiparameter fuzzy controller and to the third input of each of the devices for calculating the control signals of the jth row, inputs 2 through K +1 of the jth fuzzifier are connected with correspondingly to the outputs from ( j 1) ⋅ K +1 to j ⋅ K of the memory block, the second output of the jth fuzzifier is connected to the first input of each device for computing control signals of the jth row and to the jth input of the memory block, the first output the jth fuzzifier is connected to the second input of each device for calculating the control signals of the jth row, the output N × K +1 of the memory block is connected to the fourth input of the first device for calculating the control signals of the first row, the output N × K +2 of the memory block is connected to the fifth input first control signal computing device s of the first row, (N × K +2 i 1 ) th output of the storage unit is connected to the fourth input of ith control signal calculating unit of the first row, (N × K +2 i) th output of the storage unit is connected to a fifth input the i th device for calculating the control signals of the first row ( j varies from 1 to N , i varies from 1 to 2 ^{N1} ), the output of the first device for calculating the control signals of the first row is connected to the fourth input of the first device for calculating the control signals of the second row, the output of the second devices for computing control signals of the first row under for prison to the fifth input of the first control signal calculation device of the second row output (2 i 1) th calculating unit control signals jth row is connected to a fourth input of the i th control signal calculation device (j +1) th row, yield 2 j th device for calculating the control signals of the j th row is connected to the fifth input of the i th device for calculating the control signals of the ( j +1) th row, the output of the device for calculating the control signals of the N th row is connected to the output of a multiparameter fuzzy processor.
In the patented method, the values of the control signal
obtained as a result of experimental or theoretical studies of the behavior of the control object at discrete points of the N dimensional region of the change in the values of the set of parameters of the control object { U _{1} , U _{2} , U _{3} , ..., U _{ N } }, where k _{ i } is the number of the parameter U _{ i } in the discrete models of the N dimensional domain. The instantaneous values of the parameters U _{ a } _{ 1 } , U _{ a } _{ 2 } , ..., U _{ a } _{ N } are received from the control object. Using the fuzzifier, the intervals U _{1} _{ min } ≤ U _{ a } _{ 1 } < U _{1} _{ max } , U _{2} _{ min } ≤ U _{ a } _{ 2 } < U _{2} _{ max } , ... , U _{ N } _{ min } ≤ U _{ a } _{ N } ≤ U _{ N } _{ max } belong to the values of the parameters of the control object with respect to neighboring discrete points of the N dimensional region. In the interval [ U _{1} _{ min } , U _{1} _{ max } ] using 2 ^{ N } ^{1} devices for calculating the control signals of the first row, 2 ^{ N } ^{1} values of the control signal are calculated by the formula,
where i is the number of the control signal computing device in the first row,
,  extreme values of the parameters of the control object, limiting the selected subdomain in the discrete model of the N dimensional region,,
In the interval [ U _{ j } _{ min } , U _{ j } _{ max } ] using 2 ^{ N } ^{} ^{ j } devices for calculating the control signals of the jth row, 2 ^{ N } ^{  } ^{ j } values of the control signal F are calculated by the formula
where i is the number of the control signal computing device in the jth row ( j varies from 2 to N ). Using the device for calculating control signals in the Nth row, the final value of the control signal F _{y is} calculated, which is transmitted to the object control device.
The structural diagram of a multiparameter processor (hereinafter referred to as the processor), intended for control systems and automatic control of technical and technological objects, is shown in figure 1. The device contains N fuzzifiers 1, N rows of control signal calculating devices (DCS) 2, a memory block 3. In the first row there are 2 ^{ N } ^{1} UVUS 2, in the jth row 2 ^{ N } ^{  } ^{ j } UVUS. Each row of UVUS is connected with its corresponding fuzzifier number.
Consider the processor. Before starting work in the control mode, the values of the control signal are stored in the processor memory block 3, which are received and recorded in advance according to the results of experimental or theoretical studies of the N dimensional control function F. Each discrete point is represented in the memory block by a tuple.
where F is the value of the control signal, and  values of the parameters of the control object. Tuples define a set of rules for choosing the value of a control signal at discrete points. Each parameter of the control object U _{ j } can take one of the fixed values at discrete points where K is the number of rules for each parameter. Between the discrete points of the control area are areas of uncertainty in which the control signal takes fuzzy values. The task of the processor is to determine the values of the control signal F in the areas of uncertainty.In control mode, the current analog values of the parameters of the control object are received at the processor inputsU _{ a } _{ one },U _{ a } _{ 2 }, ...,U _{ a } _{ N }that are transmitted to the first inputs 1_{one} fuzzifiers corresponding to parameter number 1. Inputs from 1_{2}1st to_{K + 1} jth fuzzifier are connected respectively to outputs from 4_{(} _{ j } _{1) ⋅} _{ K } _{+1},..., four_{ j } _{⋅} _{ K } memory blocks from which analog values come
parameters of the control object at discrete points of the change regionjth parameter of the control object. Using fuzzifiers 1 determine the ownership of each parameter of the control objectU _{ a } _{ one },U _{ a } _{ 2 }, ...,U _{ a } _{ N } intervalsU _{one} _{ min }≤U _{ a } _{ one }<U _{one} _{ max },U _{2} _{ min }≤U _{ a } _{ 2 }<U _{2} _{ max }, ...,U _{ N } _{ min }≤U _{ a } _{ N }<U _{ N } _{ max } between adjacent discrete points of the area of change of the parameters of the control object. From outputs 1_{o1}and 1_{o2} fuzzifiers selected boundary values of the subdomain of the current parameters of the control object (U _{one} _{ min },U _{one} _{ max }), (U _{2} _{ min } _{,} U _{2} _{ max }), ..., (U _{ N } _{ min },U _{ Nmax }) are fed to inputs 2_{one} and 2_{2} UVUS located in the rows corresponding to the fuzzifier. From exit 1_{o2} jfuzzifier of valueU _{ j } _{ min } fed to input 5_{ j } block 3, setting the starting address of the selected subdomain of the current parameter valueU _{ aj } management object. Using start addressU _{one} _{ min },U _{2} _{ min }, ...,U _{N} _{ min } from the selected subdomain, memory unit 3 extracts the values of the control signal all discrete points in the cornersNdimensional cube of the selected subdomain bounded by the interval values of the parameters of the control object (U _{one} _{ min },U _{one} _{ max }), (U _{2} _{ min },U _{2} _{ max }), ..., (U _{ N } _{ min },U _{ N } _{ max })From memory block 3 with outputs 6 _{1} , ..., 6 _{S} , where S = 2 ^{ N } , the values of control signals are transmitted to the UVUS of the first row:
output 6 _{2} _{ i } _{1} and from exit 6 _{2} _{ i } , where,
i  the number of UVUS in the first row, j  the number of the parameter U (varies from 2 to N ). A signal is transmitted to the fourth input of the 2 _{4} i th UVS of the first row
a signal is transmitted to the fifth input of the 2 _{5} i th UVS of the first row At the output of ith UVUS in the first row is the intermediate value of the control signal , which is calculated by the formula (1). When j > 1, the fourth signal is transmitted to the fourth input of the 2 _{4} i th UVS of the j th row a signal is transmitted to the fifth input of the 2 _{5th} i th UVC of the jth row At the output of the i th UVUSS of the jth row there will be an intermediate value of the control signal , which is calculated by the formula (2). In the Nth row, the output of the UVUS 2 will be the value of the control signal F _{y} , which is transmitted to the processor output.Fuzzifiers and UVUS are implemented on analog circuits containing comparators, controlled keys, analog computing circuits, situational processors [Patent No. 2541850, IPC G06G7 / 25]. The memory block is implemented on digitaltoanalog circuits and microprocessors. Limitations on the number of parameters of the control object is determined only by the applied element base. The proposed method is effective with the monotonous nature of the control function in the intervals between adjacent experimental values in the N dimensional region of its definition.
The use of this invention will reduce the response time of existing automatic controllers used to control complex technical and technological objects with fastmoving processes, through parallel hardware processing of the rules for calculating the control function.
Claims (2)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

RU2017111665A RU2649791C1 (en)  20170406  20170406  Multiparameter fuzzy processor for automatic regulators and method for synthesis of control signal 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

RU2017111665A RU2649791C1 (en)  20170406  20170406  Multiparameter fuzzy processor for automatic regulators and method for synthesis of control signal 
Publications (1)
Publication Number  Publication Date 

RU2649791C1 true RU2649791C1 (en)  20180404 
Family
ID=61866931
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

RU2017111665A RU2649791C1 (en)  20170406  20170406  Multiparameter fuzzy processor for automatic regulators and method for synthesis of control signal 
Country Status (1)
Country  Link 

RU (1)  RU2649791C1 (en) 
Cited By (1)
Publication number  Priority date  Publication date  Assignee  Title 

RU2708008C1 (en) *  20190618  20191203  Федеральное государственное автономное образовательное учреждение высшего образования "Сибирский федеральный университет"  Plant for dehydration of sewage sludge by freezing 
Citations (4)
Publication number  Priority date  Publication date  Assignee  Title 

RU2212046C2 (en) *  20010301  20030910  Физический институт им. П.Н.Лебедева РАН  Optoelectronic processor 
US20100205137A1 (en) *  20090210  20100812  International Business Machines Corporation  Optimizing Power Consumption and Performance in a Hybrid Computer Evironment 
RU2417442C2 (en) *  20081219  20110427  Учреждение Российской академии наук Институт конструкторскотехнологической информатики РАН (ИКТИ РАН)  Method of constructing fuzzy logic systems and device for implementing said method 
RU2477525C2 (en) *  20110617  20130310  Алексей Евгеньевич Васильев  Microcontroller with hardware variable structure fuzzy computer 

2017
 20170406 RU RU2017111665A patent/RU2649791C1/en active
Patent Citations (4)
Publication number  Priority date  Publication date  Assignee  Title 

RU2212046C2 (en) *  20010301  20030910  Физический институт им. П.Н.Лебедева РАН  Optoelectronic processor 
RU2417442C2 (en) *  20081219  20110427  Учреждение Российской академии наук Институт конструкторскотехнологической информатики РАН (ИКТИ РАН)  Method of constructing fuzzy logic systems and device for implementing said method 
US20100205137A1 (en) *  20090210  20100812  International Business Machines Corporation  Optimizing Power Consumption and Performance in a Hybrid Computer Evironment 
RU2477525C2 (en) *  20110617  20130310  Алексей Евгеньевич Васильев  Microcontroller with hardware variable structure fuzzy computer 
Cited By (1)
Publication number  Priority date  Publication date  Assignee  Title 

RU2708008C1 (en) *  20190618  20191203  Федеральное государственное автономное образовательное учреждение высшего образования "Сибирский федеральный университет"  Plant for dehydration of sewage sludge by freezing 
Similar Documents
Publication  Publication Date  Title 

Wang et al.  Robust adaptive neural tracking control for a class of stochastic nonlinear interconnected systems  
CN103853046B (en)  A kind of Adaptivelearning control method of piezoelectric ceramic actuator  
Shoorehdeli et al.  Identification using ANFIS with intelligent hybrid stable learning algorithm approaches and stability analysis of training methods  
Azeem et al.  Generalization of adaptive neurofuzzy inference systems  
Jiménez et al.  Paretooptimal solutions in fuzzy multiobjective linear programming  
Meng et al.  Adaptive neural control of highorder uncertain nonaffine systems: A transformation to affine systems approach  
Reznik et al.  PID plus fuzzy controller structures as a design base for industrial applications  
Barbounis et al.  Locally recurrent neural networks for longterm wind speed and power prediction  
Jang et al.  Neurofuzzy modeling and control  
Huang et al.  Development and application of a novel radial basis function sliding mode controller  
Ahn et al.  Online tuning fuzzy PID controller using robust extended Kalman filter  
Saini et al.  Artificial neural network based peak load forecasting using Levenberg–Marquardt and quasiNewton methods  
Hsiao et al.  Robustness design of fuzzy control for nonlinear multiple timedelay largescale systems via neuralnetworkbased approach  
Mouzouris et al.  Dynamic nonsingleton fuzzy logic systems for nonlinear modeling  
Bobyr et al.  A method of defuzzification based on the approach of areas' ratio  
CN104462015B (en)  Process the fractional order linear discrete system state updating method of nongaussian L é vy noises  
Trabelsi et al.  Fuzzy identification of a greenhouse  
Kiumarsi et al.  Optimal control of nonlinear discrete timevarying systems using a new neural network approximation structure  
Abiyev  Fuzzy wavelet neural network based on fuzzy clustering and gradient techniques for time series prediction  
Shen  Fuzzy neural networks for tuning PID controller for plants with underdamped responses  
Luo et al.  A laguerre neural networkbased ADP learning scheme with its application to tracking control in the Internet of Things  
Mann et al.  Threedimensional min–maxgravity based fuzzy PID inference analysis and tuning  
JP6457563B2 (en)  Numerical control device and machine learning device  
Ebadat et al.  New fuzzy wavelet network for modeling and control: The modeling approach  
Wu et al.  Interventional consensus for highorder multiagent systems with unknown disturbances on coopetition networks 