CN102799107B - Stepper motor proportion integration differentiation (PID) parameter self-matching method based on micro-neural network - Google Patents
Stepper motor proportion integration differentiation (PID) parameter self-matching method based on micro-neural network Download PDFInfo
- Publication number
- CN102799107B CN102799107B CN201210307712.9A CN201210307712A CN102799107B CN 102799107 B CN102799107 B CN 102799107B CN 201210307712 A CN201210307712 A CN 201210307712A CN 102799107 B CN102799107 B CN 102799107B
- Authority
- CN
- China
- Prior art keywords
- stepper motor
- value
- neural network
- learning rate
- current
- 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
Landscapes
- Feedback Control In General (AREA)
Abstract
The invention discloses a stepper motor proportion integration differentiation (PID) parameter self-matching method based on a micro-neural network, and the method comprises following steps of framing the micro-neural network; setting an initial value of a weight coefficient and a learning rate of nerve cells on an input layer; acquiring a moment phase current value of a stepper motor, and acquiring moment target current of the stepper motor; calculating an error value; calculating a value of the nerve cells on the input layer; calculating a novel weight coefficient; and calculating a value of nerve cells on an output layer, and adopting the value of the nerve cells as the voltage of the stepper motor. The stepper motor PID parameter self-matching method has beneficial effects that the matching of the PID parameter can be automatically completed without the manual intervention, so that the workload for the debugging of the entire machine can be alleviated, links set by users can be reduced, and the error possibility also can be avoided; the method has wider adaptability to different types of stepper motors, and optimum parameters can be automatically matched through rapid learning for a non-standard-model stepper motor; and the performance of the stepper motor can be adequately played, the vibration of the motor can be reduced, and the integral system reliability can be improved.
Description
Technical field
The present invention relates to a kind of stepper motor pid parameter control method, particularly relate to a kind of stepper motor pid parameter Self Matching method based on miniature neural network.
Background technology
In some real world applications occasions, a usual driver may need the stepper motor of adaptive Multiple Type, and not only the electromechanics parameter of same model and electric parameter have difference, and the motor of different model to obtain parameter difference huge especially.The effect of driver plays the maximum performance of motor as much as possible, if driver control parameter is bad with stepper motor parameter matching, can affect stepper motor performance.The control method that present driver is conventional is, for the motor of different models in advance by parameter read-in program, in experimental stage by a large amount of tests, find out the mean value of the parameter of motor in a certain model, then the parameter that driver of adjusting out successively matches with it, and prior write-in program.Replace generally being worth control motor with representative value like this.
(1) prior art shortcoming is cannot accurate match step motor control parameter, causes and cannot give full play to stepper motor performance; (2) higher for motor speed, and change application scenario faster, prior art cannot good tracking velocity, thus causes the problems such as response lag, thus causes motor to shake, affect machine system reliability and precision.。
Summary of the invention
The object of the invention is to the shortcoming and defect overcoming above-mentioned prior art, providing a kind of stepper motor pid parameter Self Matching method based on miniature neural network, with solving: (1) prior art cannot the phenomenon of motor of the various model of Auto-matching; (2) when rotating speed is higher and frequent variations time the phenomenon of response lag and the phenomenon of response overshoot, thus the defect of influential system reliability and precision.
Object of the present invention is achieved through the following technical solutions: based on the stepper motor pid parameter Self Matching method of miniature neural network, comprise the following steps:
(1) the micro-neural network of framework, this micro-neural network comprises input layer and output layer, and described input layer comprises three neurons, and these three neurons are respectively ratio, differential and integration, and output layer comprises a neuron;
(2) initial value and the learning rate of the weight coefficient of input layer is set;
(3) obtain stepper motor certain phase current values current, be designated as Ia; Obtain the target current Iexa that stepper motor is current, namely target current Iexa is the rated current of stepper motor at current operating state;
(4) difference e=Iexa – Ia of current time size of current and target current, obtains the difference of current flow at K moment stepper motor and target current, i.e. error amount e (K):
e(K)=Iexa(K)–Ia(K) （1）
Wherein K is 0,1,2,3
(5) input layer value is calculated;
(6) new weight coefficient is calculated;
(7) calculate the neuron value of output layer, this value is as the voltage of stepper motor.
Further, namely above-mentioned learning rate is ratio learning rate
, differential learning rate
with integration learning rate
, ratio learning rate
, integration learning rate
with differential learning rate
span be 0 ~ 1.And ratio learning rate
optimum value be 0.01, differential learning rate
optimum value be 0.1 and integration learning rate
optimum value be 0.001.
The initial value of above-mentioned weight coefficient, namely proportional roles coefficient is
, differential weight coefficient is
, integration weight coefficient is
.
Above-mentioned input layer value through type (2) calculates:
（2）
In formula wherein
for the ratio input value in k moment,
for the differential input value in k moment,
for the integration input value in k moment.
Above-mentioned new weight coefficient through type (3) calculates:
（3）。
The neuron value through type (4) of above-mentioned output layer calculates:
（4）。
Micro-neural network, as a kind of neural network of simplification, has on-line study and adaptive ability, and not only structure is simple, and is easy to calculate, and has stronger robustness, after in conjunction with pid algorithm, possesses especially and responds the features such as quick.Therefore the invention has the beneficial effects as follows:
(1) automatically can complete the coupling of pid parameter without the need to manual intervention, alleviate the workload of machine debugging, decrease user and link is set, thus it also avoid the possibility of makeing mistakes;
(2) there is adaptability widely to all types of stepper motor, for the stepper motor of non-standard model, optimized parameter can be gone out by Self Matching by Fast Learning;
(3) no matter can give full play to the performance of stepper motor, be at low speed, or under high-speed case, have good performance, especially at high speeds, can reduce motor vibrations, improve machine system reliability compared to prior art.
Accompanying drawing explanation
Fig. 1 is control principle drawing of the present invention;
Fig. 2 is the iptimum speed change curve comparison diagram adopting motor of the present invention, adopt common pid algorithm motor and motor in theory;
Fig. 3 is the best torque frequency feature curve comparison diagram adopting motor of the present invention, adopt common pid algorithm motor and motor in theory.
Embodiment
Below in conjunction with embodiment, the present invention is described in further detail, but structure of the present invention is not limited only to following examples:
[embodiment]
As shown in Figure 1, based on the stepper motor pid parameter Self Matching method of miniature neural network, comprise the following steps:
(1) the micro-neural network of framework, this micro-neural network comprises input layer and output layer, and described input layer comprises three neurons, and these three neurons are respectively ratio, differential and integration, and output layer comprises a neuron;
(2) initial value and the learning rate of the weight coefficient of input layer is set;
(3) obtain stepper motor certain phase current values current, be designated as Ia; Obtain the target current Iexa that stepper motor is current, namely target current Iexa is the rated current of stepper motor at current operating state;
(4) difference e=Iexa – Ia of current time size of current and target current, obtains the difference of current flow at K moment stepper motor and target current, i.e. error amount e (K)=Iexa (K) – Ia (K), K is 0,1,2,3
(5) input layer value is calculated;
(6) new weight coefficient is calculated;
(7) calculate the neuron value of output layer, this value is as the voltage of stepper motor.
Further, namely above-mentioned learning rate is ratio learning rate
, differential learning rate
with integration learning rate
, ratio learning rate
, integration learning rate
with differential learning rate
span be 0 ~ 1.And ratio learning rate
optimum value be 0.01, differential learning rate
optimum value be 0.1 and integration learning rate
optimum value be 0.001.
The initial value of above-mentioned weight coefficient, namely proportional roles coefficient is
, differential weight coefficient is
, integration weight coefficient is
.
The calculating formula of above-mentioned input layer value is
（1）
In formula wherein
for the ratio input value in k moment,
for the differential input value in k moment,
for the integration input value in k moment.
The calculating formula of above-mentioned new weight coefficient is
（2）。
The neuron value calculating formula of above-mentioned output layer is
（3）。
The present invention relates to initial proportion coefficient, learning rate, weighting coefficient is isoparametric determines, these parameters have a great impact study and Control platform.Following rule is had about the setting of these parameters and adjustment:
Rule 1: learning rate is to the rapidity, the robustness that improve system, and very greatly, learning rate can not be excessive for elimination overshoot and static error impact, otherwise the easy overshoot of neuron regulator; Learning rate can not be too small, otherwise neuron regulator adjustment process is slow.Generally determine optimum coefficient by Multi simulation running or test.
Rule 2: after once learning speed is selected, the initial value of weight can change within the specific limits, and the performance of not influential system.According to the algorithm of micro-neural network, neuron makes weights change towards making the direction of system stability by study, if but the selection of initial weight has exceeded neuronic range of adjustment, and system cannot restrain.Simulation result shows, suitable learning rate can make initial weight choose in a big way, after Multi simulation running research, just obtains good value: ratio learning rate
optimum value be 0.01, differential learning rate
optimum value be 0.1 and integration learning rate
optimum value be 0.001; Proportional roles coefficient is
, differential weight coefficient is
, integration weight coefficient is
.
The present invention and prior art is compared below by experiment:
(1) first by using the driver of general pid algorithm to connect upper stepper motor, then by repetitious experiment, manually pid parameter being adjusted to optimum, then the speed change curves of its control step motor being recorded, on the other hand, upper stepper motor is connected by using the driver of micro-Neural network PID parameter Self Matching, adjust without artificial parameter, then also speed change curves is recorded, finally two groups of results are analyzed, experimental result is illustrated in fig. 2 shown below: in figure, solid-line curve is the rate profile that general pid algorithm obtains, thick dashed line is the rate profile that the present invention obtains, fine dotted line is theoretic iptimum speed curve, can obtain like this: general pid control algorithm, after being manually adjusted to optimized parameter, under the low speed and the little occasion of speed conversion has good performance, but for high speed, can still there is larger fluctuation in the occasion that rate variation is large, therefore vibrations can be there are in motor, the phenomenons such as heating, thus the precision of influential system and reliability.And micro-Neural network PID parameter Self Matching, when manually not adjusting parameter completely, almost consistent with theoretical curve performance can be reached, thus effectively can improve precision and the reliability of whole system.
(2) first will the driver of general pid algorithm be used to connect upper stepper motor, this driver have mated same model stepper motor in advance, and parameter is set to optimum.Then the torque frequency feature curve record of its control step motor is got off.On the other hand, upper stepper motor is connected by using the driver of miniature Neural network PID parameter Self Matching, without artificial adjustment, then its torque frequency feature curve is also recorded, finally two groups of results are analyzed, experimental result is illustrated in fig. 3 shown below: in figure, upper curve is best torque frequency feature curve in theory, intermediate curve is the torque frequency feature curve adopting the present invention to obtain, lower curve is the torque frequency feature curve adopting general pid algorithm to obtain, known, for general pid control algorithm, its parameter is manually adjusted to Optimum Matching same model stepper motor, then same model motor is replaced, its torque frequency feature curve comparatively theoretical curve has some to decline.And for micro-Neural network PID parameter Self Matching, when manually not adjusting parameter completely, replace same model stepper motor howsoever, all still can reach almost consistent with theoretical curve performance.
Experimental result shows, micro-Neural network PID parameter Self Matching compensate for the deficiency of the Controlling model of pure PID, can link up its Fast Learning ability, have adaptability widely, stepper motor performance can be given full play to without the need to manual intervention, and ensure precision and the reliability of machine system.
Claims (4)
1., based on the stepper motor pid parameter Self Matching method of miniature neural network, it is characterized in that, the method comprises the following steps:
(1) the micro-neural network of framework, this micro-neural network comprises input layer and output layer, and described input layer comprises three neurons, and these three neurons are respectively ratio, differential and integration, and output layer comprises a neuron;
(2) initial value and the learning rate of the weight coefficient of input layer is set;
(3) obtain stepper motor certain phase current values current, be designated as Ia; Obtain the target current Iexa that stepper motor is current, namely target current Iexa is the rated current of stepper motor at current operating state;
(4) difference e=Iexa – Ia of current time size of current and target current, obtains the difference of current flow at k moment stepper motor and target current, i.e. error amount e (k):
e(k)=Iexa(k)–Ia(k) （1）
Wherein k is 0,1,2,3
(5) input layer value is calculated;
(6) new weight coefficient is calculated;
(7) calculate the neuron value of output layer, this value is as the voltage of stepper motor;
Namely described learning rate is ratio learning rate, differential learning rate and integration learning rate, described ratio learning rate
value be 0.01, differential learning rate
value be 0.1 and integration learning rate
value be 0.001; The initial value of described weight coefficient, namely proportional roles coefficient is
, differential weight coefficient is
, integration weight coefficient is
.
2. the stepper motor pid parameter Self Matching method based on miniature neural network according to claim 1, it is characterized in that, the input layer value through type (2) described in step (5) calculates:
（2）
In formula wherein
for the ratio input value in k moment,
for the differential input value in k moment,
for the integration input value in k moment.
3. the stepper motor pid parameter Self Matching method based on miniature neural network according to claim 2, is characterized in that, described new weight coefficient through type (3) calculates:
（3）。
4. the stepper motor pid parameter Self Matching method based on miniature neural network according to claim 3, it is characterized in that, the neuron value through type (4) of described output layer calculates:
（4）。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210307712.9A CN102799107B (en) | 2012-08-28 | 2012-08-28 | Stepper motor proportion integration differentiation (PID) parameter self-matching method based on micro-neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210307712.9A CN102799107B (en) | 2012-08-28 | 2012-08-28 | Stepper motor proportion integration differentiation (PID) parameter self-matching method based on micro-neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102799107A CN102799107A (en) | 2012-11-28 |
| CN102799107B true CN102799107B (en) | 2015-03-18 |
Family
ID=47198237
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210307712.9A Active CN102799107B (en) | 2012-08-28 | 2012-08-28 | Stepper motor proportion integration differentiation (PID) parameter self-matching method based on micro-neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102799107B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103439882B (en) * | 2013-09-02 | 2016-08-17 | 北京经纬恒润科技有限公司 | The method of adjustment of a kind of controller parameter and device |
| CN103853046B (en) * | 2014-02-14 | 2017-10-10 | 广东工业大学 | A kind of Adaptive-learning control method of piezoelectric ceramic actuator |
| CN104076702A (en) * | 2014-06-26 | 2014-10-01 | 天津市松正电动汽车技术股份有限公司 | Motor parameter self-matching method |
| CN106682735B (en) * | 2017-01-06 | 2019-01-18 | 杭州创族科技有限公司 | The BP neural network algorithm adjusted based on PID |
| CN110529419A (en) * | 2019-09-02 | 2019-12-03 | 苏州贝舒医疗科技有限公司 | The pressure output control method of noninvasive ventilator blower |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6078843A (en) * | 1997-01-24 | 2000-06-20 | Honeywell Inc. | Neural network including input normalization for use in a closed loop control system |
| CN101571705A (en) * | 2008-04-29 | 2009-11-04 | 北京航空航天大学 | Position servo system and method |
-
2012
- 2012-08-28 CN CN201210307712.9A patent/CN102799107B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6078843A (en) * | 1997-01-24 | 2000-06-20 | Honeywell Inc. | Neural network including input normalization for use in a closed loop control system |
| CN101571705A (en) * | 2008-04-29 | 2009-11-04 | 北京航空航天大学 | Position servo system and method |
Non-Patent Citations (5)
| Title |
|---|
| 《单神经元自适应PID控制算法》;郝志红;《冶金自动化》;20120501;第228-229页 * |
| 《基于单神经元自适应控制的多电机同步控制系统的研究》;姚燕春;《江苏大学硕士学位论文》;20070331;第6页 * |
| 付华,冯爱伟,徐耀松,王传英,孟宪敬.《基于神经元控制器的异步电动机矢量控制》.《中国电机工程学报》.2006,第26卷(第1期),第129页. * |
| 吴凌云,陈春霞.《基于单神经元PID控制器的PLC直流电机控制系统》.《制造业自动化》.2010,第32卷(第9期),第102页. * |
| 陶永华,尹怡欣,葛芦生.《自适应PID控制》.《新型PID控制及其应用》.机械工业出版社,1988,第38-43页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102799107A (en) | 2012-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102799107B (en) | Stepper motor proportion integration differentiation (PID) parameter self-matching method based on micro-neural network | |
| CN108489015B (en) | Air conditioning system temperature control method based on pole allocation and Pade approximation | |
| CN104300863A (en) | Self-adaption sliding mode control method for speed regulation of variable-load permanent magnet synchronous motor | |
| CN105867136A (en) | Parameter identification based multi-motor servo system synchronization and tracking control method | |
| CN103410660A (en) | Wind power generation variable pitch self-learning control method based on support vector machine | |
| CN108757192B (en) | Diesel engine electronic control speed regulation and test method based on fuzzy variable structure | |
| CN110552961A (en) | Active magnetic bearing control method based on fractional order model | |
| CN105888971B (en) | A kind of large scale wind power machine blade active load shedding control system and method | |
| CN102509152A (en) | Switched reluctance motor on-line modeling method based RBF neural network | |
| CN107863910B (en) | Permanent magnet synchronous motor optimal fractional order PID control method with strong tracking performance | |
| CN103368474A (en) | Motor rotation speed control method | |
| WO2021237910A1 (en) | Active magnetic bearing controller construction method based on table lookup method | |
| CN103246201A (en) | Improved fuzzy model-free adaptive control system and method for radial mixing magnetic bearing | |
| CN110361974B (en) | Water turbine speed regulating system optimization method based on BP-FOA hybrid algorithm | |
| CN108512476B (en) | Induction motor rotating speed estimation method based on Longbeige observer | |
| CN110515348B (en) | Servo motor model selection method of machine tool | |
| CN102393645A (en) | Control method of high-speed electro-hydraulic proportional governing system | |
| CN104199289A (en) | Identification-free single neuron self-adaption PID (Proportion Integration Differentiation) control method for magnet power source rectification system | |
| CN113014167A (en) | Permanent magnet motor nonsingular terminal sliding mode control method based on disturbance observer | |
| CN111219293B (en) | Variable pitch controller design method based on linear active disturbance rejection control | |
| CN108566137A (en) | A kind of discrete time-domain parameterization design method of electric machine position servo controller | |
| CN108919642B (en) | Optimal setting method for controller parameters of furnace-following machine coordination control system | |
| CN108267970B (en) | Time-lag rotor active balance control system and method based on Smith model and single neuron PID | |
| CN112761796B (en) | Power closed-loop control system and method thereof | |
| CN107359835A (en) | A kind of ultrahigh speed permagnetic synchronous motor method for controlling number of revolution based on adaptive robust control |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |