CN115180461B - Tension data driving control method for new energy automobile compressor motor production equipment - Google Patents
Tension data driving control method for new energy automobile compressor motor production equipment Download PDFInfo
- Publication number
- CN115180461B CN115180461B CN202210831079.7A CN202210831079A CN115180461B CN 115180461 B CN115180461 B CN 115180461B CN 202210831079 A CN202210831079 A CN 202210831079A CN 115180461 B CN115180461 B CN 115180461B
- Authority
- CN
- China
- Prior art keywords
- winding machine
- swing angle
- tension
- output
- value
- 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
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65H—HANDLING THIN OR FILAMENTARY MATERIAL, e.g. SHEETS, WEBS, CABLES
- B65H63/00—Warning or safety devices, e.g. automatic fault detectors, stop-motions ; Quality control of the package
- B65H63/04—Warning or safety devices, e.g. automatic fault detectors, stop-motions ; Quality control of the package responsive to excessive tension or irregular operation of apparatus
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65H—HANDLING THIN OR FILAMENTARY MATERIAL, e.g. SHEETS, WEBS, CABLES
- B65H63/00—Warning or safety devices, e.g. automatic fault detectors, stop-motions ; Quality control of the package
- B65H63/02—Warning or safety devices, e.g. automatic fault detectors, stop-motions ; Quality control of the package responsive to reduction in material tension, failure of supply, or breakage, of material
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02K—DYNAMO-ELECTRIC MACHINES
- H02K15/00—Methods or apparatus specially adapted for manufacturing, assembling, maintaining or repairing of dynamo-electric machines
- H02K15/04—Methods or apparatus specially adapted for manufacturing, assembling, maintaining or repairing of dynamo-electric machines of windings, prior to mounting into machines
- H02K15/0435—Wound windings
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65H—HANDLING THIN OR FILAMENTARY MATERIAL, e.g. SHEETS, WEBS, CABLES
- B65H2701/00—Handled material; Storage means
- B65H2701/30—Handled filamentary material
- B65H2701/34—Handled filamentary material electric cords or electric power cables
Abstract
The invention discloses a tension data driving control method of new energy automobile compressor motor production equipment, which is mainly used for solving the problems of loose winding and uneven distribution of enameled wires, easy breaking of enameled wires in the winding process and the like caused by difficult modeling and low tension control precision in the tension control process of the new energy automobile compressor motor production equipment, and mainly comprises the following steps: collecting a swing angle signal; establishing a discretization equation of the swing angle relation dynamics of the winding machine; designing an improved model-free self-adaptive control method; designing a fuzzy BP neural network to obtain a pseudo-partial derivative value and an iterative weight correction formula; an improved model-free self-adaptive control method for the tension tight format of a winding machine based on a fuzzy BP neural network is designed. The model-free self-adaptive control method solves the problem of difficult modeling of the traditional winding machine, and the result shows that the dynamic performance and the steady-state performance of the winding machine control system are effectively improved and the tension control precision is improved through the comparison verification with the existing method.
Description
Technical Field
The invention relates to the technical field of tension control of new energy automobile compressor production equipment, in particular to a tension data driving control method of new energy automobile compressor production equipment.
Background
In recent years, new energy automobiles in China are continuously developed, the market competitiveness of the new energy automobiles is obviously improved, and the relation technologies of power batteries, driving motors, vehicle operating systems and the like are greatly broken through. According to the national and development trend of new energy automobile industry, the production of the new energy automobile compressor motor has wide market prospect, the tension control of a winding machine system is a main winding method of a compressor motor winding, cheng Mingyang and the like propose a full-automatic winding machine iterative sliding mode manufacturing method, zhang and the like propose a winding machine speed adjusting control method based on a rectangular wire mode, haouari fonad and the like propose a novel fault-tolerant winding machine control method based on a RST-Backstepping controller, and the development of the winding machine control method is promoted.
In practical studies, modeling of the tension control system is complicated, and thus the design of the controller is more difficult. Zheng Bao provides a design method of a tension compensation system based on model prediction equally, and the model prediction control method is applied to tension control to reduce a tension peak by more than 56%; the fuzzy self-adaptive tension control is not proposed, and the fuzzy control is applied to the tension control, so that the tension control precision is improved; mi Junjie et al propose an integral robust control method based on neural network disturbance compensation for tension control, and the ability of the tension control system to inhibit external disturbance is obviously enhanced. However, the control methods are all model-based control, the control precision and the self-adaptation degree are low, and as the production equipment of the new energy automobile compressor motor is continuously changed along with the increase of the winding quantity, the method can cause the problem that the tension control accuracy is reduced due to the inaccuracy of model establishment.
Therefore, in order to solve the problem of difficult modeling in the tension control system, cao Xiaohua and the like propose constant tension control research based on BP neural network, so that overshoot of the control system is reduced, and the tension control effect is improved; hu Yawei et al propose a tension control system based on BP neural network, which overcomes the defects of the traditional PID control and ensures the stable control of the tension by the tension control system. However, when the BP neural network is used as a controller, the essence of the BP neural network is a gradient descent method, the essence of the gradient descent method is that the objective function value is reduced in each iteration, and as long as the iteration times are large, the minimum value of the function value can be obtained necessarily, so that when the BP neural network faces the control function of a complex system, a 'zigzag phenomenon' can be generated necessarily, and the response speed of the control system is reduced. In order to solve the problem of the decline of tension control precision caused by inaccurate model establishment, a model-free adaptive control (MFAC) theory is proposed herein, which is that Hou Zhongsheng teacher in 1994 formally proposes the theory for the first time in its doctor paper, and a data driving control method for single input and single output is proposed, and by introducing a pseudo partial derivative phi (k), a nonlinear system of new energy automobile compressor production equipment can be converted into a simple 'generalized model', so that the tight format dynamic linearization of the model is realized by introducing the pseudo partial derivative phi (k), and the model-free adaptive control method only depends on the I/O data of the controlled system, and overcomes the difficulty that the mathematical model is required by the traditional controller. Because the pseudo partial derivative phi (k) is a time-varying parameter, the description is difficult to be carried out by using mathematical language, the characteristic that the pseudo partial derivative phi (k) can approach any function according to the BP neural network becomes the most widely used parameter evaluation method at present, and the pseudo partial derivative phi (k) has been successfully applied to aspects such as PID, intelligent PID, ADRC and the like. In practical application, the swing angle difference value of the winding machine at any moment is also influenced by external related factors, such as winding diameter change, inertia of a paying-off wheel motor and the like, so that the value of the pseudo partial derivative phi (k) and the actual value are larger in and larger out, the fuzzy BP neural network is utilized to approximately calculate the value of the pseudo partial derivative phi (k), the problem that the system response speed is reduced due to the 'zigzag phenomenon' when the BP neural network is used as a controller and faces a complex control function is avoided, the problem that the control method precision is reduced due to the time variability of the pseudo partial derivative phi (k) is avoided, and the tension control precision and the control system response speed of new energy automobile compressor motor production equipment are improved.
Disclosure of Invention
In view of the above, the invention discloses a tension control method of a new energy automobile compressor motor production device under data driving, which comprises the steps of firstly collecting a swing angle signal of a tension control system of a winding machine, and secondly establishing a dynamic discretization equation of the swing angle relation of the winding machine.
1. In order to achieve the above purpose, the present invention adopts the following technical scheme:
s1: swing angle signal acquisition
S2: establishing a discretization equation of the swing angle relation dynamics of the winding machine;
s3: designing an improved model-free self-adaptive control law;
s4: designing an improved winding machine tension tight format model-free self-adaptive control method (FNN-iCFMFAC) based on a fuzzy BP neural network;
further, in step S1, the swing angle signal acquisition includes:
firstly, an actuator outputs a paying-off linear speed signal to enable a swing rod spring mechanism to deform so as to output a swing rod swing angle value, a model-free self-adaptive control method is used as a controller to output a rotating speed control signal, and when a tension control system of a winding machine works, the controller receives a swing angle measured value theta output by a sensor o And the swing angle is adjusted to be a set value theta i And obtaining the swing angle theta of the tension system of the winding machine after the difference is made.
Further, in step S2, the establishing a discretization equation of the winding machine swing angle relation dynamics specifically includes:
secondly, setting an O point as a coordinate origin, taking the vertical upward direction of the O point as a y-axis positive direction, and taking the horizontal rightward point of the O point as an x-axis positive direction, so as to establish an xOy plane rectangular coordinate system; setting the mass of the swing rod as m, the moment of inertia taking O as the center as J, the length of the absolute value OE as R, the length of the absolute value OP as R and the swing angle as theta; assuming that the elastic modulus of the spring is K, the length of the spring is l, the coordinates of each point are O (0, 0), S (x) s ,y s )，D(x D ,y D )，F(x F ,y F )，E(-R sinθ,R cosθ)，P(-r sinθ,r cosθ)；
Let T be 1 And T 2 The same enamelled wire passes through the porcelain ring to pull the swing rod forwards and backwards, and the moment of the swing rod to the O point is M respectively 1 And M 2 ；
T 1 The moment for the point O is as follows:
T 2 the moment for the point O is as follows:
the moment of the gravity of the swing rod to the O point is as follows:
the tension of the spring to the swing rod is F, and the moment to the O point is as follows:
the swing rod winds around the point O under the moment, and a is acceleration. The moment balance equation is:
M 1 +M 2 +M mg +M F ＝Ja
to sum up, the tension T S The one-to-one functional relationship with the swing angle theta of the swing rod can be established as follows:
here we only consider the relationship between the swing angle and the tension when the swing lever is in steady state, i.e. a=0, and therefore by measuring the swing angle θThe tension T can be indirectly obtained by the variation s Is a variable amount of (a);
from the above analysis, the tension T of the winding machine can be found S The function relation with the swing angle theta can indirectly control the tension of the winding machine by controlling the swing angle, and according to the working principle of the winding machine, it can be known that if the tension of the winding machine is just not loosened or pulled off at the moment t, the swing angle theta of the winding machine and the winding speed V of the winding machine 1 And the paying-off speed V of the winding machine 2 The following relationship must exist:
after discretizing the formula:
θ(k+1)＝θ(k)+h(f(θ(k),V 1 (k),V 2 (k))
Δθ(k+1)＝Δθ(k)+h(f(θ(k),V 1 (k),V 2 (k)-f(θ(k),V 1 (k-1),V 2 (k-1))+f(θ(k),V 1 (k-1),V 2 (k-1))-f(θ(k-1),V 1 (k-1),V 2 (k-1)))
order theIn combination with the above
Let V 3 (k)＝V 1 (k)-V 2 (k)，f * ＝Δθ(k)+h(f(θ(k),V 1 (k),V 2 (k)-f(θ(k),V 1 (k-1),V 2 (k-1))
From the above sum Cauchy differential median theorem, the above formula can be written as follows:
when Δθ (k+1) +.0, there must be a time-varying parameter φεR called PPD, so that the system can be converted into the CFDL data model as follows
Δθ(k+1)＝φ(k)ΔV 3 (k)
In the method, in the process of the invention,the derivative value of the swing angle theta of the swing rod is represented, h represents the sampling period eta of a tension system of the winding machine * (t) represents the unique solution of the existence of the function, and f (…) represents the nonlinear function constituting the system;
taking the minimum swing angle error law of a tension control system of a winding machine as an input criterion function:
J(V 3 (k))＝|V 3 * (k+1)-V 3 (k+1)| 2 +α|θ(k)-θ(k-1)| 2
bringing the above into control laws of the available compact format model-free adaptive control method (CF-MFAC)
Wherein θ (k) represents the swing angle value at time k of the winding machine, θ (k-1) represents the swing angle value at time k-1 of the winding machine, φ (k) represents the pseudo partial derivative, ρ represents the step factor, λ represents the limit swing angle difference variation amount, V 3 * (k+1) represents a desired output rotational speed signal of the motor at time k+1, V 3 (k) The motor rotating speed at the moment k of the motor is represented;
further, in step S3, the design-improved model-free adaptive control law specifically includes:
the following specific steps of the improved tight format model-free adaptive control for improving the control law are as follows:
first, to illustrate that even the simplest linear system, the value of the pseudo-partial derivative phi (k) is unknown, time-varying, in order for the control law described above to be applicable in a particular environment, we will use the following criterion function to estimate the value of the pseudo-partial derivative online:
in the method, in the process of the invention,for the estimated value of the pseudo partial derivative phi, lambda is a weighting coefficient, solving the above formula for +.>Let it be zero, can get +.>Is used for the online estimation algorithm:
in the formula DeltaV 3 (k) The rotation speed increment of the motor of the winding machine at the moment k is given, lambda is a weighting coefficient, and delta theta (k-1) is a swing angle difference value at the moment k-1;
we introduce the following parameter reset mechanism:
where ε > 0 is a number infinitely close to 0,is->Is a value of the initial time of (a);
the original CF-MFAC control law is known to be essentially time-varying integral control, and only contains one time-varying integral control term, so that accurate control is difficult to perform on a complex system, and therefore, the control law is improved here, and the improved compact format model-free adaptive (iCF-MFAC) control law is designed as follows:
wherein k is p ≥0，k i The weight coefficient is more than 0, the improved compact format model-free control law consists of a time-varying proportional term and a time-varying integral control term, and the improved compact format model-free control law has better control effect for a nonlinear winding machine system;
in the above discussion we can see thatFor the approximation of the pseudo-partial derivative phi (k), we have also obtained +.>The estimation algorithm of (1) is as follows:
in the formula DeltaV 3 (k) In practice, the difference in the swing angle of the winding machine is affected by external related factors such as radius change of winding diameter, inertia of the motor of the paying-off wheel, etc., so that the difference in the swing angle of the winding machine isIn actual application, the estimation algorithm of (a) has larger access to the true value of the pseudo partial derivative phi (k) so as to cause the control accuracy of model-free adaptive control to be reduced, and the fuzzy is utilizedBP neural network to approximate to find pseudo partial derivative +.>For improving the model-free adaptive control method of the winding machine;
further, in step S4, the improved winding machine tension tight format model-free adaptive control method (FNN-ifemfac) based on the fuzzy BP neural network specifically includes:
first, assuming that the system state variable is e (k), the specific expression is as follows:
e(k)＝y * (k)-y(k)
secondly, normalizing the systematic error by calculating e/r, and dividing the systematic error into a plurality of grades in a closed interval [0,1] to finish fuzzy quantization;
wherein E is a fuzzy domain of systematic errors, and then multiplied by a reduction coefficient to adjust to the order of magnitude of 0-1, namely, the error E is converted into a concept value and sent to a neural network NN;
the neural network learning algorithm essentially aims at solving the minimum value of the controlled error function by using a 'steepest descent method'. The input layer nodes are M, the hidden layers are Q, and one output layer node, and the output value of the output layer node is the pseudo partial derivative in the model-free self-adaptive control parameter;
the output value of the output layer of the neural network, namely the pseudo-partial derivative value, is calculated from the forward output angle of the BP neural network, and the input node corresponds to the system state variable after fuzzy quantization processing:
O j (1) ＝E k-j ,j＝1,…M,k≥M
E k-j the value of e (k-j) is the fuzzy quantized value of the system error at the moment of k-j, the input variable M depends on the complexity of the controlled system, and the hidden layer output of the network is as follows:
in the method, in the process of the invention,zeta is the weight coefficient of the hidden layer i The superscripts (1), (2) and (3) represent an input layer, an hidden layer and an output layer; only one output layer of the neural network, so the output layer outputs as follows:
in the method, in the process of the invention,for the output layer weighting coefficients ζ is the threshold and g (x) is the Sigmoid function, i.e.>
Where β is a constant, and saturation is prevented in order to increase the range of output values;
the following criterion functions and inertial terms of search acceleration are introduced:
where η is learning efficiency and α is inertia coefficient but
Unknown, use->Replace->Wherein->Is V (V) 3 (k+1).
The paper quotes the forward prediction equation
Therefore, the weighting coefficient formula of the output layer is:
the weighting coefficient formula of the obtained hidden layer is as follows:
the error value of the pseudo partial derivative is further reduced by adjusting the network weight from the back propagation angle of the BP neural network, and the output error of the output layer neuron is set as follows:
e kj ＝d kj -y kj
wherein d kj To output the expected output of the jth neuron of the output layer, y kj For the actual output of the jth neuron of the output layer in the kth training, since the number of neurons of the output layer is 1 in the above formula, j=1 is taken here, and after H samples are input, the total node error sum of the output layer is:
the network weight correction algorithm is as follows:
in the formula, k is the training times,is the differential quotient of the error and the weight, delta is the learning step length
Here we refer to the gradient descent method with additional momentum terms, the weight correction of the algorithm is:
ω ij (k+1)＝ω ij (k)+Δω ij (k)
training output error E of cyclic utilization of two adjacent samples H (k) And E is H-1 (k) The iterative learning neural network weight correction formula is provided as follows:
wherein H is a sample set, lambda is learning gain
Based on the analysis, the improved winding machine control law of the winding machine tension tight-format model-free self-adaptive control method (FNN-iCFMFAC) based on the fuzzy BP neural network can be obtained as follows:
in the method, in the process of the invention,for the output layer weighting coefficients, g (x) is the Sigmoid function, i.e.>
Where β is a constant, and saturation is prevented in order to increase the range of output values.
Drawings
FIG. 1 is a flow chart of tension data driving control of new energy automobile compressor motor production equipment;
fig. 2 is a plane rectangular graph of the relationship between the swing angle and the tension of the winding machine;
FIG. 3 is a schematic diagram of a winding machine tension control method based on model-free adaptive control of a fuzzy BP neural network;
FIG. 4 is a graph showing the comparison of the swing angle of the winding machine under the control method and PID control method and model-free self-adaptive control method;
fig. 5 is a graph showing the comparison of the swing angles of the winding machine under unknown disturbance of the control method, the PID control method and the model-free self-adaptive control method.
Detailed Description
The invention will be described in more detail below with reference to the accompanying drawings:
referring to fig. 1,2 and 3, the embodiment of the present application specifically includes the following steps:
step S1: establishing a relation between a swinging angle and tension of a winding machine:
firstly, an O point is taken as a coordinate origin, a vertical upward direction of the O point is taken as a y-axis positive direction, a horizontal rightward point of the O point is taken as an x-axis positive direction, and an xOy plane rectangular coordinate system is established. Setting the mass of the swing rod as m, the moment of inertia taking O as the center as J, the length of the absolute value OE as R, the length of the absolute value OP as R and the swing angle as theta; let the elastic modulus of the spring be K and the length of the spring be l. Let the coordinates of each point be O (0, 0), S (x) s ,y s )，D(x D ,y D )，F(x F ,y F )，E(-R sinθ,R cosθ)，P(-r sinθ,r cosθ)；
Let T be 1 And T 2 The same enamelled wire passes through the porcelain ring to pull the swing rod forwards and backwards, and the moment of the swing rod to the O point is M respectively 1 And M 2 。
T 1 The moment for the point O is as follows:
T 2 the moment for the point O is as follows:
the moment of the gravity of the swing rod to the O point is as follows:
the tension of the spring to the swing rod is F, and the moment to the O point is as follows:
the swing rod winds around the point O under the moment, and a is acceleration. The moment balance equation is:
M 1 +M 2 +M mg +M F ＝Ja
to sum up, the tension T S The one-to-one functional relationship with the swing angle theta of the swing rod can be established as follows:
therefore, the tension T can be indirectly obtained by measuring the variation of the swing angle theta s Is a variable amount of (a).
Step S2: establishing a motor compact format dynamic linearization data model:
the compact format linearization method mainly aims at applying a linearization method designed by a control system, and has the characteristics of simple structure, convenience for controller design, convenience for direct utilization of input and output data and the like. The rotating speed system of the winding machine motor is a single-input single-output nonlinear discrete system and can be expressed as follows:
V 3 (k+1)＝f(V(k),···,V(k-n y ),θ(k),···,θ(k-n u ))
wherein V is 3 (k+1) represents the output of the system, i.e., motor rotation speed control signal, V 3 (k) The output quantity representing the previous moment of the system is the motor rotating speed control signal of the previous moment, theta (k) represents the input quantity representing the previous moment of the system is the swing angle difference, n y And n u Representing the order of the system; f (··) represents the nonlinear function that makes up the system.
Suppose 1: the input and output of the motor system of the winding machine are considerable and controllable, for all the given bounded output signals V 3 (k+1) there is always a bounded input signal to make the output of the system equal to the desired output V 3 (k+1)。
Suppose 2: in addition to the limited point of time, f (& gtis & lt, & gt) is related to (n) y The partial derivatives of +2) variables are continuous.
Suppose 3: except for the limited time points, the system meets the generalized Lipschitz condition, i.e. for any k 1 ≠k 2 ，k 2 0 and u (k) 1 )≠u(k 2 ) Has the following components
|V 3 (k 1 +1)-V 3 (k 2 +1)|≤b|θ(k 1 )-θ(k 2 )|
Wherein V is 3 (k i +1)＝f(V 3 (k i ),···V 3 (k i -n y ),θ(k i )，···θ(k i -n u ) I=1, 2; b > 0 is a constant.
Thus, the above assumption is satisfied, and when |Δu (k) | is not equal to 0, the following relationship must exist:
ΔV 3 (k+1)＝φ(k)Δθ(k)
phi (k) is bounded for any instant k, demonstrated as follows:
ΔV 3 (k+1)＝f(V 3 (k),···V 3 (k-n y ),θ(k),···,θ(k-n u ))-f(V 3 (k),···,V 3 (k-n y ),θ(k-1),θ(k-1),···,θ(k-n u ))+f(V 3 (k),···,V 3 (k-n y ),θ(k-1),θ(k-1),···,θ(k-n u ))-f(V 3 (k-1),···,V 3 (k-n y -1),θ(k-1),···,θ(k-n u -1))
ψ(k)＝f(V 3 (k),···,V 3 (k-n y ),θ(k-1),θ(k-1),···,θ(k-n u ))-f(V 3 (k-1),···,V 3 (k-n y -1),θ(k-1),···,θ(k-n u -1))
the median theorem can be derived from assumptions 2 and Cauchy derivatives as follows:
in the method, in the process of the invention,represents f (& gtis & lt- & gtis about (n) y Partial derivatives of +2) variables at
[V 3 (k),···,V 3 (k-n y ),θ(k-1),θ(k-1),···,θ(k-n u )] T And [ V 3 (k),···,V 3 (k-n y ),θ(k),θ(k-1),···θ(k-n u )] T A certain point value in between.
The following data equation is available:
ψ(k)＝η(k)Δu(k)
since the |Δu (k) | noteq 0 equation must have a unique solution η * (k)
Order theThe equation can be written as follows:
ΔV 3 (k+1)＝φ(k)Δθ(k)
the phi (k) is available in a bounded manner because the phi (k) is approximated by a fuzzy BP neural network, wherein the neural network output layer output value is the value of the pseudo partial derivative, and the specific steps are as follows:
the input nodes correspond to the system state variables after fuzzy quantization processing:
O j (1) ＝E k-j ,j＝1,…M,k≥M
E k-j the value of e (k-j) is the fuzzy quantized value of the system error at the moment of k-j, and the input variable M depends on the complexity of the controlled system. The hidden layer output of the network is:
in the method, in the process of the invention,zeta is the weight coefficient of the hidden layer i The superscripts (1), (2), and (3) represent input layers, hidden layers, and output layers for the threshold. Only one output layer of the neural network, so the output layer outputs as follows:
in the method, in the process of the invention,for the output layer weighting coefficients ζ is the threshold and g (x) is the Sigmoid function, i.e.>
Where β is a constant, and saturation is prevented in order to increase the range of output values.
The following criterion functions and inertial terms of search acceleration are introduced:
where η is learning efficiency and α is inertia coefficient but
Unknown, use->Replace->Wherein->Is V (V) 3 (k+1).
The paper quotes the forward prediction equation
Therefore, the weighting coefficient formula of the output layer is:
the weighting coefficient formula of the obtained hidden layer is as follows:
step S3: the improved winding machine tension tight format model-free self-adaptation (FNN-iCFMFAC) control law based on the fuzzy BP neural network is designed:
referring to fig. 4 and 5, a winding machine swing angle control system model is built through MATLAB/Simulink and simulation verification is carried out, a swing angle response curve of a winding machine in a starting process is shown in the figure, the adjustment time of FNN-iCFMFAC, MFAC and PID response curves is 0.23s, 0.3s and 0.36s respectively, the peak time is 0.07s, 0.09s and 0.08s respectively, the maximum swing angle difference of the response curves is 7 DEG, 12 DEG and 16.5 DEG, an unknown disturbance is added to the system in 1.25s, the swing angle can be stabilized in 1.4s by FNN-iCFMFAC, and the control effect is obviously improved.
Therefore, the data driving control of the new energy automobile compressor motor production equipment provided by the invention verifies the effectiveness of an improved winding machine tension tight format model-free self-adaptive control (FNN-iCFMFAC) method based on the fuzzy BP neural network through theoretical analysis and simulation.
Claims (4)
1. The tension data driving control method of the new energy automobile compressor motor production equipment is characterized by comprising the following steps of: the swing angle difference of the swing rod of the winding machine is used as input, and the tension control is realized by dynamically adjusting a wire wheel control motor of the winding machine by a model-free self-adaptive control method, which comprises the following steps:
s1: collecting a swing angle signal;
s2: establishing a discretization equation of the swing angle relation dynamics of the winding machine;
s3: designing an improved model-free self-adaptive control law, wherein a pseudo partial derivative evaluation method is provided;
s4: the improved winding machine tension tight format model-free self-adaptive control method based on the fuzzy BP neural network is designed:
only one output layer of the neural network, so the output layer outputs as follows:
in the method, in the process of the invention,for the weighting coefficient of the output layer, xi is a threshold value, g (x) is a Sigmoid function, and superscripts (2) and (3) represent hidden layers and output layers, < ->
Where β is a constant, and saturation is prevented in order to increase the range of output values;
based on the pseudo partial derivative evaluation method provided in the step S3, according to the characteristics of BP neural network reverse learning and high requirements on the tension control precision of the new energy automobile compressor motor, the problem of inaccurate pseudo partial derivative caused by external disturbance is provided by utilizing iterative weight correction, and the specific process is as follows:
e kj ＝d kj -y kj (2)
wherein d kj To output the expected output of the jth neuron of the output layer, y kj For the actual output of the jth neuron of the output layer in the kth training, since the number of neurons of the output layer is 1 in the above formula, j=1 is taken here, and after H samples are input, the total node error sum of the output layer is:
the network weight correction algorithm is as follows:
in the formula, k is the training times,is the differential quotient of the error and the weight, delta is the learning step length
Here we refer to the gradient descent method with additional momentum terms, the weight correction of the algorithm is:
ω ij (k+1)＝ω ij (k)+Δω ij (k) (5)
training output error E of cyclic utilization of two adjacent samples H (k) And E is H-1 (k) The iterative learning neural network weight correction formula is provided as follows:
wherein H is a sample set, lambda is learning gain
The control law of the winding machine based on the improved winding machine tension tight format model-free self-adaptive control method based on the fuzzy BP neural network is as follows:
。
2. the method for controlling the driving of tension data of the production equipment of the compressor motor of the new energy automobile according to claim 1, wherein in the step S1, the swing angle signal is acquired:
the system outputs a paying-off linear speed signal through an actuator to enable a swing rod spring mechanism to deform so as to output a swing rod swing angle value, a model-free self-adaptive control method is designed to serve as a controller to output a rotating speed control signal, and when a tension control system of a winding machine works, the controller receives a swing angle measured value theta output by a sensor o It is subjected toAnd the swing angle set value theta i And obtaining the swing angle theta of the tension system of the winding machine after the difference is made.
3. The method for controlling the driving of tension data of a new energy automobile compressor motor production device according to claim 1, wherein in step S2, the discrete equation of the winding machine swing angle relation dynamics is established:
secondly, setting an O point as a coordinate origin, taking the vertical upward direction of the O point as the positive y-axis direction, and taking the horizontal rightward point of the O point as the positive x-axis direction, so as to establish an xOy plane rectangular coordinate system; setting the mass of the swing rod as m, the moment of inertia taking O as the center as J, the length of the absolute value OE as R, the length of the absolute value OP as R and the swing angle as theta; assuming that the elastic modulus of the spring is K, the length of the spring is l, the coordinates of each point are O (0, 0), S (x) s ,y s )，D(x D ,y D )，F(x F ,y F ) E (-Rsin θ, rcos θ), P (-Rsin θ, rcos θ); tension T of winding machine S The relation with the swing angle theta is as follows:
if the tension of the winding machine is not relaxed or broken at time t, the swing angle θ of the winding machine and the winding speed V of the winding machine 1 And the paying-off speed V of the winding machine 2 The following relationship must exist:
after discretizing the above formulas (11) and (12):
θ(k+1)＝θ(k)+h(f(θ(k),V 1 (k),V 2 (k)) (13)
order theCombining the above can be achieved:
order the
V 3 (k)＝V 1 (k)-V 2 (k)，f * ＝Δθ(k)+h(f(θ(k),V 1 (k),V 2 (k)-f(θ(k),V 1 (k-1),V 2 (k-1)) (16)
From the median theorem of the differential values of equation (16) and Cauchy, equation (14) above can be written as follows:
when Δθ (k+1) +.0, there must be a time-varying parameter φεR called PPD, so that the system can be converted into the CFDL data model as follows:
Δθ(k+1)＝φ(k)ΔV 3 (k) (18)
in the method, in the process of the invention,the derivative value of the swing angle theta of the swing rod is represented, h represents the sampling period eta of a tension system of the winding machine * (t) represents the unique solution existing for the function, f (& gtis & lt- & gt) represents the composition of the composition a nonlinear function of the system;
taking the minimum swing angle error law of a tension control system of a winding machine as an input criterion function:
J(V 3 (k))＝|V 3 * (k+1)-V 3 (k+1)| 2 +α|θ(k)-θ(k-1)| 2 (20)
bringing the above (20) into the control law of a tightly-available format model-free adaptive control method
Wherein θ (k) represents the swing angle value at time k of the winding machine, θ (k+1) represents the swing angle value at time k+1 of the winding machine, φ (k) represents the pseudo partial derivative, ρ represents the step factor, λ represents the limit swing angle difference variation amount, V 3 * (k+1) represents a desired output rotational speed signal of the motor at time k+1, V 3 (k) The motor speed at the moment k of the motor is indicated.
4. The method for controlling the driving of tension data of the motor production equipment of the compressor of the new energy automobile according to claim 1, wherein in the step S3, the improved model-free adaptive control law is designed, and a pseudo partial derivative evaluation method is proposed:
the following specific steps of the improved compact format model-free self-adaptive control method obtained by improving the control law formula (21) are as follows:
estimating a pseudo partial derivative value using the following performance indicators:
in the method, in the process of the invention,for the estimated value of the pseudo partial derivative phi, lambda is a weighting coefficient, solving the above formula for +.>Is made to be zero, and can be obtainedIs used for the online estimation algorithm:
in the formula DeltaV 3 (k) The rotation speed increment of the motor of the winding machine at the moment k is given, lambda is a weighting coefficient, and delta theta (k-1) is a swing angle difference value at the moment k-1;
we introduce the following parameter reset mechanism:
where ε > 0 is a number infinitely close to 0,is->Is a value of the initial time of (a);
designing a control law of an improved compact format model-free adaptive method of the following form:
wherein k is p ≥0，k i And the weight coefficient is more than 0, the improved compact format model-free control law consists of a time-varying proportional term and a time-varying integral control term, and the improved compact format model-free control law can have better control effect for a nonlinear winding machine system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210831079.7A CN115180461B (en) | 2022-07-14 | 2022-07-14 | Tension data driving control method for new energy automobile compressor motor production equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210831079.7A CN115180461B (en) | 2022-07-14 | 2022-07-14 | Tension data driving control method for new energy automobile compressor motor production equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115180461A CN115180461A (en) | 2022-10-14 |
| CN115180461B true CN115180461B (en) | 2023-09-01 |
Family
ID=83520189
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210831079.7A Active CN115180461B (en) | 2022-07-14 | 2022-07-14 | Tension data driving control method for new energy automobile compressor motor production equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115180461B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117476364B (en) * | 2023-12-22 | 2024-03-12 | 日特机械工程（苏州）有限公司 | Winding machine control method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108777556A (en) * | 2018-06-29 | 2018-11-09 | 江苏大学 | The model-free adaption robust decoupling control method of two motor speed regulation systems based on neural network contrary modeling |
| CN112925208A (en) * | 2021-02-04 | 2021-06-08 | 青岛科技大学 | Disturbance compensation method for data-driven electro-hydraulic servo system of well drilling machine |
| CN214298669U (en) * | 2020-12-14 | 2021-09-28 | 山东中际智能装备有限公司 | Servo tension control mechanism for winding machine |
| CN215755635U (en) * | 2021-07-23 | 2022-02-08 | 河南巨力钢丝绳制造有限公司 | Steel wire rope pay-off device with automatic adjustment function |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10234848B2 (en) * | 2017-05-24 | 2019-03-19 | Relativity Space, Inc. | Real-time adaptive control of additive manufacturing processes using machine learning |
-
2022
- 2022-07-14 CN CN202210831079.7A patent/CN115180461B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108777556A (en) * | 2018-06-29 | 2018-11-09 | 江苏大学 | The model-free adaption robust decoupling control method of two motor speed regulation systems based on neural network contrary modeling |
| CN214298669U (en) * | 2020-12-14 | 2021-09-28 | 山东中际智能装备有限公司 | Servo tension control mechanism for winding machine |
| CN112925208A (en) * | 2021-02-04 | 2021-06-08 | 青岛科技大学 | Disturbance compensation method for data-driven electro-hydraulic servo system of well drilling machine |
| CN215755635U (en) * | 2021-07-23 | 2022-02-08 | 河南巨力钢丝绳制造有限公司 | Steel wire rope pay-off device with automatic adjustment function |
Non-Patent Citations (1)
| Title |
|---|
| 紧格式无模型自适应控制在直线电机中的仿真;李萍;宁夏大学学报;第35卷(第4期);324-327 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115180461A (en) | 2022-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111982302B (en) | Temperature measurement method with noise filtering and environment temperature compensation | |
| CN112613253B (en) | Vehicle mass and road gradient combined self-adaptive estimation method considering environmental factors | |
| CN111931411B (en) | Duhem dynamic hysteresis modeling method for piezoelectric driving micro-positioning platform | |
| CN109033585B (en) | Design method of PID controller of uncertain network control system | |
| CN108777556B (en) | Model-free adaptive robust decoupling control method for two-motor speed regulation system | |
| CN108803647B (en) | Model-free data driving control method for spacecraft attitude control | |
| CN110374804B (en) | Variable pitch control method based on gradient compensation of depth certainty strategy | |
| CN115180461B (en) | Tension data driving control method for new energy automobile compressor motor production equipment | |
| CN111716360B (en) | Fuzzy logic-based flexible joint mechanical arm sampling control method and device | |
| CN113110048B (en) | Nonlinear system output feedback adaptive control system and method adopting HOSM observer | |
| CN112077839B (en) | Motion control method and device for mechanical arm | |
| CN111443718B (en) | High-speed train state feedback prediction control method and system based on prediction error method | |
| CN116382071B (en) | Pneumatic parameter intelligent identification method for deep learning network correction compensation | |
| CN110442026B (en) | Extended state observer based on error correction, anti-interference control system and design method | |
| CN116991068A (en) | Motor control method and system based on distributed preset time gradient descent method | |
| CN108988710B (en) | Networked H-infinity model reference direct current motor speed regulation method and system considering long time delay | |
| CN116373846A (en) | Back wheel steering vehicle stability control method based on BP neural network optimization | |
| CN109995278B (en) | Motor servo system self-adjustment control method considering input limitation | |
| CN113283090B (en) | Friction compensation method for electric servo system of airplane steering engine based on bee colony algorithm | |
| CN110824922A (en) | Smith estimation compensation method based on six-order B-spline wavelet neural network | |
| Sun et al. | Event-triggered intelligent critic control with input constraints applied to a nonlinear aeroelastic system | |
| CN110932609B (en) | Self-adaptive recursion control method and system for multi-motor servo system | |
| CN113977571B (en) | Flexible joint robot output torque control method | |
| Wang et al. | Adaptive fuzzy fixed-time control for a class of strict-feedback stochastic nonlinear systems | |
| CN113589862A (en) | Intelligent PID temperature control method and control chip based on network topology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |