Testing machine for simulating load of servo motor of injection molding machine and self-adaptive control method

Abstract

The invention belongs to the technical field of injection molding equipment, and discloses a testing machine for simulating the load of a servo motor of an injection molding machine and a self-adaptive control method.The testing machine comprises a motor input unit, a gear reduction unit, a load unit and a monitoring unit; and dynamic load characteristics in an injection molding process are simulated. According to the self-adaptive control method, a BP-PSO-PID algorithm is adopted, a load signal is collected in real time and compared with a target value to generate an error, the error is input into a BP neural network to output PID parameters, the weight of the neural network is optimized online through a particle swarm optimization algorithm, and dynamic adjustment is achieved. According to the method, load changes in the glue injection, pressure maintaining and glue melting stages are simulated through a three-section type load curve, closed-loop control is formed in combination with the quick response characteristic of the eddy current brake, and the test precision and the self-adaptive capacity are remarkably improved. The dynamic load is accurately reproduced, the research and development cost is reduced, and the research and development period is shortened.

Description

Technical Field

[0001] The invention belongs to the technical field of injection molding equipment, and in particular relates to a testing machine for simulating the load of a servo motor of an injection molding machine and an adaptive control method. Background Art

[0002] Traditional injection molding machines face complex dynamic load variations in actual production. Testing directly on production machines presents challenges such as high cost, high risk, and difficulty adjusting parameters. Simulated injection molding machine load testing is a crucial method for evaluating and testing the performance and load characteristics of injection molding machines. Traditional load testing methods typically rely on external devices such as mechanical devices or dampers to provide the load. These methods are unable to accurately replicate the complex dynamic load characteristics of the injection molding process, struggle to meet the testing accuracy required for servo system development, and can be affected by changes in the external environment, making it difficult to adapt to varying operating conditions. These limitations severely restrict the development efficiency and testing reliability of injection molding machine servo drive systems.

[0003] In existing injection molding machine servo motor load testing technologies, the traditional PID control algorithm, due to its fixed parameters, struggles to adapt to the complex nonlinear load variations encountered in the injection molding process, resulting in insufficient test accuracy. While BP neural networks possess nonlinear mapping capabilities, their inherent shortcomings, such as slow convergence and a tendency to fall into local optimality, limit their application in real-time control systems. While the particle swarm optimization (PSO) algorithm alone can optimize parameters, it lacks the ability to continuously adjust the parameters online. More importantly, existing test systems generally fail to effectively integrate the advantages of these three methods, resulting in an inability to meet the requirements for high-precision dynamic load simulation, rapid response, and multi-condition adaptive simulation required in the development of injection molding machine servo systems. Summary of the Invention

[0004] The present invention aims to provide a test machine and an adaptive control method for simulating the load of a servo motor of an injection molding machine, so as to solve the above-mentioned technical problems.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention, a test machine for simulating the load of a servo motor of an injection molding machine and an adaptive control method are as follows:

[0006] A test machine for simulating the load of a servo motor of an injection molding machine, comprising a motor input unit, a gear reduction unit, a load unit and a monitoring unit, wherein the motor input unit comprises a servo motor, a first coupling and a motor bracket, wherein the motor bracket is fixed on a platform plane, the servo motor is fixed on the motor bracket, the servo motor output shaft is connected to the first coupling, and the first coupling is connected to the monitoring unit; the monitoring unit comprises a torque sensor, a second coupling and a torque sensor support, wherein the torque sensor support is fixed on the platform plane, the torque sensor is fixed on the torque sensor support, one end of the torque sensor is connected to the first coupling via an internal shaft, and the other end is connected to the second coupling via an internal shaft, and the second coupling is connected to Connected to the gear reduction unit; the gear reduction unit includes a transmission shaft, a first bearing support, a second bearing support, a gear set, and a third coupling. The gear set includes a large gear and a pinion. The first bearing support and the second bearing support are fixed on the platform plane and are respectively used to support the axles of the pinion and the large gear. One end of the transmission shaft is connected to the second coupling, and the other end is connected to the axle of the pinion. The pinion is meshed with the large gear, and the axle of the large gear is connected to the third coupling. The third coupling is connected to the load unit; the load unit includes an eddy current brake and a brake support. The brake support is fixed on the platform plane, and the eddy current brake is fixed on the brake support; the eddy current brake is connected to the third coupling.

[0007] Furthermore, the gear set is meshed in a manner that the two axes are horizontal, and all the matching modes of the shaft bodies are key and keyway matching modes.

[0008] Furthermore, pads are provided at the bottom of the torque sensor and the bottom of the eddy current brake.

[0009] Furthermore, a PLC control unit is used to control the eddy current brake. The PLC control unit includes an adaptive algorithm, which monitors the injection molding machine load parameters in real time and makes adaptive adjustments according to the load characteristics. Based on the real-time monitored load parameters, the PLC control unit adjusts the current and phase of the eddy current brake through the BP-PSO-PID control algorithm, and achieves precise control of the simulated injection molding machine load by continuously optimizing the control strategy.

[0010] The present invention also discloses an adaptive control method for a testing machine for simulating the load of a servo motor of an injection molding machine, comprising the following steps:

[0011] Step 1: After the system is started, the hardware of the test machine is initialized and the target load curve is set. The servo motor is set to torque control mode, the eddy current brake pre-excitation current is set, the torque sensor is zeroed, and the initial offset is eliminated.

[0012] Simultaneously load the preset load curve r(t);

[0013] Step 2: Use a high-precision torque sensor to collect the actual load signal y(t) and compare it with the target value r(t) to obtain the error

[0014] e(t):

[0015] e(t)=r(t)-y(t)

[0016] The error e(t), set value r(t) and output value y(t) are input into the three-layer BP neural network, and the number of PSO parameters and inertia weight are set. The neural network output is optimized online by the particle swarm algorithm to generate the optimal PID parameter K p ,K i ,K d ;

[0017] Step 3: Parameters are input into a digital PID controller to generate a control voltage. After power amplification, the eddy current brake is driven to generate precise braking torque, which is transmitted to the servo motor through a gear set to form a closed-loop control. During this process, the PSO algorithm continuously optimizes the neural network weights, enabling the system to adaptively adjust parameters to adapt to dynamic load changes at each stage of the injection molding process. Furthermore, step 1 includes the following steps:

[0018] Configure the servo motor working mode to torque control mode, and its dynamic equation is as follows:

[0019]

[0020] Among them, T m is the motor output torque (unit: Nm); I q is the torque current of the servo motor; K t is the torque constant, set at 1.2Nm / A; J m is the motor rotor inertia, ω is the real-time speed, and B is the damping coefficient;

[0021] The initial excitation current I0 applied to the eddy current brake is 0.4Nm / A, and its current-torque relationship is:

[0022] T l =K t (T)·I 2

[0023] Where: T l is the load torque, T is the brake temperature, and is executed through the PLC analog output module; K t (T) represents the temperature-related eddy current brake torque coefficient, I is the brake excitation current, and the torque sensor is automatically zeroed under no-load conditions;

[0024] Generate a load curve and define a three-segment load curve r(t) to simulate the injection molding process:

[0025] (1) Injection stage: 0≤t<0.1s, load increases linearly to 150Nm, the slope reflects the injection speed:

[0026]

[0027] (2) Pressure holding stage: 0.1≤t<0.5s, superimposed Gaussian noise to simulate hydraulic fluctuations:

[0028]

[0029] (3) Melting stage: t≥0.5s, sinusoidal fluctuation simulates screw retraction:

[0030] r(t)=80+30sin(4π(t-0.5)).

[0031] Furthermore, the step 2 includes the following steps:

[0032] The control error e(t) is:

[0033] e(t)=r(t)-y(t)

[0034] Take e(t), r(t), and y(t) as the 3-node input vector of the BP neural network, set the hidden layer to 7 nodes, and obtain the output PID parameters through the forward propagation formula:

[0035] h(t)=ReLU(W ih I(t)+b h )

[0036] [K p ,K i ,K d ] T =W ho h(t)+b o

[0037] Among them: K p ,K i ,K d is the PID output parameter, I(t) represents the normalized control signal input vector [r(t); y(t); e(t)], h(t) is the output vector of the hidden layer, W is the weight matrix from the input layer to the hidden layer, W ho is the weight matrix from the hidden layer to the output layer, b h is the bias vector of the hidden layer, b o is the bias vector of the output layer, ReLU(·) represents the rectified linear unit activation function;

[0038] The PSO weight optimization is performed on the output parameters. The specific optimization process of PSO is as follows:

[0039] First, initialize the PSO-BP parameters, the number of particles N is 50, and each particle encodes all the weights of the BP network:

[0040]

[0041] Where: X i Represents the position vector of the i-th particle, encoding all adjustable parameters of the BP neural network; W ih 、W ho 、b h 、b o Same meaning as above;

[0042] The fitness function is:

[0043]

[0044] Among them: e(t) is the control error, u(t) is the PID output control quantity;

[0045] Then update the particle position and velocity:

[0046]

[0047]

[0048] in, represents the velocity vector of the i-th particle at the k-th iteration, represents the position vector of the i-th particle at the k-th iteration, pbest i is the individual best historical position of the i-th particle, gbest represents the global best historical position (the best solution among all particles), w(k) is the inertia weight at the k-th iteration, which decreases from 0.9 to 0.4, k is the current iteration number, k max is the maximum number of iterations. c1, c2 are learning factors, r1, r2 are random numbers between 0 and 1;

[0049] Finally, the global optimal weight gbest is output, and the optimal weight W is obtained by decoding gbest. ih * (:),W ho * (:),b h * ,b o * , input the real-time input I(t) = [r(t); y(t); e(t)] into the optimized BP network,

[0050] h(t)=ReLU(W ih * I(t)+bh * )

[0051] [K p ,K i ,K d ] T =W ho * h(t)+b o *

[0052] After parameter denormalization, the optimal PID parameter K optimized by PSO is generated. p * ,K i * ,K d * .

[0053] Furthermore, the step three includes the following steps:

[0054] The PLC controller uses a position PID algorithm, and the optimized PID parameter K p * ,K i * ,K d * , update the control voltage u(t) every 1ms:

[0055]

[0056] Where: T s is the control period, which is 1ms; e(k) is the error at the current moment; ∑e(j) is the error integral term; is the error differential term;

[0057] The power amplifier converts the voltage u(t) into the brake current I(t) and dynamically modifies the torque constant K t (T), the current-torque relationship is:

[0058] T l (t) = K t (T)·I 2 (t)

[0059] Where: T l (t) is the actual braking torque, K t (T) is the dynamic correction torque constant, T is the operating temperature of the eddy current brake coil;

[0060] The braking torque is transmitted to the servo motor through the gear set, forming a complete electromechanical coupling system;

[0061] Its dynamic behavior is described by the second-order differential equation:

[0062]

[0063] Among them: J eq is the equivalent moment of inertia, is the angular acceleration, b eq is the equivalent damping coefficient, is the angular velocity, T m (t) is the output torque of the servo motor, T l (t) is the actual braking torque, i is the gearbox speed ratio;

[0064] When the load changes suddenly, the PSO algorithm monitors the error change rate. Automatically trigger weight optimization.

[0065] The present invention provides a test machine for simulating the load of a servo motor of an injection molding machine and an adaptive control method thereof, which have the following advantages:

[0066] (1) The testing machine provided by the present invention can replace the actual injection molding machine in a laboratory environment to complete the performance test and parameter adjustment of the servo drive system by accurately simulating the dynamic load characteristics of the entire cycle of the injection molding process, eliminating the production equipment occupation, raw material consumption and energy consumption costs, greatly shortening the R&D verification cycle, and at the same time reducing equipment maintenance costs and site requirements, significantly improving test efficiency and economy.

[0067] (2) This invention adopts a collaborative control strategy based on the particle swarm optimization algorithm (PSO) and the BP neural network PID controller. By collecting the load torque error, set value, and measured value in real time, a three-layer neural network input is constructed to dynamically adjust the PID parameter mapping relationship. The PSO algorithm optimizes the neural network weights online. Combined with the rapid response characteristics of the eddy current brake, a closed-loop control chain of "error perception-intelligent decision-making-precise execution" is formed to achieve high steady-state control accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a specific structural diagram of the servo motor load testing machine for simulating an injection molding machine according to the present invention;

[0069] Figure 2 This is a structural flow chart of the testing machine and the adaptive control method of the present invention;

[0070] Figure 3 This is a flow chart of the adaptive control method of the present invention;

[0071] Explanation of marks in the figure: 1. Servo motor; 2. Motor bracket; 3. Torque sensor; 4. Drive shaft; 5. Bearing support 1; 6. Gear set; 7. Bearing support 2; 8. Coupling 3; 9. Eddy current brake; 10. Brake support; 11. Coupling 2; 12. Torque sensor support; 13. Coupling 1. DETAILED DESCRIPTION

[0072] In order to better understand the purpose, structure and function of the present invention, the following is a further detailed description of a test machine for simulating the load of a servo motor of an injection molding machine and an adaptive control method of the present invention in conjunction with the accompanying drawings.

[0073] like Figure 1 As shown, the present invention is a test machine for simulating the load of a servo motor of an injection molding machine, comprising a motor input unit, a gear reduction unit, a load unit, and a monitoring unit. The motor input unit is connected to the load unit through the monitoring unit and the gear reduction unit. Specifically:

[0074] The motor input unit includes a servo motor 1, a coupling 13, and a motor bracket 2. The motor bracket 2 is fixed on the platform plane, and the servo motor 1 is fixed on the motor bracket 2. The output shaft of the servo motor 1 is connected to the coupling 13, and the coupling 13 is connected to the monitoring unit.

[0075] The monitoring unit includes a torque sensor 3, a coupling 2 11 and a torque sensor support 12. The torque sensor support 12 is fixed on the platform plane, and the torque sensor 3 is fixed on the torque sensor support 12. One end of the torque sensor 3 is connected to the coupling 1 13 through an internal shaft, and the other end is connected to the coupling 2 11 through an internal shaft. The coupling 2 11 is connected to the gear reduction unit.

[0076] The gear reduction unit includes a transmission shaft 4, bearing support 1 5, bearing support 2 7, gear set 6, and coupling 3 8. Gear set 6 includes a large gear and a pinion. Bearing support 1 5 and bearing support 2 7 are fixed to a flat platform and support the axles of the pinion and large gear, respectively. One end of transmission shaft 4 is connected to coupling 2 11, and the other end is connected to the axle of the pinion. The pinion meshes with the large gear, and the axle of the large gear is connected to coupling 3 8, which is connected to the load unit. The transmission of gear set 6 achieves a speed reduction effect.

[0077] The load unit includes an eddy current brake 9 and a brake support 10. The brake support 10 is fixed on the platform plane, and the eddy current brake 9 is fixed on the brake support 10. The eddy current brake 9 is connected to the coupling 3 8.

[0078] In order to improve the machinability of the entire platform and optimize the structural design, the gear set 6 adopts a horizontal meshing method with two axes to ensure that the two bearing seats are exactly the same size. All the matching methods of the shaft body adopt the key and keyway matching method, and the gear meshing adopts a horizontal meshing method. This can ensure that the two axis centers that match the gear set 6 are at the same height, so that the bearing seats designed with the same size can be used to fix two shafts in different positions.

[0079] In order to ensure that the center heights of the components of the entire platform are the same, the torque sensor 3 and the eddy current brake 9 are provided with pads at the bottom.

[0080] To ensure a continuous and gradual transition between no-load and full-load operation, the eddy-current brake 9 is controlled by a programmable logic controller (PLC) control unit, ensuring data transmission with the host computer to monitor the output torque during operation. The PLC control unit incorporates an adaptive algorithm that monitors the injection molding machine's load parameters in real time and makes adaptive adjustments based on these load characteristics. This algorithm dynamically adjusts the eddy-current brake's operating state based on real-time load changes, achieving precise control.

[0081] The monitoring unit is responsible for real-time monitoring of the injection molding machine load parameters, such as load force, speed, etc. Sensors or other measuring devices can be used to obtain these load parameters and feed them back to the control unit.

[0082] Based on real-time monitored load parameters, the PLC control unit uses the BP-PSO-PID control algorithm to adjust the current and phase of the eddy current brake. By continuously optimizing the control strategy, precise control of the simulated injection molding machine load can be achieved.

[0083] The entire testing machine system needs to be integrated and optimized to achieve a smooth workflow. Reasonable connections and data exchange are required between the eddy current brake, control unit, and monitoring unit to ensure accurate real-time monitoring and control.

[0084] like Figure 2 As shown, in this embodiment, the adaptive control method of the above-mentioned simulated injection molding machine servo motor load testing machine can adopt the following steps:

[0085] Step 1: After the system starts, initialize the test machine hardware and set the target load curve. Set the servo motor operating mode to torque control, set the eddy current brake pre-excitation current, zero the torque sensor, eliminate the initial offset, and apply the preset load curve r(t).

[0086] Step 2: Use a high-precision torque sensor to collect the actual load signal y(t) and compare it with the target value r(t) to obtain the error e(t):

[0087] e(t)=r(t)-y(t)

[0088] The error e(t), set value r(t) and output value y(t) are input into the three-layer BP neural network, and the number of PSO parameters and inertia weight are set. The neural network output is optimized online by the particle swarm algorithm to generate the optimal PID parameter K p ,K i ,K d .

[0089] Step 3: Parameters are input into a digital PID controller to generate a control voltage. After power amplification, the eddy current brake is driven to produce precise braking torque, which is transmitted to the servo motor through a gear set to form a closed-loop control. During this process, the PSO algorithm continuously optimizes the neural network weights, enabling the system to adaptively adjust parameters to adapt to dynamic load changes at each stage of the injection molding process.

[0090] like Figure 3 As shown, the method of the present invention is specifically described below through examples.

[0091] Step 1: After the system is started, the hardware units of the test machine are initialized and configured. The servo motor working mode is configured to torque control mode. Its dynamic equation is as follows:

[0092]

[0093] Among them, T m is the motor output torque (unit: Nm); I q is the torque current of the servo motor; K t is the torque constant, set at 1.2Nm / A; J m is the motor rotor inertia, ω is the real-time speed, and B is the damping coefficient.

[0094] The initial excitation current I0 applied to the eddy current brake is 0.4Nm / A, and its current-torque relationship is:

[0095] T l =K t (T)·I 2

[0096] Where: T l is the load torque, T is the brake temperature, and is executed through the PLC analog output module; K t (T) represents the temperature-dependent eddy current brake torque coefficient, and I is the brake excitation current. The torque sensor is automatically zeroed under no-load conditions.

[0097] Generate a load curve and define a three-segment load curve r(t) to simulate the injection molding process:

[0098] (1) Injection stage (0≤t<0.1s), the load increases linearly to 150Nm, and the slope reflects the injection speed:

[0099]

[0100] (2) During the pressure holding stage (0.1 ≤ t < 0.5 s), Gaussian noise is superimposed to simulate hydraulic fluctuations:

[0101]

[0102] (3) Melting stage (t≥0.5s), sinusoidal fluctuation simulates screw retraction:

[0103] r(t)=80+30sin(4π(t-0.5))

[0104] Step 2: First, the actual load signal y(t) is collected by a high-precision torque sensor and compared with the target value r(t) to obtain the control error e(t):

[0105] e(t)=r(t)-y(t)

[0106] Take e(t), r(t), and y(t) as the 3-node input vector of the BP neural network, set the hidden layer to 7 nodes, and obtain the output PID parameters through the forward propagation formula:

[0107] h(t)=ReLU(W ih I(t)+b h )

[0108] [K p ,K i ,K d ] T =W ho h(t)+b o

[0109] Among them: K p ,K i ,K d is the PID output parameter, I(t) represents the normalized control signal input vector [r(t); y(t); e(t)], h(t) is the output vector of the hidden layer, W ih is the weight matrix from the input layer to the hidden layer, W ho is the weight matrix from the hidden layer to the output layer, b h is the bias vector of the hidden layer, b o is the bias vector of the output layer, and ReLU(·) represents the rectified linear unit activation function.

[0110] The PSO weight optimization is performed on the output parameters. The specific optimization process of PSO is as follows:

[0111] First, initialize the PSO-BP parameters, the number of particles N is 50, and each particle encodes all the weights of the BP network:

[0112]

[0113] Where: X i Represents the position vector of the i-th particle, encoding all adjustable parameters of the BP neural network; W ih 、Who 、b h 、b o Same meaning as above.

[0114] The fitness function is:

[0115]

[0116] Where: e(t) is the control error, u(t) is the PID output control quantity.

[0117] Then update the particle position and velocity:

[0118]

[0119]

[0120] in, represents the velocity vector of the i-th particle at the k-th iteration, represents the position vector of the i-th particle at the k-th iteration, pbest i is the individual best historical position of the i-th particle, pbest represents the global best historical position (the best solution among all particles), w(k) is the inertia weight at the k-th iteration, which decreases from 0.9 to 0.4, k is the current number of iterations, k max is the maximum number of iterations. c1, c2 are learning factors, and r1, r2 are random numbers between 0 and 1.

[0121] Finally, the global optimal weight pbest is output, and the optimal weight W is obtained by decoding pbest. ih * (:),W ho * (:),b h * ,b o * , input the real-time input I(t) = [r(t); y(t); e(t)] into the optimized BP network,

[0122] h(t)=ReLU(W ih * I(t)+b h * )

[0123] [K p ,K i ,K d ] T =W ho * h(t)+b o *

[0124] After parameter denormalization, the optimal PID parameter K optimized by PSO is generated. p * ,K i * ,K d * .

[0125] Step 3: The PLC controller uses the position PID algorithm, and the optimized PID parameter K p * ,K i * ,K d * , update the control voltage u(t) every 1ms:

[0126]

[0127] Where: T s is the control period, which is 1ms; e(k) is the error at the current moment; ∑e(j) is the error integral term; is the error differential term.

[0128] The power amplifier converts the voltage u(t) into the brake current I(t) and dynamically modifies the torque constant K t (T), the current-torque relationship is:

[0129] T l (t) = K t (T)·I 2 (t)

[0130] Where: T l (t) is the actual braking torque, K t (T) is the dynamic correction torque constant, and T is the operating temperature of the eddy current brake coil.

[0131] The braking torque is transmitted to the servo motor through the gear set, forming a complete electromechanical coupling system.

[0132] Its dynamic behavior can be described by the second-order differential equation:

[0133]

[0134] Among them: J eq is the equivalent moment of inertia, is the angular acceleration, b eq is the equivalent damping coefficient, is the angular velocity, T m (t) is the output torque of the servo motor, T l (t) is the actual braking torque, and i is the gearbox speed ratio.

[0135] When the load changes suddenly, the PSO algorithm monitors the error change rate. Automatically trigger weight optimization. For example, in the injection stage, PSO guides the neural network to increase the proportional coefficient through the fitness function, while in the pressure holding stage, the weight is adjusted to optimize the steady-state performance.

[0136] It will be understood that the present invention is described by way of some embodiments, and it will be appreciated by those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are intended to be protected by the present invention.

Claims

1. A test machine for simulating the load of a servo motor of an injection molding machine, comprising a motor input unit, a gear reduction unit, a load unit, and a monitoring unit, characterized in that: The motor input unit comprises a servo motor (1), a coupling 1 (13) and a motor bracket (2), wherein the motor bracket (2) is fixed on a platform plane, the servo motor (1) is fixed on the motor bracket (2), the output shaft of the servo motor (1) is connected to the coupling 1 (13), and the coupling 1 (13) is connected to the monitoring unit; the monitoring unit comprises a torque sensor (3), a coupling 2 (11) and a torque sensor support (12), wherein the torque sensor support (12) is fixed on the platform plane, the torque sensor (3) is fixed on the torque sensor support (12), one end of the torque sensor (3) is connected to the coupling 1 (13) through an internal shaft, and the other end is connected to the coupling 2 (11) through an internal shaft, and the coupling 2 (11) is connected to the gear reduction unit; the gear reduction unit comprises The invention comprises a transmission shaft (4), a bearing support 1 (5), a bearing support 2 (7), a gear set (6), and a coupling 3 (8), wherein the gear set (6) comprises a large gear and a small gear, the bearing support 1 (5) and the bearing support 2 (7) are fixed on the platform plane and are used to support the axles of the small gear and the large gear respectively, one end of the transmission shaft (4) is connected to the coupling 2 (11), and the other end is connected to the axle of the small gear, the small gear is meshed with the large gear, the axle of the large gear is connected to the coupling 3 (8), and the coupling 3 (8) is connected to the load unit; the load unit comprises an eddy current brake (9) and a brake support (10), the brake support (10) is fixed on the platform plane, and the eddy current brake (9) is fixed on the brake support (10); the eddy current brake (9) is connected to the coupling 3 (8).

2. The test machine for simulating the load of the servo motor of an injection molding machine according to claim 1, characterized in that: The gear set (6) is meshed in a manner that the two axes are horizontal, and all the matching modes of the shaft bodies are the matching modes of keys and keyways.

3. The test machine for simulating the load of the servo motor of an injection molding machine according to claim 1, characterized in that: Pads are provided at the bottoms of the torque sensor (3) and the eddy current brake (9).

4. The test machine for simulating the load of the servo motor of an injection molding machine according to claim 1, characterized in that: A PLC control unit is used to control the eddy current brake (9). The PLC control unit includes an adaptive algorithm. The algorithm monitors the injection molding machine load parameters in real time and performs adaptive adjustment according to the load characteristics. Based on the real-time monitored load parameters, the PLC control unit adjusts the current and phase of the eddy current brake through a BP-PSO-PID control algorithm. By continuously optimizing the control strategy, precise control of the simulated injection molding machine load is achieved.

5. An adaptive control method for a test machine for simulating the load of a servo motor of an injection molding machine according to any one of claims 1 to 4, characterized in that: The steps include: Step 1: After the system is started, the hardware of the test machine is initialized and the target load curve is set. The servo motor is set to torque control mode, the eddy current brake pre-excitation current is set, the torque sensor is zeroed, and the initial offset is eliminated. Simultaneously load the preset load curve r(t); Step 2: Use a high-precision torque sensor to collect the actual load signal y(t) and compare it with the target value r(t) to obtain the error e(t): e(t)=r(t)-y(t) The error e(t), set value r(t) and output value y(t) are input into the three-layer BP neural network, and the number of PSO parameters and inertia weight are set. The neural network output is optimized online by the particle swarm algorithm to generate the optimal PID parameter K p ,K i ,K d ; Step 3: Parameters are input into a digital PID controller to generate a control voltage. After power amplification, the eddy current brake is driven to produce precise braking torque, which is transmitted to the servo motor through a gear set to form a closed-loop control. During this process, the PSO algorithm continuously optimizes the neural network weights, enabling the system to adaptively adjust parameters to adapt to dynamic load changes at each stage of the injection molding process.

6. The adaptive control method according to claim 5, characterized in that: The step 1 includes the following steps: Configure the servo motor working mode to torque control mode, and its dynamic equation is as follows: Among them, T m is the motor output torque (unit: Nm); I q is the torque current of the servo motor; K t is the torque constant, set at 1.2Nm / A; J m is the motor rotor inertia, ω is the real-time speed, and B is the damping coefficient; The initial excitation current I0 applied to the eddy current brake is 0.4Nm / A, and its current-torque relationship is: T l =K t (T)·I 2 Where: T l is the load torque, T is the brake temperature, and is executed through the PLC analog output module; K t (T) represents the temperature-related eddy current brake torque coefficient, I is the brake excitation current, and the torque sensor is automatically zeroed under no-load conditions; Generate a load curve and define a three-segment load curve r(t) to simulate the injection molding process: (1) Injection stage: 0≤t<0.1s, load increases linearly to 150Nm, the slope reflects the injection speed: (2) Pressure holding stage: 0.1≤t<0.5s, superimposed Gaussian noise to simulate hydraulic fluctuations: (3) Melting stage: t≥0.5s, sinusoidal fluctuation simulates screw retraction: r(t)=80+30sin(4π(t-0.5)).

7. The adaptive control method according to claim 5, characterized in that: The second step includes the following steps: The control error e(t) is: e(t)=r(t)-y(t) Take e(t), r(t), and y(t) as the 3-node input vector of the BP neural network, set the hidden layer to 7 nodes, and obtain the output PID parameters through the forward propagation formula: h(t)=ReLU(W ih I(t)+b h ) [K p ,K i ,K d ] T =W ho h(t)+b o Among them: K p ,K i ,K d is the PID output parameter, I(t) represents the normalized control signal input vector [r(t); y(t); e(t)], h(t) is the output vector of the hidden layer, W ih is the weight matrix from the input layer to the hidden layer, W ho is the weight matrix from the hidden layer to the output layer, b h is the bias vector of the hidden layer, b o is the bias vector of the output layer, ReLU(·) represents the rectified linear unit activation function; The PSO weight optimization is performed on the output parameters. The specific optimization process of PSO is as follows: First, initialize the PSO-BP parameters, the number of particles N is 50, and each particle encodes all the weights of the BP network: Where: X i Represents the position vector of the i-th particle, encoding all adjustable parameters of the BP neural network; W ih 、W ho 、b h 、b o Same meaning as above; The fitness function is: Among them: e(t) is the control error, u(t) is the PID output control quantity; Then update the particle position and velocity: in, represents the velocity vector of the i-th particle at the k-th iteration, represents the position vector of the i-th particle at the k-th iteration, pbest i is the individual best historical position of the i-th particle, gbest represents the global best historical position (the best solution among all particles), w(k) is the inertia weight at the k-th iteration, which decreases from 0.9 to 0.4, k is the current iteration number, k max is the maximum number of iterations. c1, c2 are learning factors, r1, r2 are random numbers between 0 and 1; Finally, the global optimal weight gbest is output, and the optimal weight W is obtained by decoding gbest. ih * (:),W ho * (:),b h * ,b o * , input the real-time input I(t) = [r(t); y(t); e(t)] into the optimized BP network, h(t)=ReLU(W ih * I(t)+b h * ) [K p ,K i ,K d ] T =W ho * h(t)+b o * After parameter denormalization, the optimal PID parameter K optimized by PSO is generated. p * ,K i * ,K d * .

8. The adaptive control method according to claim 5, characterized in that: The step three includes the following steps: The PLC controller uses a position PID algorithm, and the optimized PID parameter K p * ,K i * ,K d * , update the control voltage u(t) every 1ms: Where: T s is the control period, which is 1ms; e(k) is the error at the current moment; ∑e(j) is the error integral term; is the error differential term; The power amplifier converts the voltage u(t) into the brake current I(t) and dynamically modifies the torque constant K t (T), the current-torque relationship is: T l (t)=K t (T)·I 2 (t) Where: T l (t) is the actual braking torque, K t (T) is the dynamic correction torque constant, T is the operating temperature of the eddy current brake coil; The braking torque is transmitted to the servo motor through the gear set, forming a complete electromechanical coupling system; Its dynamic behavior is described by the second-order differential equation: Among them: J eq is the equivalent moment of inertia, is the angular acceleration, b eq is the equivalent damping coefficient, is the angular velocity, T m (t) is the output torque of the servo motor, T l (t) is the actual braking torque, i is the gearbox speed ratio; When the load changes suddenly, the PSO algorithm monitors the error change rate. Automatically trigger weight optimization.

Images