Method and device for measuring on-orbit self-adaptive minimum starting torque of whole spaceflight motor

Abstract

The invention is suitable for the technical field of aerospace scanning stepping motor control, and provides a method and a device for measuring the overall on-orbit self-adaptive minimum starting torque of an aerospace motor, a storage medium and a processor, and the measurement method optimizes the overall on-orbit self-adaptive minimum starting torque measurement of the aerospace motor based on a Q algorithm. And through collaborative design of discretized parameters and multi-dimensional reward / penalty terms, the adaptation problem of a traditional Q algorithm in a motor parameter measurement scene is solved. According to the method, the minimum starting current meeting the error requirement can be rapidly measured in real time, measurement can be implemented in the normal operation state of the motor, the minimum starting current of the whole machine is determined according to the estimated current when the error is minimum, and then the minimum starting torque of the whole machine is obtained through calculation. Compared with the prior art, the estimated value is more accurate, and the method is suitable for any environment and scene in which the motor can be operated. The device adopting the method, the storage medium storing the method and the processor running the method also have the same technical effects.

Description

Technical Field

[0001] The present application belongs to the field of aerospace scanning stepper motor control technology, and in particular relates to an on-orbit adaptive minimum starting torque measurement method and device for an aerospace motor, a storage medium, and a processor. Background Art

[0002] The current selection of aerospace scanning stepper motor drive torque is usually based on the minimum starting torque plus torque margin. However, the existing minimum starting torque can only be measured through calculation and estimation based on system design parameters, or roughly tested using counterweights on customized tooling. It is difficult to accurately measure the actual minimum starting torque of the entire machine under normal load.

[0003] Especially in harsh conditions such as vacuum and low temperature on orbit, the minimum starting torque of the scanning stepper motor will change, and it is difficult to simulate and measure it on the ground.

[0004] The Q-learning algorithm (hereinafter referred to as the Q-algorithm) is a model-free reinforcement learning algorithm that learns optimal policies by iteratively updating the Q-function (state-action value function). It is suitable for decision-making problems involving discrete states and actions. However, typical Q-learning algorithms are characterized by a single goal orientation, such as minimizing error to maximize reward. When applied to calculating motor parameters, this approach can lead to excessive pursuit of error minimization, resulting in uncontrolled current increases; frequent current fluctuations, affecting motor stability and lifespan; and an inability to identify the critical point between error and current. Therefore, existing technologies have shortcomings. Summary of the Invention

[0005] This application aims to provide an on-orbit adaptive minimum starting torque measurement method and device for aerospace motors. Through discretization and the collaborative design of multi-dimensional reward / penalty terms, this method addresses the inaccurate measurement of traditional motor minimum starting torque and the inability to measure under extremely challenging conditions. Measurements can be performed even when the motor is operating normally. This method is applicable to a variety of harsh and extreme experimental conditions and is of great significance to the system design of scanning motors.

[0006] In one aspect, the present application provides an on-orbit adaptive minimum starting torque measurement method for an aerospace motor, the measurement method comprising the following steps: s1. Initialize the Q table from the discretized motor operating state parameters and the discretized current adjustment action parameters, set the corresponding learning rate and discount factor, and initialize the sliding average filter; s2. The actual motor rotation angle is acquired in real time through an encoder, and after calculating the deviation from the target angle, the deviation state and the corresponding motor current state are mapped to the discrete motor operating state parameters; the current is adjusted by changing the PWM value to achieve the change of the motor output torque, and the current adjustment amplitude generated by the current state is mapped to the discrete current adjustment action parameters; s3. Designing a reward function based on the association between the deviation state and the corresponding current state, as well as the current adjustment amplitude taken, to avoid the Q algorithm falling into a single goal guide; s4. Use the ε-greedy strategy to select actions to execute and converge the Q table; s5. Calculate the new state and reward and update the Q value; s6. If the convergence condition is met, the minimum starting current is output, otherwise repeat steps s2-s5; s7. Based on the linear relationship between current and torque, the minimum starting torque is calculated according to the minimum starting current.

[0007] On the other hand, the present application also provides an on-orbit adaptive minimum starting torque measurement device for an aerospace motor, which adopts the above-mentioned measurement method; the device includes a main control CPU, an interface expansion logic FPGA, a scanning motor and an encoder electrically connected in sequence; the encoder collects the rotation angle of the scanning motor and feeds it back to the main control CPU; The main control CPU determines the mechanism rotation angle information fed back by the encoder, iteratively calculates and updates the motor control current parameters, and controls the PWM duty cycle through the interface expansion logic FPGA to update the control current; ultimately, the minimum starting current of the whole machine is determined, and the minimum starting torque of the whole machine is obtained by calculation.

[0008] On the other hand, the present application also provides a storage medium storing a program file capable of implementing the above-mentioned on-orbit adaptive minimum starting torque measurement method for aerospace motors.

[0009] On the other hand, the present application also provides a processor, which is used to run a program, wherein the program executes the above-mentioned on-orbit adaptive minimum starting torque measurement method for aerospace motors when running.

[0010] In this application's measurement method, the Q-algorithm-based motor's overall minimum starting torque is discretized and designed in conjunction with multi-dimensional reward / penalty terms, addressing the difficulties of adapting the traditional Q-algorithm to motor parameter measurement scenarios. This Q-algorithm can measure the minimum starting current that meets error requirements. Measurements can be performed when the motor is in a normal state.

[0011] The measuring device of the present application runs the measuring method, and its main control CPU determines the motor rotation angle information fed back by the encoder, takes the rotation angle under normal load as close to the step angle as possible as a standard, iteratively calculates and updates the motor control current parameters, and realizes the update of the motor control current by controlling the PWM duty cycle. Among them, the main control CPU updates the PWM parameters to the interface expansion logic FPGA, the interface expansion logic FPGA controls the driver chip, sends a control signal to the scanning motor, controls the scanning motor to rotate an angle, the encoder collects the rotation angle of the scanning motor and feeds it back to the main control CPU, the main control CPU determines whether the difference between the rotation angle of the scanning motor and the step angle meets the error standard, and determines the minimum starting current of the whole machine with the estimated current at the minimum error, and then calculates the minimum starting torque of the whole machine. Compared with the value estimated by the prior art, it is more accurate and applicable to any environment and scenario where the motor can be operated. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of the on-orbit adaptive minimum starting torque measurement method for an aerospace motor provided in Example 1 of the present application; Figure 2 This is a module diagram of the on-orbit adaptive minimum starting torque measurement device for an aerospace motor provided in Example 2 of the present application. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0014] The following describes the specific implementation of this application in detail with reference to specific embodiments: Example 1: Figure 1 The following is an implementation flow of the method for measuring the minimum starting torque of a scanning stepper motor provided in the first embodiment of the present application. For ease of illustration, only the part related to the embodiment of the present application is shown, which is detailed as follows: In one aspect, the present application provides an on-orbit adaptive minimum starting torque measurement method for an aerospace motor, the measurement method comprising the following steps: s1. Initialize the Q table from the discretized motor operating state parameters and the discretized current adjustment action parameters, set the corresponding learning rate and discount factor, and initialize the sliding average filter; s2. The actual motor rotation angle is acquired in real time through an encoder, and after calculating the deviation from the target angle, the deviation state and the corresponding motor current state are mapped to the discrete motor operating state parameters; the current is adjusted by changing the PWM value to achieve the change of the motor output torque, and the current adjustment amplitude generated by the current state is mapped to the discrete current adjustment action parameters; s3. Designing a reward function based on the association between the deviation state and the corresponding current state, as well as the current adjustment amplitude taken, to avoid the Q algorithm falling into a single goal guide; s4. Use the ε-greedy strategy to select actions to execute and converge the Q table; s5. Calculate the new state and reward and update the Q value; s6. If the convergence condition is met, the minimum starting current is output, otherwise repeat steps s2-s5; s7. Based on the linear relationship between current and torque, the minimum starting torque is calculated according to the minimum starting current.

[0015] In general, the calculation formula for the minimum dynamic torque of the moving parts is: ; Where: - Minimum driving torque of moving parts, in Nm ( ); ——Dynamic inertia moment of moving parts, in N·m ( ); ——The maximum resistance torque of the moving parts, in N·m ( ).

[0016] 1. The formula for load inertia resistance moment is: ; Where: — total moment of inertia of the rotating mechanism; ——maximum angular acceleration; The friction torque of the bearing is ; The rotational resistance torque of the harness is ; Therefore: ; In actual measurement, the friction torque of the bearing Rotational resistance torque with harness It is difficult to measure accurately, especially in extreme conditions, where these resistance torques fluctuate greatly. , here we consider that the output torque and current have a linear relationship, that is: .

[0017] In this way, the minimum output torque of the aerospace scanning stepper motor can be indirectly measured based on the minimum starting current.

[0018] In this paper, we use a table-based Q-learning algorithm to discretize the motor's operating state parameters (the deviation between the actual rotation angle and the target angle, the current) and the action space parameters (control PWM changes), divide them into a finite number of state intervals, and construct a complete state space.

[0019] Furthermore, in step s1, the state space of the Q table is defined by discretized motor operating state parameters, constituting rows of the Q table; The action space of the Q table is defined by discretized current adjustment action parameters, which constitute the columns of the Q table.

[0020] Further, the number of rows of the Q table is the product of the number of the deviation states and the number of the current states; The number of columns in the Q table corresponds to the number of times the current adjustment amplitude is adjusted; The dimension of the Q table is the product of the number of rows and the number of columns.

[0021] Specifically, the rows (states) and columns (actions) of the Q-table are completely defined by discretized parameters, and the granularity of the parameters directly determines the size of the Q-table. The state space is "5×5" (25 states in total), and the action space consists of five discrete actions, resulting in a Q-table dimension of 25 × 5 (25 rows of states, 5 columns of actions). Initially, all Q-values ​​are set to 0. By dividing continuous quantities such as angle deviation and current into discrete intervals and quantizing current regulation actions into fixed amplitudes, the algorithm can learn the "state-action-reward" mapping relationship within a limited Q-table. This discretization helps balance accuracy and computational complexity, and supports the closed-loop control of "continuous input → discrete decision → continuous output" throughout the algorithm's initialization, state perception, action selection, reward calculation, Q-value update, and convergence determination steps.

[0022] Furthermore, in the process of designing the reward function in step s3, the deviation state is divided into the following five intervals: 1. When the angular error percentage between the actual rotation angle and the target angle is greater than 20%, it is determined that the actual rotation angle is much greater than the target angle; 2. When the angular error percentage between the actual rotation angle and the target angle is greater than 10% and less than 20%, the actual rotation angle is determined to be slightly greater than the target angle: 3. When the angular error percentage between the actual rotation angle and the target angle is ≥-10% and ≤10%, the actual rotation angle is determined to be close to the target angle; 4. When the angular error percentage between the actual rotation angle and the target angle is less than -10% and greater than -20%, the actual rotation angle is determined to be slightly smaller than the target angle; 5. When the angular error percentage between the actual rotation angle and the target angle is less than -20%, it is determined that the actual rotation angle is much smaller than the target angle; The current state is divided into the following five intervals: 1. The measured current is less than 20% of the rated current, which is considered to be in the extremely low current range; 2. When the measured current is ≥20% and <25% of the rated current, it is determined to be in the low current range; 3. When the measured current is ≥25% and ≤30% of the rated current, it is determined to be in the medium current range; 4. When the measured current is greater than 30% and less than or equal to 35% of the rated current, it is considered to be in the high current range; 5. When the measured current is greater than 35% of the rated current, it is determined to be in the extremely high current range; The current adjustment range is divided into the following five intervals: 1. If the current increases by 1% of the PWM value based on the current, it is considered a slight increase in current; 2. If the current decreases by 1% of the PWM value based on the current, it is judged as a slight current decrease; 3. If the current increases by 10% of the PWM value, it is considered a significant increase in current. 4. If the current decreases by 10% of the PWM value based on the current, it is determined to be a significant current reduction; 5. Keep the current PWM value unchanged and determine it as maintaining current.

[0023] Furthermore, in the process of designing the reward function in step s3, the reward function is constructed by using a combination of basic reward + penalty term; the reward function is: r t = Basic reward - Angle error penalty - Current penalty - Fluctuation penalty + Critical point exploration reward; Among them, the basic reward is used to encourage the angle error to approach the target, which is generally limited to the range of ±2°; the angle error penalty is designed based on the idea that the larger the error, the heavier the penalty, and is used to suppress large errors; the current penalty is also designed based on the idea that the higher the current, the heavier the penalty, and is used to encourage the realization of low current state; the fluctuation penalty adopts the design idea that the more drastic the current change, the heavier the penalty, and is used to suppress frequent current adjustments; the critical point exploration reward is used to give additional rewards when the current decreases and the error does not increase significantly to guide the search for a balance point with low current and small error.

[0024] The following scheme is used to implement the reward function: 1. Basic Rewards When the motor is in the "actual rotation angle close to the target angle" state, a positive reward of +10 is given; When the motor is in the state of "the actual rotation angle is slightly greater than the target angle" or "the actual rotation angle is slightly less than the target angle", a positive reward of +5 is given; When the motor is in the "actual rotation angle is much greater than the target angle" or "actual rotation angle is much less than the target angle" state, it indicates that the current current torque is far from sufficient for the motor to complete the target angle rotation under load. A significant increase in current is required, and a basic negative reward of -5 is given.

[0025] 2. Angle error penalty Angle error penalty = absolute value of error percentage * penalty coefficient; The penalty coefficient can be set according to specific circumstances. In the present invention, it is set to 0.5.

[0026] 3. Current Penalty The higher the current level, the greater the penalty (incentivizing lower current): Very low current: 0; Low current: 0.1 × PWM duty cycle; Medium current: 0.2 × PWM duty cycle; High current: 0.5 × PWM duty cycle; Very high current: 1.0 × PWM duty cycle; 4. Volatility Penalty Calculate the absolute difference between the current and the current in the previous cycle, that is: Fluctuation penalty = (current cycle PWM duty cycle - previous cycle PWM duty cycle) * fluctuation penalty coefficient In the present invention, the fluctuation penalty coefficient is set to 0.3. 5. Tipping Point Exploration Rewards When the current is reduced without a significant increase in error, an additional bonus is given: If the current is less than the current in the previous cycle and the angle error change is ≤5%, reward +3; If the current is less than the current in the previous cycle and the angle error change is ≤10%: reward +1; When the current increases but the error decreases significantly, a modest reward is given: If the current is greater than the current in the previous cycle and the angle error is reduced by ≥15%: reward +2; Furthermore, the Q value in step s5 represents the comprehensive benefit of the minimum starting current and the minimum angle error that can be obtained after taking a certain action in a certain state; the calculation formula for updating the Q value is as follows: ; in: t For the current cycle; is the current state; The action currently being taken; is the learning rate; is the immediate reward calculated according to the reward function; The reward after the action is executed for k cycles. The Q value can more accurately evaluate the long-term effect of the action and avoid policy mislearning caused by time lag or single error. is the time lag compensation factor, which is set to 0.8 in the present invention; is the discount factor, which indicates the degree of influence of future decisions on current decisions and is set to 0.95 in this invention; To perform an action The new state after New state The maximum Q value among all possible actions.

[0027] Specifically, in the Q-value update mechanism, the discretization parameters define "what the algorithm needs to learn" (the granularity of the state, the granularity of the action, and the boundaries of the physical constraints); the Q-value parameters define "how to learn" (the learning speed, the degree of attention to future impacts, and the intensity of exploration). The state discretization granularity (the number of intervals) corresponds to the learning rate α; the larger the state space, the smaller α. The action adjustment amplitude (step size) corresponds to the discount factor γ. A larger action adjustment amplitude requires a larger γ. The number of actions corresponds to the exploration rate ε: a larger number of actions requires a higher initial ε and a slower decay rate.

[0028] At the same time, the calculation process introduces a time delay compensation factor , which helps to more accurately evaluate the long-term effects of actions and avoid strategy mislearning caused by time lag or single error.

[0029] Furthermore, in step s6, the exploration rate is set , when the test starts Set to 0.5, and then decay to 0.1, using a greedy strategy. The probability of randomly selecting an action is The probability of choosing the current state Take the action with the largest Q value; gradually reduce the exploration rate based on the state and action feedback , until the angular error percentage and current change indicators meet the standards, the Q value table converges, and the motor can run stably with the minimum current.

[0030] Specifically, during step s6, when the algorithm converges (i.e., the Q value changes by less than 0.01 for 10 consecutive cycles), the "critical point fine-tuning" mode is initiated: a "minor current reduction" action is forced with a 5% probability. If the error does not increase significantly after five consecutive executions, the base current level is permanently reduced. If the absolute value of the angular error percentage increases again, the search is terminated, and the minimum starting current, and thus the minimum starting torque, is calculated based on the PWM duty cycle value before the absolute value of the angular error percentage increased.

[0031] During the convergence process, the essence of the Q value is to quantify the "combined benefit of minimum starting current and minimum angle error achieved after taking a certain action in a certain state." Its core role is reflected in three aspects, directly supporting the exploration and determination of the minimum starting current: 1. As a decision criterion for the "current-error" trade-off; 2. Accumulate and learn the mapping relationship of the “starting critical point”; 3. Suppress “over-adjustment” and “conservative strategies” during the startup process.

[0032] The above-mentioned measurement method of the present application also has the following technical effects: 1. Suitable for precise measurement of the load torque of moving mechanisms in extremely harsh conditions such as vacuum and low temperatures. While ensuring the normal operation of the moving mechanism under extremely harsh conditions, the system accurately measures the minimum starting torque and current of the motor, reducing power consumption and heat loss, avoiding resource waste, and improving the safety and stability of motor operation.

[0033] 2. No need to estimate the starting test current; the current increase and decrease strategy can be dynamically adjusted, achieving rapid algorithm convergence. Through continuous interaction with the motor environment, the Q-value table is autonomously optimized, and the optimal strategy is learned directly from state-action feedback, without requiring prior knowledge of the motor's physical model or load characteristics. Even if motor parameters drift slightly over time, the algorithm maintains optimal control by continuously iterating and updating the strategy.

[0034] 3. The Q-value table is continuously updated through iterative optimization. As the number of learning times increases, the algorithm's judgment of the optimal action for various states will become increasingly accurate. Even if a single test angle value is abnormal, it can respond quickly based on historical learning results to avoid excessive adjustments.

[0035] 4. Strong anti-interference ability, setting fluctuation penalty to reduce mechanical vibration during the measurement process.

[0036] 5. Critical point exploration and fine-tuning mode actively explores the limit of current reduction and finds the optimal current value that minimizes the operating angle error.

[0037] Example 2: On the other hand, as attached Figure 2 As shown, the present application also provides a device for measuring the overall minimum starting torque of a scanning stepper motor, which adopts the above method; the device includes a main control CPU, an interface expansion logic FPGA, a scanning motor and an encoder electrically connected in sequence; the encoder collects the rotation angle of the scanning motor and feeds it back to the main control CPU; The main control CPU determines the mechanism rotation angle information fed back by the encoder, iteratively calculates and updates the motor control current parameters, and controls the PWM duty cycle through the interface expansion logic FPGA to update the control current; ultimately, the minimum starting current of the whole machine is determined, and the minimum starting torque of the whole machine is obtained by calculation.

[0038] Based on the above-mentioned measuring device, taking a two-phase hybrid stepper motor as an example, a two-phase dual four-beat control method is adopted. The motor rotation angle is observed through the encoder, and the stepper motor accurately moves a step angle as the standard. The motor control current is updated through iterative calculation. The minimum starting torque of the scanning stepper motor in the whole machine state can be accurately measured, effectively improving the measurement accuracy of the minimum starting torque of the whole machine.

[0039] Example 3: On the other hand, the present application also provides a storage medium storing a program file capable of implementing the above-mentioned on-orbit adaptive minimum starting torque measurement method for aerospace motors.

[0040] Those skilled in the art will understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc.

[0041] Example 4: On the other hand, the present application also provides a processor, which is used to run a program, wherein the program executes the above-mentioned on-orbit adaptive minimum starting torque measurement method for aerospace motors when running.

[0042] In the embodiments of the present application, the method for measuring the on-orbit adaptive minimum starting torque of an aerospace motor can be implemented by corresponding hardware or software units. Each unit can be an independent hardware or software unit, or can be integrated into a single hardware or software unit, which is not intended to limit the present application. The specific implementation of each unit can be referred to the description of Example 1 and will not be repeated here.

[0043] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for measuring the minimum starting torque of an aerospace motor on-orbit, characterized in that: The measuring method comprises the following steps: s1. Initialize the Q table from the discretized motor operating state parameters and the discretized current adjustment action parameters, set the corresponding learning rate and discount factor, and initialize the sliding average filter; s2. The actual motor rotation angle is acquired in real time through an encoder, and after calculating the deviation from the target angle, the deviation state and the corresponding motor current state are mapped to the discrete motor operating state parameters; the current is adjusted by changing the PWM value to achieve the change of the motor output torque, and the current adjustment amplitude generated by the current state is mapped to the discrete current adjustment action parameters; s3. Designing a reward function based on the association between the deviation state and the corresponding current state, as well as the current adjustment amplitude taken, to avoid the Q algorithm falling into a single goal guide; s4. Use the ε-greedy strategy to select actions to execute and converge the Q table; s5. Calculate the new state and reward and update the Q value; s6. If the convergence condition is met, the minimum starting current is output, otherwise repeat steps s2-s5; s7. Based on the linear relationship between current and torque, the minimum starting torque is calculated according to the minimum starting current.

2. The measuring method according to claim 1, wherein In step s1, the state space of the Q table is defined by discretized motor operating state parameters, which constitute rows of the Q table; The action space of the Q table is defined by discretized current adjustment action parameters, which constitute the columns of the Q table.

3. The measuring method according to claim 2, wherein: described The number of rows of the Q table is the product of the number of the deviation states and the number of the current states; The number of columns in the Q table corresponds to the number of times the current adjustment amplitude is adjusted; The dimension of the Q table is the product of the number of rows and the number of columns.

4. The measuring method according to claim 3, wherein: In the process of designing the reward function in step s3, the deviation state is divided into the following five intervals: (1) When the angular error percentage between the actual rotation angle and the target angle is greater than 20%, it is determined that the actual rotation angle is much greater than the target angle; (2) When the angular error percentage between the actual rotation angle and the target angle is greater than 10% and less than or equal to 20%, it is determined that the actual rotation angle is slightly greater than the target angle: (3) When the angular error percentage between the actual rotation angle and the target angle is ≥-10% and ≤10%, it is determined that the actual rotation angle is close to the target angle; (4) When the angular error percentage between the actual rotation angle and the target angle is less than -10% and greater than or equal to -20%, it is determined that the actual rotation angle is slightly smaller than the target angle; (5) When the angular error percentage between the actual rotation angle and the target angle is less than -20%, it is determined that the actual rotation angle is much smaller than the target angle; The current state is divided into the following five intervals: (1) The measured current is less than 20% of the rated current, which is considered to be in the extremely low current range; (2) When the measured current is ≥20% and <25% of the rated current, it is determined to be in the low current range; (3) When the measured current is ≥25% and ≤30% of the rated current, it is determined to be in the medium current range; (4) When the measured current is greater than 30% and less than or equal to 35% of the rated current, it is determined to be in the high current range; (5) When the measured current is greater than 35% of the rated current, it is determined to be in the extremely high current range; The current adjustment range is divided into the following five intervals: (1) If the current increases by 1% of the PWM value based on the current, it is judged as a slight increase in current; (2) If the current decreases by 1% of the PWM value based on the current, it is judged as a slight decrease in current; (3) If the current increases by 10% of the PWM value based on the current, it is judged as a significant increase in current; (4) If the current decreases by 10% of the PWM value based on the current, it is judged as a significant current reduction; (5) Keep the current PWM value unchanged and determine it as maintaining current.

5. The measuring method according to claim 4, wherein: In the process of designing the reward function in step s3, the reward function is constructed by using a combination of basic reward + penalty term; the reward function is: r t = Basic reward - Angle error penalty - Current penalty - Fluctuation penalty + Critical point exploration reward; Among them, the basic reward is used to encourage the angle error to approach the target; the angle error penalty is used to suppress large errors; the current penalty is used to encourage the realization of the low current state; the fluctuation penalty is used to suppress frequent adjustments; the critical point exploration reward is used to give additional rewards to guide the search for a balance point with low current and small error when the current decreases and the error does not increase significantly.

6. The measuring method according to claim 1, wherein: The Q value in step s5 represents the comprehensive benefit of the minimum starting current and the minimum angle error that can be obtained after taking a certain action in a certain state. The calculation formula for updating the Q value is as follows: ； in: is the current state; The action currently being taken; is the learning rate; is the immediate reward calculated according to the reward function; The reward after the action is executed for k cycles. The Q value can more accurately evaluate the long-term effect of the action and avoid policy mislearning caused by time lag or single error. is the time lag compensation factor, which is set to 0.8 in the present invention; is the discount factor, which indicates the degree of influence of future decisions on current decisions and is set to 0.95 in this invention; To perform an action The new state after New state The maximum Q value among all possible actions.

7. The measuring method according to claim 1, wherein: In step s6, the exploration rate is set , when the test starts Set to 0.5, and then decay to 0.1, using a greedy strategy. The probability of randomly selecting an action is The probability of choosing the current state Take the action with the largest Q value; gradually reduce the exploration rate based on the state and action feedback , until the angular error percentage and current change indicators meet the standards, the Q value table converges, and the motor can run stably with the minimum current.

8. An on-orbit adaptive minimum starting torque measurement device for an aerospace motor, characterized in that: The measuring method according to any one of claims 1 to 7 is used; the device comprises a main control CPU, an interface expansion logic FPGA, a scanning motor, and an encoder electrically connected in sequence; the encoder collects the rotation angle of the scanning motor and feeds it back to the main control CPU; The main control CPU determines the mechanism rotation angle information fed back by the encoder, iteratively calculates and updates the motor control current parameters, and controls the PWM duty cycle through the interface expansion logic FPGA to update the control current; ultimately, the minimum starting current of the whole machine is determined, and the minimum starting torque of the whole machine is obtained by calculation.

9. A storage medium, characterized in that: The storage medium stores a program file capable of implementing the on-orbit adaptive minimum starting torque measurement method for an aerospace motor as described in any one of claims 1 to 7.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the on-orbit adaptive minimum starting torque measurement method for an aerospace motor as described in any one of claims 1 to 7.

Images