A Servo Motor Adaptive Energy-Saving Control Method
By deploying the excitation optimization algorithm on an external independent computing unit, the servo motor achieves efficient energy consumption optimization and dynamic response under variable load conditions, solving the problems of excitation current drift and noise interference, and improving the safety and stability of the system.
Patent Information
- Application Number
- CN202611486758.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-23
- Publication Date
- 2026-11-17
AI Technical Summary
Existing servo motors suffer from problems such as increased energy consumption and decreased torque response performance due to excitation current operating point drift under variable load conditions. Furthermore, existing control schemes suffer from excessive algorithm resource consumption, significant noise interference, low optimization accuracy, and safety hazards.
The excitation optimization algorithm is deployed on an external independent computing unit. It distinguishes between load and noise through timing analysis, adaptively switches the optimization mode, and configures disturbance blocking logic and fault tolerance mechanism to achieve efficient excitation control.
It improves the energy efficiency and dynamic response performance of servo motors under variable load conditions, ensures system safety and stability, and reduces the impact of control delay and noise interference.
Smart Images

Figure CN123122401A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of motor control technology, and more specifically, to a servo motor adaptive energy-saving control method. Background Technology
[0002] Servo motors are widely used in high-end equipment such as metal processing, CNC machine tools, and robot joints. During their operation, the load typically exhibits a slow, gradual change. Under such conditions, the optimal excitation current operating point of the servo motor will drift nonlinearly with changes in load. If the excitation current setpoint continuously deviates from this optimal operating point, it will directly lead to increased copper and iron losses in the motor, and a significant decrease in the overall system efficiency.
[0003] To address the excitation energy-saving control requirements under the aforementioned variable load conditions, existing technical solutions mainly suffer from the following technical bottlenecks:
[0004] The existing solution embeds the excitation optimization algorithm module directly into the firmware of the servo drive, which is prone to problems: the computing resources of the drive's main control chip are severely squeezed, the current loop control cycle with the highest real-time requirements is forced to be extended, and the torque dynamic response performance is limited to further improvement; the control logic and optimization logic are highly coupled, and the algorithm upgrade of any module requires a comprehensive regression test of the entire firmware of the drive.
[0005] Although existing systems are equipped with external torque sensors to collect load information, analog signals are prone to electromagnetic interference during long-distance transmission, introducing high-frequency noise components. The industry standard solution is to use low-pass delay filters to smooth the acquired signals and suppress noise. However, the inherent group delay characteristic of low-pass filters inevitably introduces signal phase lag. This lag directly degrades the real-time torque response of the servo system under dynamic operating conditions, causing a timing deviation between the load information used by the optimization algorithm and the actual load state, thus reducing optimization accuracy.
[0006] Existing optimization techniques that incorporate excitation disturbance testing generally lack online identification and classification of load change rates, continuing to apply periodic disturbances even under non-gradual load change conditions such as sudden load jumps. This disturbance component directly superimposes onto the motor's excitation windings, causing instantaneous fluctuations in torque output and severely compromising the system's dynamic torque response and operational stability under sudden load changes. The lack of a periodic online reference calibration mechanism leads to accumulated load measurement reference offsets, directly impacting the input accuracy and convergence reliability of the optimization algorithm. Furthermore, there is a general lack of active interlocking logic for excitation disturbances during abnormal conditions such as emergency stops, overloads, and overspeeds. When these abnormal conditions occur, the disturbance may continue to be output, posing a safety hazard. Summary of the Invention
[0007] To overcome the above shortcomings, this invention provides an adaptive energy-saving control method for servo motors, aiming to improve the problem of excessively high energy consumption caused by the lack of adaptive regulation in servo motors.
[0008] This invention is implemented as follows:
[0009] This invention provides an adaptive energy-saving control method for servo motors, applicable to variable load conditions of servo motors. The method is executed by a torque sensor, an external independent processing unit, and a servo driver.
[0010] The torque sensor acquires analog load torque signals and transmits them to the external independent processing unit. All core algorithms are handled by this external independent processing unit, which outputs analog control signals to the servo driver. The servo driver performs internal secondary calculations on the analog control signals and then outputs drive commands. The method includes:
[0011] S1. The external independent processing unit performs time-series analysis processing on the load torque analog signal, extracts the continuously smooth changing component as the real load component, identifies the high-frequency jumping component as the transmission noise component, performs compensation and cancellation on the noise component, and outputs the compensated and corrected load torque signal.
[0012] S2. Calculate the load change rate based on the compensated and corrected load torque signal; when it is determined to be a slow and gradual change condition, activate the small-amplitude trial disturbance optimization mode of the excitation current; when it is determined to be a rapid change in load, forcibly close the optimization mode, and the external computing unit directly outputs the excitation setpoint to maintain the current torque output.
[0013] S3. In the optimization mode, an excitation test disturbance of preset amplitude is superimposed on the excitation given signal, and real-time energy consumption data of the servo motor is collected simultaneously; based on the changing trend of the real-time energy consumption data, the optimal excitation current operating point under the current load condition is iteratively searched.
[0014] S4. The optimal excitation current operating point obtained by searching is converted into an analog control quantity that matches the input specifications of the servo driver and output to the servo driver; after receiving the analog control quantity, the servo driver converts it into a drive command to drive the servo motor to run.
[0015] By adopting the above technical solution, the independent method deploys all core excitation optimization algorithms on an external independent computing unit. The servo driver is only responsible for analog signal reception and internal secondary calculations, forming a three-layer physically decoupled architecture of data acquisition, algorithm decision-making, and drive execution. This architecture frees up the driver's underlying computing power, allowing it to focus on ensuring the real-time closed-loop control accuracy of current loop, speed loop, etc. At the same time, it extracts the real load components through time-series analysis and compensates for noise, and adaptively switches the optimization mode according to the load change rate, thereby improving the overall energy efficiency optimization effect and dynamic response performance under variable load conditions.
[0016] Preferably, the specific process of iteratively searching for the optimal excitation current operating point in S3 is as follows:
[0017] Positive and negative test disturbances are superimposed on the current excitation signal, and the real-time energy consumption values under the two disturbance states are collected respectively.
[0018] By comparing the two sets of energy consumption values, the excitation setpoint is iteratively updated in the direction of reducing energy consumption.
[0019] The process is repeated until the energy consumption difference between two adjacent iterations is less than the preset energy consumption convergence threshold. Once the optimization convergence is determined, the current optimal excitation current operating point is locked, the excitation test disturbance output is paused, and the current optimal excitation current is maintained for continuous output.
[0020] The external independent computing unit records the adjustment step size of the excitation setpoint in each iteration and dynamically adjusts the step size according to the changing trend of the energy consumption difference between the two iterations.
[0021] When the trend of energy consumption difference slows down, and the energy consumption difference between two adjacent iterations satisfies ΔP(k) < ΔP(k-1), it is determined that the trend of change has slowed down, and the step size is multiplied by the preset decay coefficient 0.8.
[0022] The energy consumption convergence threshold is set to 0.5%-2% of the rated operating power of the servo motor, and is configured online by the external independent computing unit or preset by the host computer.
[0023] By adopting the above technical solution, the specific process of iteratively searching for the optimal excitation current operating point is further refined. This includes superimposing positive and negative trial disturbances and comparing energy consumption values, updating the excitation setpoint in the direction of reducing energy consumption, dynamically adjusting the step size according to the trend of energy consumption difference, and locking the operating point and pausing the disturbance after the difference is less than the convergence threshold of 0.5%-2% of the rated power. This can achieve fast and high-precision optimization convergence while avoiding additional losses caused by continuous disturbance.
[0024] Preferably, the amplitude of the excitation test disturbance superimposed in S3 does not exceed 10% of the excitation rated value; when the change amplitude of the compensated and corrected load torque signal exceeds the pre-configured load change trigger threshold, the external independent computing unit reactivates the excitation current small amplitude test disturbance optimization mode and performs the optimal excitation current operating point search again.
[0025] By adopting the above technical solution, the amplitude of the excitation test disturbance is limited to no more than 10% of the excitation rating, and the optimization mode is reactivated when the change in load torque signal exceeds the preset trigger threshold. This ensures that the impact of the test disturbance on system operation is within a safe and controllable range, while also responding promptly to significant load changes, ensuring that the optimal operating point can be re-tracked and locked after drifting with operating conditions.
[0026] Preferably, the load change rate determination in S2 specifically includes:
[0027] The external independent processing unit calculates the signal change of the compensated and corrected load torque signal within a unit sampling time interval, and compares the signal change with a pre-stored rate judgment threshold. The rate judgment threshold is generated by multiplying the steady-state fluctuation amplitude of the load torque signal collected by the external independent processing unit during the no-load start-up phase of the servo motor by a preset safety margin coefficient, or by configuring it online by the host computer.
[0028] If the signal change is less than or equal to the rate determination threshold, it is determined to be a slow and gradual change condition.
[0029] If the signal change exceeds the rate determination threshold, it is determined to be a rapid load change, and the external independent processing unit stops applying excitation test disturbances.
[0030] By adopting the above technical solution, a specific method for determining the load change rate is provided. This involves calculating the change in load torque signal within a unit sampling interval and comparing it with a rate determination threshold generated based on the no-load steady-state fluctuation amplitude multiplied by a safety margin coefficient. This distinguishes between slow, gradual changes and rapid, abrupt changes in operating conditions. This determination mechanism adaptively generates the threshold based on actual load characteristics, ensuring accuracy and reliability, and providing a quantitative basis for operating condition switching decisions. Under rapid, abrupt changes, the current optimization result can still serve as a valid approximation of the operating condition after the abrupt change. Optimization is reactivated after the abrupt change ends, thus preventing energy efficiency loss.
[0031] Preferably, the timing analysis processing in S1 employs a non-low-pass delay filtering algorithm, specifically including:
[0032] Construct a time-domain sliding window for the original analog signal of the load torque, calculate the slope variance of the signal within the window, and ensure that the sampling frequency is not less than 10 times the highest effective frequency of the torque sensor output signal.
[0033] Signal segments with slope variance below a preset smoothing threshold are marked as components of real load with slow changes, while signal segments with slope variance above a preset smoothing threshold are marked as components of analog transmission noise.
[0034] The zero-point compensation parameter is invoked to perform reverse cancellation on the analog transmission noise component. The zero-point compensation parameter is generated by the external independent computing unit periodically collecting the original analog signal of the torque sensor under no-load conditions. It is used to eliminate the measurement deviation introduced by the temperature drift of the torque sensor itself and reduce the impact of this factor.
[0035] By adopting the above technical solution, a non-low-pass delay filtering algorithm is used for time-series analysis processing. A time-domain sliding window is constructed, and the signal slope variance is calculated. The low-variance segment is marked as the true load component, and the high-variance segment is marked as the noise component. Periodically generated zero-point compensation parameters are then used for inverse cancellation. This method filters out transmission noise while completely preserving the original phase information, avoiding the phase lag problem of traditional low-pass filtering, and providing a high-fidelity, low-latency input signal for the optimization algorithm.
[0036] Preferably, the method is further configured with disturbance blocking logic:
[0037] When it is detected that the servo motor is in an emergency stop state, the speed fluctuation exceeds the speed protection threshold, or the output torque exceeds the torque protection threshold, regardless of the load change rate determination result of S2, the small amplitude test disturbance optimization mode of the excitation current is directly locked, and the superimposed output of the excitation test disturbance is stopped.
[0038] The speed protection threshold and torque protection threshold are set to 50%-150% of the rated speed and rated torque of the servo motor, respectively, and are pre-written by the external independent computing unit through the host computer configuration interface.
[0039] By adopting the above technical solution and configuring disturbance interlocking logic, when abnormal operating conditions such as emergency stop, excessive speed, or excessive torque are detected, the optimization mode is forcibly interlocked and the probing disturbance output is stopped, regardless of the load change rate determination result. This can cut off the secondary risks that disturbances may cause under dangerous operating conditions from the source, thereby improving the operational safety of the system.
[0040] Preferably, before the S4 outputs the analog control quantity, a limiting process is performed: the external independent processing unit retrieves the upper and lower limits of the servo motor's excitation current and constrains the analog control quantity corresponding to the calculated optimal excitation current within the analog input range that the servo driver can recognize.
[0041] By adopting the above technical solution, amplitude limiting processing is performed before the output analog control quantity. This involves calling the upper and lower limits of the servo motor's excitation current to constrain the analog control quantity within the input range recognizable by the servo driver. This process prevents out-of-range signals from causing driver malfunctions or output loss of control, ensuring the physical compatibility of the signal interface and the stability of system operation.
[0042] Preferably, the internal secondary calculations performed by the servo driver include at least signal range conversion calculations and excitation closed-loop feedforward compensation calculations:
[0043] The servo driver converts the received analog control quantity into an internal digital domain excitation reference value. Combined with the real-time speed and torque of the servo motor, it queries the internally stored excitation feedforward compensation table to calculate the feedforward compensation amount and generate the final excitation drive command output to the power drive circuit of the servo motor.
[0044] By adopting the above technical solution, it is clear that the internal secondary calculations performed by the servo driver include at least signal range conversion calculations and excitation closed-loop feedforward compensation calculations. That is, the driver converts the analog control quantity into an internal digital domain excitation reference value, and then performs feedforward compensation based on real-time speed and torque to generate the final drive command. This solution ensures that the analog quantity output by the external unit can be correctly parsed by the driver and converted into an excitation output adapted to the actual operating state of the motor, guaranteeing the accuracy of execution at the end of the control link.
[0045] Preferably, the method also includes fault-tolerant handling logic:
[0046] The external independent computing unit caches the load torque signal, the excitation test disturbance application status, the optimal excitation current operating point, and the real-time energy consumption of the servo motor in real time, and records the data timestamps simultaneously.
[0047] When the external independent processing unit detects an abnormal torque sensor signal or a fault in the analog signal transmission, it prioritizes retrieving historical data with the most recent timestamp and at which the excitation test disturbance was in a convergent steady state to generate an analog control quantity for output.
[0048] If the timestamp of the cached data exceeds the preset valid time limit, the pre-stored default excitation setpoint will be retrieved and output.
[0049] By adopting the above technical solution and configuring fault-tolerant processing logic, historical data such as load torque, disturbance state, optimal operating point and energy consumption are cached in real time and timestamps are recorded. When the sensor signal is abnormal or the transmission is interrupted, the most recently converged steady-state data is called first to generate the transition control quantity. If the cache times out, the default excitation setpoint is called. This can achieve a seamless and smooth transition under fault conditions, ensuring the reliability and safety of continuous production in industrial sites.
[0050] Preferably, the specific method for collecting real-time energy consumption data of the servo motor in S3 is as follows:
[0051] The external independent processing unit reads the DC bus voltage and DC bus current values of the servo driver in real time via a high-speed sampling interface.
[0052] Instantaneous power data is obtained by synchronously multiplying the DC bus voltage value and the DC bus current value.
[0053] A moving average filter is applied to the instantaneous power data to remove high-frequency switching noise from the power circuit. The processed average power value is then used as real-time energy consumption data for iterative optimization.
[0054] By adopting the above technical solution, the specific method for acquiring real-time energy consumption data is defined. Specifically, the DC bus voltage and current are synchronously read through a high-speed sampling interface and multiplied. Then, the average power value is obtained after filtering out high-frequency switching noise using a moving average filter, serving as the energy consumption feedback. This method ensures the synchronization and smoothness of energy consumption data, providing accurate and stable feedback signals for iterative optimization, thereby improving the accuracy of optimization decisions.
[0055] The beneficial effects of this invention are:
[0056] 1. In this invention, all the core algorithms for adaptive optimization of excitation current are deployed in an external computing unit independent of the servo driver. The servo driver is only responsible for receiving analog control quantities and completing internal secondary calculations.
[0057] The three-layer separation architecture of data acquisition, algorithm decision-making, and drive execution achieves physical layer decoupling of control computing power and optimization computing power. The underlying computing resources of the servo drive are freed from heavy optimization tasks, allowing it to focus more on ensuring the execution accuracy and response speed of real-time closed-loop control such as current loop and speed loop.
[0058] 2. This invention employs a time-series analysis combined with noise compensation and cancellation mechanism based on signal slope variance analysis within a time-domain sliding window. By statistically calculating the signal slope variance within the window, it distinguishes between the slowly changing components of the real load and the simulated transmission noise components, and performs reverse cancellation on the noise components based on periodically self-generated zero-point compensation parameters;
[0059] While effectively suppressing electromagnetic noise interference in the analog transmission link, it retains the original phase information of the load torque signal, avoids the phase lag problem introduced by the low-pass filter, and provides a high-fidelity, low-latency original signal input basis for the excitation optimization algorithm.
[0060] 3. This invention configures an online load change rate discrimination mechanism to achieve adaptive intelligent switching of operating conditions, activating optimization under slow and gradual load changes and blocking disturbances under rapid and sudden load changes. Only when the load is determined to be in a slow and gradual load change condition, positive and negative probing disturbances of a limited amplitude (2%-5% of the rated excitation value) are superimposed on the excitation given signal, and the optimal excitation current operating point after drift is iteratively searched based on real-time energy consumption feedback.
[0061] Once a rapid change in load is detected, the probe disturbance output is immediately forcibly shut down to avoid the optimization disturbance from degrading the dynamic torque response.
[0062] 4. This invention integrates multiple safety protection mechanisms such as disturbance interlocking logic, analog output hardware limiting, and fault-tolerant transition. It is configured with abnormal operating condition disturbance interlocking logic. When an emergency stop, speed limit overrun, or torque limit overrun is detected, the excitation test disturbance output is directly blocked, reducing the secondary safety risks caused by control disturbances.
[0063] Before the analog control output is executed, hardware receiving range limiting processing is performed to prevent out-of-range signals from causing abnormal operation of the servo drive or loss of output control. Fault tolerance processing logic is configured so that when the torque sensor signal is abnormal or the transmission link is disconnected, the historical operating data under the most recent normal operating conditions is automatically retrieved to generate transition control quantities, so as to achieve a seamless and smooth transition under fault conditions, thereby improving the stability and safety of continuous production in industrial sites. Attached Figure Description
[0064] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.
[0065] Figure 1 This is a schematic diagram of the system structure of a servo motor adaptive energy-saving control method provided by an embodiment of the present invention;
[0066] Figure 2 This is a schematic diagram of the main flow of a servo motor adaptive energy-saving control method provided by an embodiment of the present invention;
[0067] Figure 3 This is a schematic diagram of the internal algorithm logic of an external independent computing unit for a servo motor adaptive energy-saving control method provided in an embodiment of the present invention;
[0068] Figure 4 This is a schematic diagram of the working condition switching and safety protection logic of a servo motor adaptive energy-saving control method provided by an embodiment of the present invention.
[0069] In the diagram: 1. Torque sensor; 2. External independent processing unit; 3. Servo driver. Detailed Implementation
[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0071] Example
[0072] Reference Figures 1-4 An adaptive energy-saving control method for servo motors is proposed, applied to servo motors under varying load conditions. This method is executed by a torque sensor 1, an external independent processing unit 2, and a servo driver 3, wherein:
[0073] Torque sensor 1 is a strain gauge torque sensor, which outputs an analog voltage signal of 0-10V. This signal is transmitted to the analog input channel of the external independent processing unit 2 through a twisted pair shielded cable.
[0074] Reference Figure 3 The external independent processing unit 2 is a microcontroller board based on the ARM Cortex-M7 core, which includes multiple 12-bit analog-to-digital converter channels, digital-to-analog converter output channels, and a CAN communication interface. The ARM Cortex-M7 core microcontroller has a built-in double-precision floating-point unit and SIMD instruction set, which can complete the parallel execution of multiple instructions in a single cycle. The floating-point multiplication and division operations in the optimization algorithm (such as instantaneous power calculation, step size dynamic adjustment coefficient multiplication, etc.) are accelerated by the FPU hardware, which improves the efficiency by about 10-15 times compared with software simulation of floating-point operations.
[0075] Servo driver 3 is a general-purpose servo driver, whose analog input port receives 0-10V control signals.
[0076] It should be noted that the real-time control architecture of a servo drive typically contains three nested control loops:
[0077] Current loop (inner loop, response time 50-100μs);
[0078] Speed loop (middle loop, response time 1-5ms);
[0079] Location loop (outer loop, response time 10-50ms).
[0080] The real-time requirements for the current loop are the most stringent, and it needs to be completed within each control cycle:
[0081] The three-phase current is sampled to Clarke transform (transforming the three-phase stationary coordinate system into a two-phase stationary coordinate system), then to Park transform (transforming the two-phase stationary coordinate system into a two-phase rotating coordinate system), then to the PI regulator for calculation, then to the inverse Park transform, then to SVPWM (space vector pulse width modulation), and finally modulated to the power switch drive signal output.
[0082] If the iterative optimization algorithm (which requires multiple sampling, comparisons, and step size adjustments in each iteration) is embedded into the driver's main control chip, it will directly encroach on the current loop's computation time window.
[0083] Torque sensor 1 collects the load torque analog signal and transmits it to external independent processing unit 2. All core algorithms are handled by external independent processing unit 2, which outputs analog control quantity to servo driver 3. Servo driver 3 performs internal secondary calculations on the analog control quantity and outputs drive command.
[0084] In some implementations, the internal secondary operations performed by the servo driver 3 include at least signal range conversion operations and excitation closed-loop feedforward compensation operations:
[0085] Servo driver 3 converts the received analog control signal into an internal digital domain excitation reference value. Combining this with the real-time speed and torque of the servo motor, it consults the internally stored excitation feedforward compensation table to calculate the feedforward compensation amount, generating the final excitation drive command output to the servo motor's power drive circuit. It should be noted that the analog control signal (e.g., 5V voltage) received by servo driver 3 is converted into a digital value through its internal analog-to-digital conversion. This digital value is then converted into an excitation current setpoint (e.g., 2A) according to the proportional coefficient configured within the driver. The driver then reads the speed value fed back from the current rotor position sensor and the torque and current values fed back from the current sensor. It consults the internally stored excitation feedforward compensation table to obtain the corresponding compensation amount, adds this compensation amount to the excitation current setpoint, and then outputs it to the inverter's PWM modulation unit via a PI regulator to control the duty cycle of the power switching transistors.
[0086] To verify the technical superiority of the three-layer discrete architecture (torque sensor 1, external independent processing unit 2, and servo driver 3) over the traditional solution (algorithm embedded in the driver), a comparative experiment was conducted:
[0087] Table 1 Comparison Results of Three-Layer Decoupled Architecture and Algorithm Embedded Driver Technology
[0088]
[0089] It can be known that:
[0090] The three-layer discrete architecture reduces the current loop control cycle from 150μs to 80μs. This shorter control cycle improves current tracking accuracy; the tracking error of the current loop is positively correlated with the control cycle, and a shorter cycle provides stronger suppression of high-frequency harmonics. The speed loop response bandwidth is increased from 180Hz to 280Hz, meaning the system's ability to follow speed command changes is improved by approximately 55%. Regarding algorithm iteration efficiency, the dedicated computing power of the external independent computing unit 2 reduces the optimization iteration cycle from 200ms / iteration to 50ms / iteration, increasing the convergence speed by four times. Simultaneously, it reduces the CPU utilization of the driver's main control chip from 78% to 42%, freeing up computing power to further improve the control accuracy of the current loop.
[0091] like Figure 1 As shown by the middle arrow, the torque sensor 1 is connected to the external independent processing unit 2 via an analog signal link, and the external independent processing unit 2 is connected to the servo driver 3 via an analog control signal link, forming a unidirectional data flow architecture.
[0092] Reference Figure 2 The methods include:
[0093] S1, the external independent processing unit 2 performs timing analysis processing on the load torque analog signal, extracts the continuously smooth changing component as the real load component, identifies the high-frequency abrupt component as the transmission noise component, performs compensation and cancellation on the noise component, and outputs the compensated and corrected load torque signal. The external independent processing unit 2 continuously acquires the analog output voltage signal of the torque sensor 1 at a sampling frequency of 1kHz. The acquired raw signal sequence first enters a circular buffer with a length of 50 sampling points. For each new data point acquired, the unit performs timing analysis processing on all data points in the buffer.
[0094] In some implementations, the timing analysis process employs a non-low-pass delay filtering algorithm, specifically including:
[0095] A time-domain sliding window is constructed for the original analog signal of the load torque. The mechanical time constant of the servo motor is 0.5 seconds, therefore the length of the time-domain sliding window is set to 0.1 seconds, or 100 sampling points (at a sampling frequency of 1kHz). The signal slope variance within the window is calculated. The length of the time-domain sliding window is set to 1 / 5 to 1 / 3 of the mechanical time constant of the servo motor, and the sampling frequency is not less than 10 times the highest effective frequency of the torque sensor 1 output signal. The highest effective frequency of the torque sensor 1 output signal is tested to be approximately 100Hz, and the current 1kHz sampling frequency meets the requirement of being at least 10 times higher.
[0096] The transfer function of a traditional low-pass filter can be expressed as:
[0097] H(s) = 1 / (1 + sτ),
[0098] Its phase frequency response is
[0099] ,
[0100] An approximately 45° phase lag is introduced at the cutoff frequency. The direct consequence of this phase lag in the servo system is a delayed torque response. When the load changes abruptly, the low-pass filtered torque signal arrives at the controller several milliseconds late, resulting in a timing deviation between the controller's excitation adjustment decision based on the lagging signal and the actual load condition.
[0101] Signal segments with slope variance below a preset smoothing threshold are marked as slowly changing components of the actual load, while those with slope variance above the preset smoothing threshold are marked as analog transmission noise components. The signal slope variance is calculated using the least squares fitting method. The preset smoothing threshold is set to 0.05 (dimensionless value), which is related to the torque sensor range and the ADC quantization bit depth. In this embodiment, the torque sensor range is 0~10V, the ADC has a 12-bit resolution (0~4095 digital values), and 0.05 is the normalized dimensionless threshold, corresponding to an actual voltage change rate of 0.5V / s. Users can adjust this threshold via a host computer according to the actual sensor specifications.
[0102] The principle of the least squares linear fitting algorithm used is: for the signal sequence within the window...
[0103] {x1,x2,...,x N}{x1,x2,...,x N},
[0104] Slope of the fitted straight line
[0105] ,
[0106] Then calculate the residual variance
[0107] .
[0108] The residual variance reflects the degree to which the signal deviates from the smooth trend line. The actual load variation component is continuous on the mechanical time constant scale, and its residual variance is small. High-frequency noise components such as electromagnetic interference have randomness and high-frequency characteristics, and their residual variance is large.
[0109] The zero-point compensation parameter is invoked to reversely cancel the analog transmission noise component. This parameter is generated by the external independent processing unit 2 periodically acquiring the original analog signal from the torque sensor 1 under no-load conditions. It is used to eliminate the measurement deviation introduced by the temperature drift of the torque sensor 1 itself, reducing the impact of this factor. The zero-point compensation parameter is automatically acquired once after each system power-on, before the servo motor is loaded, for a duration of 200 sampling points. The average value is taken as the current zero-point offset, and it is automatically updated every 4 hours of operation thereafter.
[0110] In some implementations, zero-point temperature drift is one of the main sources of error in strain gauge torque sensors, with a typical zero-point temperature drift coefficient of ±0.05%FS / ℃. Under operating conditions with an ambient temperature variation of 20℃, an uncompensated zero-point offset of approximately 1%FS will be introduced. Periodically (every 4 hours) automatically acquiring no-load state signals and updating the zero-point compensation parameters helps to reduce the measurement deviation introduced by temperature drift.
[0111] To verify the superiority of timing analysis, a comparative experiment was conducted with traditional low-pass filtering schemes:
[0112] Table 2 Comparison of timing analysis and low-pass filtering results
[0113]
[0114] It can be known that:
[0115] Traditional first-order low-pass filtering introduces approximately 32° phase hysteresis at 100Hz, with an equivalent delay of about 0.89ms. Timing analysis achieves zero phase hysteresis by trading a slightly lower noise suppression ratio (-35dB / dec vs -40dB / dec), reducing the response delay from load abrupt changes to excitation regulation from 1.2ms to 0.15ms, an 8-fold reduction. This is particularly relevant in servo system applications where real-time torque response is critical.
[0116] S2. Calculate the load change rate based on the compensated and corrected load torque signal; when it is determined to be a slow and gradual change condition, activate the excitation current small-amplitude trial disturbance optimization mode; when it is determined to be a rapid load change, forcibly close the optimization mode, and the external arithmetic processing unit 2 directly outputs the excitation setpoint to maintain the current torque output.
[0117] In some implementations, the data flow between S1 and S2 is as follows:
[0118] After the external independent processing unit 2 completes the timing analysis and noise compensation of the original analog signal of the load torque, it writes the compensated and corrected load torque signal into the data buffer of the internal random access memory at a fixed sampling period (e.g., sampling frequency of 1kHz), and stores the data of the most recent N sampling points in a first-in-first-out manner.
[0119] In some implementations, the determination of the load change rate specifically includes:
[0120] The external independent processing unit 2 calculates the signal change of the compensated and corrected load torque signal within a unit sampling time interval. This signal change is compared with a pre-stored rate judgment threshold. The rate judgment threshold is generated by multiplying the steady-state fluctuation amplitude of the load torque signal collected by the external independent processing unit 2 during the servo motor's no-load start-up phase by a preset safety margin coefficient, or it can be configured online by the host computer. The external independent processing unit 2 reads the compensated and corrected load torque signal value T(n) at the current moment and the value T(n-1) at the previous sampling moment from the aforementioned data buffer. It calculates the signal change ΔT within a unit sampling time interval Δt according to the following formula. When performing load change rate judgment, ΔT is compared with the rate judgment threshold ΔTthr pre-stored within the unit:
[0121] ΔT = T(n) - T(n-1),
[0122] If the signal change is less than or equal to the rate determination threshold, i.e., if |ΔT| < ΔT thr If the servo motor is in a slow, gradual change mode, then it is determined that the servo motor is currently in a slow, gradual change mode.
[0123] If the signal change is greater than the rate determination threshold, i.e., if |ΔT|>ΔT thr If the servo motor experiences a rapid load change, the external independent processing unit 2 will stop applying excitation test disturbances.
[0124] The above reading and calculation operations are performed synchronously in each sampling period to ensure that the judgment result of S2 and the signal output of S1 are strictly synchronized in time, and there is no pipeline delay or data frame loss.
[0125] In this embodiment, the rate determination threshold is automatically generated when the servo motor is started under no-load for the first time. Specifically, after the servo motor has been running stably under no-load for 5 seconds, the external independent processing unit 2 collects the root mean square value of the load torque signal during this period as the steady-state fluctuation amplitude, multiplies it by the safety margin factor 4 to obtain the rate determination threshold, and the host computer can also write new threshold parameters online via the CAN bus.
[0126] The amount of signal change within a unit sampling time interval (1ms, corresponding to a 1kHz sampling frequency).
[0127] ΔT thr =K margin ×T ripple ,
[0128] Where T rippleK is the root mean square value of the load torque signal during the no-load stable operation period. margin This is the safety margin factor (taken as 4).
[0129] Since the rated load fluctuation range of a servo motor is usually ±20% of the rated torque, taking a margin factor of 4 can make the judgment threshold adapt to the load fluctuation characteristics of different specifications of servo motors, and avoid false triggering due to sensor noise.
[0130] The significance lies in the fact that the steady-state fluctuation amplitude of torque signals varies under different servo systems and installation conditions. A fixed threshold may lead to frequent misjudgments of rapid abrupt changes (i.e., false alarms) in systems with large fluctuations, or insensitivity to real abrupt changes (i.e., missed detections) in systems with small fluctuations. An adaptive threshold matches the judgment benchmark with the noise limit of the actual system, helping to improve the accuracy of the judgment. As an industrial fieldbus standard, the CAN bus has differential signal transmission and collision detection arbitration mechanisms, which can ensure the reliability of data transmission in environments with strong electromagnetic interference.
[0131] To verify the technical advantages of the adaptive threshold scheme, a comparative experiment was conducted with the fixed threshold scheme.
[0132] Table 3 Comparison results of adaptive threshold and fixed threshold
[0133]
[0134] It can be known that:
[0135] Fixed threshold schemes face a dilemma in practical applications: if the threshold is set too high, it will miss real load mutations (missed detection rate of about 8%), resulting in failure to lock out disturbances that should be locked out; if the threshold is set too low, it will frequently generate false alarms (false alarm rate of about 15%), causing unnecessary and frequent interruptions to the optimization mode. Adaptive threshold schemes, through self-learning under no-load conditions, reduce the false alarm rate from 15% to below 3% and the missed detection rate from 8% to below 2%, helping to improve the accuracy and reliability of operating condition identification.
[0136] S3. In optimization mode, a preset amplitude excitation test disturbance is superimposed on the excitation given signal, and real-time energy consumption data of the servo motor is collected simultaneously. Based on the changing trend of the real-time energy consumption data, the optimal excitation current operating point under the current load condition is iteratively searched.
[0137] In some implementations, the specific process of iteratively searching for the optimal excitation current operating point is as follows:
[0138] Positive and negative test disturbances are superimposed on the current excitation signal, and real-time energy consumption values under the two disturbance states are collected respectively.
[0139] By comparing the two sets of energy consumption values, the excitation setpoint is iteratively updated in the direction of reducing energy consumption.
[0140] The process is repeated until the energy consumption difference between two adjacent iterations is less than the preset energy consumption convergence threshold. Once the optimization convergence is determined, the current optimal excitation current operating point is locked, the excitation test disturbance output is paused, and the current optimal excitation current is maintained for continuous output.
[0141] The external independent processing unit 2 records the adjustment step size of the excitation setpoint in each iteration and dynamically adjusts the step size according to the changing trend of the energy consumption difference between the two iterations.
[0142] When the trend of energy consumption difference slows down, and the energy consumption difference between two adjacent iterations satisfies ΔP(k) < ΔP(k-1), it is determined that the trend of change has slowed down, and the step size is multiplied by the preset decay coefficient 0.8.
[0143] The energy consumption convergence threshold is set to 0.5%-2% of the servo motor's rated operating power, and is configured online by the external independent processing unit 2 or preset via the host computer. The external independent processing unit 2 is connected to an external high-speed analog-to-digital converter via its SPI interface. SPI is a high-speed synchronous serial communication protocol with a maximum communication rate of tens of MHz. The converter synchronously samples the voltage and current signals of the DC bus of the servo driver 3 at a sampling rate of 20kHz. According to the Nyquist sampling theorem, the sampling frequency must be no less than twice the highest frequency of the signal.
[0144] 0.5% of the rated power corresponds to high-precision applications (such as precision positioning), while 2% corresponds to general speed regulation applications. Users can configure this online via a host computer according to their actual application scenarios; the default value is set to 1% of the rated power.
[0145] The switching frequency of the IGBT power switch in servo driver 3 is typically in the range of 4-16kHz. The high-frequency switching noise it generates is mainly concentrated at the switching frequency and its harmonics, which helps to filter out high-frequency switching noise components. After each sampling, the unit immediately performs a multiplication operation to obtain the instantaneous power, and then sends the instantaneous power value into a moving average filter with a length of 200 sampling points. The filtered average power value is output as the energy consumption data.
[0146] The specific process for each iteration is as follows:
[0147] Let the current excitation setpoint be I0, and record the real-time energy consumption P0 at this time;
[0148] Then, I0+ΔI and I0-ΔI are output sequentially, where the initial value of ΔI is 3% of the excitation rating. After each trial disturbance is output and stabilized for 20ms, the corresponding energy consumption P+ and P- are recorded respectively.
[0149] Compare P+ and P-:
[0150] If P+<P-, the new excitation setpoint I1=I0+step;
[0151] Otherwise, I1=I0-step, where the initial value of step is 1% of the rated excitation value.
[0152] In subsequent iterations, if |P(k)-P(k-1)|<|P(k-1)-P(k-2)|, multiply step by 0.8, that is
[0153] step k +1=0.8×step k .
[0154] This mechanism enables the algorithm to quickly approach the extreme point with a large step size when far away from it (coarse search phase), and perform fine search with a small step size when close to the extreme point (fine search phase), taking into account both convergence speed and convergence accuracy; otherwise, the step size remains unchanged.
[0155] The energy consumption convergence threshold is set to 1% of the rated power of the servo motor. When the difference in energy consumption between two adjacent iterations is less than this threshold, the optimization search is determined to be convergent, the current excitation current value is locked, and the trial perturbation output is stopped.
[0156] To verify the advantages of this mechanism, the following comparative test is carried out:
[0157] Table 4 Comparison of results between fixed step size and variable step size
[0158]
[0159] It can be seen that:
[0160] The fixed step size scheme faces a dilemma: a large step size (such as 2% rated excitation) converges fast (about 10-14 times) but has poor accuracy (about ±1.2%), while a small step size (such as 0.5% rated excitation) has high accuracy (about ±0.3%) but converges slowly (about 30-40 times).
[0161] Through the dynamic step size attenuation mechanism ( attenuates to 0.8 times at the corresponding condition), convergence can be achieved within 8-12 iterations, and the convergence time is shortened from about 1 second to about 0.5 seconds. At the same time, the convergence accuracy is improved from ±0.8% to ±0.2% of the rated excitation. In a system with a rated excitation current of 10A, this means that the optimization search error is reduced from ±0.08A to ±0.02A, and the corresponding copper loss (Pcu=I²R) is optimized more fully. 20kHz high-speed synchronous sampling reduces the synchronization error of energy consumption feedback from ±2% to ±0.3%, providing a more accurate decision-making basis for iterative optimization.
[0162] In another embodiment, the amplitude of the superimposed excitation test disturbance does not exceed 10% of the excitation rated value; the load change trigger threshold is set to 1.8 times the rate determination threshold. When the cumulative change of the compensated and corrected load torque signal exceeds this trigger threshold within 10 consecutive sampling periods, the external independent processing unit 2 reactivates the optimization mode, and starts a new round of optimization even if it is not currently in the optimization state. When the change amplitude of the compensated and corrected load torque signal exceeds the pre-configured load change trigger threshold, the external independent processing unit 2 reactivates the small-amplitude excitation current test disturbance optimization mode and performs the optimal excitation current operating point search again.
[0163] In yet another embodiment, the specific method for collecting real-time energy consumption data of the servo motor is as follows:
[0164] The external independent processing unit 2 reads the DC bus voltage and DC bus current values of the servo driver 3 in real time via a high-speed sampling interface.
[0165] Instantaneous power data is obtained by synchronously multiplying the DC bus voltage value and the DC bus current value.
[0166] A moving average filter is applied to the instantaneous power data to remove high-frequency switching noise from the power circuit. The processed average power value is then used as real-time energy consumption data for iterative optimization.
[0167] S4. The optimal excitation current operating point obtained by searching is converted into an analog control quantity that matches the input specifications of the servo driver 3 and output to the servo driver 3. After receiving the analog control quantity, the servo driver 3 converts it into a drive command to drive the servo motor.
[0168] In some implementations, a limiting process is performed before outputting the analog control quantity: the external independent processing unit 2 retrieves the upper and lower limits of the servo motor excitation current and constrains the analog control quantity corresponding to the optimal excitation current to within the analog input range that the servo driver 3 can recognize.
[0169] The servo driver 3 can recognize an analog input range of 0-10V. The external independent processing unit 2 has an internally stored excitation current with an upper and lower limit of 2A and 8A respectively, with a conversion factor of 1A corresponding to 1V.
[0170] When the optimal excitation current obtained through optimization is 5.5A, the converted analog control quantity is 5.5V. This value is within the 0-10V range and is directly output. If the converted value exceeds this range, it will be output at the upper limit of 10V or the lower limit of 0V.
[0171] In another implementation, the method also configures fault-tolerant processing logic:
[0172] The independent computing and processing unit 2 caches the load torque signal, the excitation test disturbance application status, the optimal excitation current operating point, and the real-time energy consumption of the servo motor in real time, and records the data timestamps synchronously.
[0173] When the external independent processing unit 2 detects an abnormal signal from the torque sensor 1 or a fault in the analog signal transmission, it prioritizes retrieving historical data with the most recent timestamp and at which time the excitation test disturbance was in a convergent steady state to generate an analog control quantity for output.
[0174] In this embodiment, a circular buffer is allocated in the internal RAM of the external independent processing unit 2 to store the operating history data for the most recent hour. Each data entry includes a timestamp, load torque value, disturbance application state, optimal excitation current value, and real-time energy consumption value. When the output signal of the torque sensor 1 remains below 0.5V or above 9.5V for more than 100ms, it is determined to be a signal abnormality. At this time, the unit searches for the data entry with the most recent timestamp and a disturbance application state that is in a convergent steady state from the circular buffer, and converts its optimal excitation current value into an analog control output.
[0175] If the timestamp of the latest converged steady-state data in the cache is more than 10 minutes earlier than the current time, the cache data is considered to have timed out, and the default excitation setpoint stored in the EEPROM is output instead (this default value is written during the initial debugging of the device and is 50% of the rated excitation current of the servo motor. The default excitation setpoint is written to the EEPROM during the initial debugging of the device through the minimum energy consumption calibration test under no-load rated speed. The specific method is: gradually adjust the excitation current under no-load conditions, record the input power under each excitation current, and take the excitation current value corresponding to the lowest power point as the default excitation setpoint).
[0176] To further explain, EEPROM is a non-volatile memory that does not lose data after the system loses power. It is suitable for storing critical parameters such as default excitation setpoints written during the initial debugging of the device. The circular buffer adopts a first-in, first-out (FIFO) strategy to store the operating history data of the most recent hour, and each data entry includes a timestamp to identify the timeliness of the data.
[0177] If the timestamp of the cached data exceeds the preset valid time limit, the pre-stored default excitation setpoint will be retrieved and output.
[0178] Furthermore, the method is also equipped with disturbance interlocking logic, which has a higher priority than the operating condition determination in S2, and is a hard protection mechanism prioritizing safety. The emergency stop signal is independently accessed through a digital input interface and is physically isolated from the torque analog signal channel. Even if the analog signal transmission link completely fails, the emergency stop signal can still be reliably identified.
[0179] To verify the advantages of the above three-layer protection measures, a comparison was made with traditional solutions (no / few layers of protection):
[0180] Table 5 Comparison of results between three-layer protection measures and no / few-layer protection measures:
[0181]
[0182] It can be known that:
[0183] Traditional solutions often directly trigger driver failure and shutdown when a sensor disconnects, causing production line interruptions with recovery times typically exceeding 5 minutes. By using a circular buffer (storing 1 hour of historical data) and fault-tolerant transition logic, when a sensor signal is abnormal (signal continuously below 0.5V or above 9.5V for more than 100ms), the system can switch to the most recent "converged steady-state historical data" for continued output within 2ms. The disturbance lockout logic cuts off the probe disturbance output within 1ms under emergency stop or overload conditions (speed > 120% of rated speed or torque > 130% of rated torque), reducing the probability of secondary failures caused by abnormal operating conditions from 12% to below 1%.
[0184] When any of the following conditions are detected: the servo motor is in an emergency stop state, the speed fluctuation exceeds the speed protection threshold, or the output torque exceeds the torque protection threshold, regardless of the load change rate determination result of S2, the small amplitude trial disturbance optimization mode of the excitation current is directly blocked, and the superimposed output of the excitation trial disturbance is stopped.
[0185] The speed protection threshold and torque protection threshold are set to 50%-150% of the servo motor's rated speed and rated torque, respectively, and are pre-written by the external independent processing unit 2 through the host computer configuration interface. The speed protection threshold is set to 120% of the rated speed, and the torque protection threshold is set to 130% of the rated torque. The external independent processing unit 2 receives these threshold parameters from the host computer via the CAN communication interface and stores them in its internal EEPROM. When the unit detects that the speed signal fed back by the servo driver 3 exceeds the above thresholds, or the received torque signal exceeds the above thresholds, it immediately sets the disturbance lockout flag and stops any probing disturbance output. The emergency stop status signal is connected to the external independent processing unit 2 through a digital input interface independent of the torque signal.
[0186] In some implementations, after receiving the 0-10V analog control signal, the servo driver 3 converts it into a digital domain excitation reference value via an internal analog-to-digital converter, which typically uses 12-bit resolution, corresponding to a 0-4095 digital value. The core of feedforward compensation lies in prediction rather than feedback. Feedback control only starts adjusting after detecting a deviation, resulting in inherent lag; feedforward compensation, on the other hand, calculates the required excitation compensation amount in advance based on the real-time values of speed and torque current and adds it to the given value.
[0187] The purpose of compensation is to counteract the effect of armature reaction on the excitation magnetic field. When the armature current increases, the armature reaction will produce a demagnetizing or magnetizing effect.
[0188] In permanent magnet synchronous motors, the demagnetizing effect of armature reaction may cause irreversible demagnetization of permanent magnets;
[0189] In induction motors, this will affect the stability of the air gap magnetic flux.
[0190] The feedforward compensation table is typically a two-dimensional lookup table, with the horizontal axis representing speed and the vertical axis representing torque and current. The table stores the optimal compensation coefficients obtained through calibration tests. This two-dimensional lookup table is obtained through bench calibration tests before the motor leaves the factory. Specifically, the optimal excitation current value is measured under different combinations of servo motor speeds (from 0 to rated speed, in 200 rpm steps) and torque loads (from 0 to rated torque, in 10% of rated torque steps), and then entered into the two-dimensional table. The calibration data is stored in the servo driver's internal non-volatile memory. The compensated excitation reference value is then output to the inverter's PWM modulation unit via a PI regulator to control the duty cycle of the power switching transistors.
[0191] To verify the advantages of feedforward compensation, a comparative experiment was conducted:
[0192] Table 6 Comparison results with and without feedforward compensation
[0193]
[0194] It can be known that:
[0195] Feedforward compensation reduces the excitation current overshoot from ±12% to ±3% during load abrupt changes, and the steady-state error from ±2.5% to ±0.5%. Torque response delay is reduced from 12ms to 5ms, and torque output capability in the field-weakening operating region (high-speed operating region, where the excitation current needs to be reduced to weaken the magnetic field and increase speed) is increased from 85% to 95% of rated torque. The excitation current tracking bandwidth is increased from 150Hz to 350Hz. Higher bandwidth means stronger ability of the excitation current to follow command changes, resulting in better control accuracy under high-frequency dynamic conditions.
[0196] It should be noted that the specific model and specifications of the servo motor need to be selected and determined based on the actual specifications of the device. The specific selection and calculation method adopts existing technology in this field, so it will not be described in detail here.
[0197] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.
Claims
1. A servo motor adaptive energy-saving control method, characterized in that, This method is applied to the variable load conditions of servo motors and is executed by a torque sensor (1), an external independent processing unit (2), and a servo driver (3). The torque sensor (1) collects the load torque analog signal and transmits it to the external independent processing unit (2). All core algorithms are handled by the external independent processing unit (2), which outputs analog control quantities to the servo driver (3). The servo driver (3) performs internal secondary calculations on the analog control quantities and then outputs drive commands. The method includes: S1. The external independent processing unit (2) performs time-series analysis processing on the load torque analog signal, extracts the continuous smooth change component as the real load component, identifies the high-frequency jump component as the transmission noise component, performs compensation and cancellation on the noise component, and outputs the compensated and corrected load torque signal. S2. Calculate the load change rate based on the compensated and corrected load torque signal; when it is determined to be a slow and gradual change condition, activate the excitation current small-amplitude trial disturbance optimization mode; when it is determined to be a rapid load change, forcibly close the optimization mode, and the external computing unit (2) directly outputs the excitation setpoint to maintain the current torque output. S3. In the optimization mode, an excitation test disturbance of preset amplitude is superimposed on the excitation given signal, and real-time energy consumption data of the servo motor is collected simultaneously; based on the changing trend of the real-time energy consumption data, the optimal excitation current operating point under the current load condition is iteratively searched. S4. The optimal excitation current operating point obtained by searching is converted into an analog control quantity that matches the input specification of the servo driver (3) and output to the servo driver (3); after receiving the analog control quantity, the servo driver (3) converts it into a drive command to drive the servo motor to run.
2. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The specific process of iteratively searching for the optimal excitation current operating point in S3 is as follows: Positive and negative test disturbances are superimposed on the current excitation signal, and the real-time energy consumption values under the two disturbance states are collected respectively. By comparing the two sets of energy consumption values, the excitation setpoint is iteratively updated in the direction of reducing energy consumption. The process is repeated until the energy consumption difference between two adjacent iterations is less than the preset energy consumption convergence threshold. Once the optimization convergence is determined, the current optimal excitation current operating point is locked, the excitation test disturbance output is paused, and the current optimal excitation current is maintained for continuous output. The external independent processing unit (2) records the adjustment step size of the excitation setpoint in each iteration and dynamically adjusts the step size according to the changing trend of the energy consumption difference between the two iterations. When the trend of energy consumption difference slows down, and the energy consumption difference between two adjacent iterations satisfies ΔP(k) < ΔP(k-1), it is determined that the trend of change has slowed down, and the step size is multiplied by the preset decay coefficient 0.
8. The energy consumption convergence threshold is set to 0.5%-2% of the rated operating power of the servo motor, and is configured online by the external independent computing unit (2) or preset by the host computer.
3. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The amplitude of the excitation test disturbance superimposed in S3 does not exceed 10% of the excitation rated value; when the change amplitude of the compensated and corrected load torque signal exceeds the pre-configured load change trigger threshold, the external independent computing processing unit (2) reactivates the excitation current small amplitude test disturbance optimization mode and performs the optimal excitation current operating point search again.
4. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The load change rate determination described in S2 specifically includes: The external independent processing unit (2) calculates the signal change of the compensated and corrected load torque signal within a unit sampling time interval, and compares the signal change with a pre-stored rate judgment threshold. The rate judgment threshold is generated by multiplying the steady-state fluctuation amplitude of the load torque signal collected by the external independent processing unit (2) during the no-load start-up phase of the servo motor by a preset safety margin coefficient, or is configured online by the host computer. If the signal change is less than or equal to the rate determination threshold, it is determined to be a slow and gradual change condition. If the signal change is greater than the rate determination threshold, it is determined to be a rapid load change, and the external independent processing unit (2) stops applying excitation test disturbance.
5. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The timing analysis processing in S1 employs a non-low-pass delay filtering algorithm, specifically including: Construct a time-domain sliding window for the original analog signal of the load torque, calculate the slope variance of the signal inside the window, and the sampling frequency is not less than 10 times the highest effective frequency of the output signal of the torque sensor (1). Signal segments with slope variance below a preset smoothing threshold are marked as components of real load with slow changes, while signal segments with slope variance above a preset smoothing threshold are marked as components of analog transmission noise. The zero-point compensation parameter is invoked to perform reverse cancellation on the analog transmission noise component. The zero-point compensation parameter is generated by the external independent processing unit (2) periodically collecting the original analog signal of the torque sensor (1) under no-load conditions. It is used to eliminate the measurement deviation introduced by the temperature drift of the torque sensor (1) itself and reduce the influence caused by this factor.
6. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The method is also configured with disturbance blocking logic: When it is detected that the servo motor is in an emergency stop state, the speed fluctuation exceeds the speed protection threshold, or the output torque exceeds the torque protection threshold, regardless of the load change rate determination result of S2, the small amplitude test disturbance optimization mode of the excitation current is directly locked, and the superimposed output of the excitation test disturbance is stopped. The speed protection threshold and torque protection threshold are set to 50%-150% of the rated speed and rated torque of the servo motor, respectively, and are pre-written by the external independent computing unit (2) through the host computer configuration interface.
7. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, Before the S4 outputs the analog control quantity, it performs a limiting process: the external independent arithmetic processing unit (2) retrieves the upper and lower limits of the servo motor excitation current and constrains the analog control quantity corresponding to the optimal excitation current to the analog input range that the servo driver (3) can recognize.
8. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The internal secondary operations performed by the servo driver (3) include at least signal range conversion operations and excitation closed-loop feedforward compensation operations: The servo driver (3) converts the received analog control quantity into the excitation reference value in the digital domain inside the driver, combines the real-time speed and real-time torque of the servo motor, queries the internally stored excitation feedforward compensation table, calculates the feedforward compensation amount, and generates the final excitation drive command output to the power drive circuit of the servo motor.
9. The servo motor adaptive energy-saving control method according to claim 1, characterized in that, The method also includes fault tolerance handling logic: The external independent computing unit (2) caches the load torque signal, the excitation test disturbance application status, the optimal excitation current operating point, and the real-time energy consumption of the servo motor in real time, and records the data timestamps synchronously. When the external independent processing unit (2) detects that the torque sensor (1) signal is abnormal or the analog signal transmission is disconnected, it prioritizes retrieving the historical data with the most recent timestamp and at that time the excitation test disturbance was in a convergent steady state to generate an analog control quantity for output. If the timestamp of the cached data exceeds the preset valid time limit, the pre-stored default excitation setpoint will be retrieved and output.
10. The servo motor adaptive energy-saving control method according to claim 3, characterized in that, The specific method for collecting real-time energy consumption data of servo motor operation in S3 is as follows: The external independent processing unit (2) reads the DC bus voltage and DC bus current values of the servo driver (3) in real time via a high-speed sampling interface; Instantaneous power data is obtained by synchronously multiplying the DC bus voltage value and the DC bus current value. A moving average filter is applied to the instantaneous power data to remove high-frequency switching noise from the power circuit. The processed average power value is then used as real-time energy consumption data for iterative optimization.
