Reduction gear multi-field loss equalization method and system based on working condition adaptive feedback

Abstract

The application relates to the field of mechanical engineering and discloses a reducer multi-field loss balancing method and system based on working condition self-adaptive feedback, aiming to solve the problems of local loss concentration, temperature rise aggravation and low energy efficiency caused by neglecting the multi-physical field coupling relationship in the prior art. The method comprises the following steps: collecting mechanical, thermal, fluid and electromagnetic parameters in the operation of the reducer through a distributed sensor to generate a standardized working condition state vector; inputting the standardized working condition state vector into a mechanical-thermal-fluid-electromagnetic four-field coupling model, and solving the time and space loss distribution of key components in combination with a data-driven correction term. Through the above scheme, the application realizes fine prediction and closed-loop regulation and control of the multi-field loss distribution, significantly reduces local temperature difference and energy consumption concentration, improves the overall energy efficiency by 8%-15%, prolongs the service cycle, and is suitable for high-end equipment transmission systems under high dynamic heavy load working conditions.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of mechanical engineering, and particularly relates to a reducer multi-field loss balancing method and system based on working condition adaptive feedback. BACKGROUND

[0002] With the continuous advancement of industrial automation and intelligent manufacturing, as a core component in the transmission system, the reducer is widely used in the fields of robots, numerical control machine tools, wind power equipment and new energy vehicles, and its running efficiency and energy consumption performance directly affect the performance and reliability of the whole machine system. Under complex and variable working conditions, the reducer involves the coupling of multiple physical fields such as mechanical field, thermal field and lubrication field, and the energy loss between the fields affects each other, resulting in problems such as local loss concentration, temperature rise and life attenuation. Therefore, the coordinated optimization of loss distribution under multi-field environment has become the key to improving the comprehensive efficiency of the transmission system.

[0003] Among them, for the energy consumption control in the running process of the reducer, the existing technology mainly focuses on the optimization design of a single field, such as reducing mechanical friction loss through tooth profile modification, or improving lubrication characteristics by using high-efficiency lubricating oil, or increasing heat dissipation structure to relieve heat accumulation effect. However, such methods often ignore the dynamic coupling relationship between different physical fields, making it difficult to adapt to non-steady-state conditions such as load fluctuation, frequent start-stop and environmental temperature change in actual operation, resulting in limited optimization effect, and even in some working conditions, the negative effect of "this reduces that increases" occurs, that is, the reduction of a certain type of loss may cause a significant increase in another type of loss.

[0004] The existing technology has not established a multi-field loss dynamic balancing mechanism based on actual working condition feedback, and lacks the ability of unified modeling and real-time control of multi-source loss paths such as mechanical, thermal and lubrication. At the same time, most schemes do not introduce adaptive feedback strategies, and cannot adjust the control parameters according to the running state to achieve global efficiency optimization. In addition, there is a lack of precise perception and prediction ability for the change law of loss distribution characteristics under different working conditions, which causes the system to run at a suboptimal working point for a long time, resulting in energy waste and accelerated aging of components. These problems are particularly prominent in high-dynamic and heavy-load application scenarios, seriously restricting the development of high-end equipment transmission systems towards high efficiency and intelligence, and there is an urgent need for a new control method and system architecture that can realize working condition adaptation, multi-field coordination and loss balancing. SUMMARY

[0005] The purpose of the present application is to make up for the deficiencies of the prior art, and provide a reducer multi-field loss balancing method and system based on working condition adaptive feedback, which can effectively solve the problems in the background art. When the current reducer operates in complex variable load, variable speed and frequent start-stop non-steady state conditions, there is a strong coupling effect between the internal mechanical, thermal, lubrication and electromagnetic (such as integrated motor drive) multi-physical fields, resulting in problems such as local stress concentration, uneven temperature rise, oil film rupture and unbalanced energy loss distribution. Traditional optimization strategies usually only focus on a single performance indicator for static design or open-loop control, and cannot achieve coordinated adjustment of multi-field dynamic response, making it difficult to balance efficiency, service life and reliability. The local over-loss phenomenon caused by this significantly reduces the overall energy efficiency and service life of the system, restricting the further development of high-end equipment transmission systems.

[0006] To achieve the above purpose, the present application provides the following technical solutions: on the one hand, a reducer multi-field loss balancing system based on working condition adaptive feedback, the system comprises the following components: a real-time working condition sensing module for collecting input torque, output speed, shell vibration signal, lubricating oil temperature and pressure, environmental temperature and humidity, and driving motor current and voltage and other multi-source sensing data during the operation of the reducer, and performing time synchronization and preprocessing on the data to generate a standardized working condition state vector; a multi-field coupling modeling module receives the working condition state vector, based on a pre-set physical mechanism model and a data-driven correction mechanism, constructs a time-varying model of gear mesh stiffness, a bearing support stiffness model, a dynamic evolution model of lubricating oil film thickness, and an electromagnetic loss distribution model, and then realizes mechanical-thermal-fluid-electromagnetic four-field coupling solution through field interface variable transmission, and outputs the spatial loss density distribution cloud and time sequence change curve of each key component; a loss balancing decision module receives the multi-field loss distribution results, combines a pre-set global objective function with "loss balancing" as the core orientation, comprehensively considers three sub-goals of maximum single-point loss suppression, regional cumulative loss smoothness and system total efficiency improvement, and generates an optimal control instruction set using a multi-objective weighted optimization algorithm; an adaptive execution module receives the control instruction set, adjusts the lubricating oil pump flow / pressure, cooling fan speed, active lubrication nozzle switch timing, clutch engagement pressure or motor output characteristic parameters, realizes closed-loop intervention and dynamic reconstruction of the multi-field distribution inside the reducer; a data feedback updating module collects the actual operating state data after control, compares the predicted loss distribution with the measured indirect indicators (such as infrared thermal image partition temperature difference, vibration energy spectrum distribution), calculates the deviation residual and uses it to correct the multi-field coupling model parameters online, forming a complete feedback link from sensing to execution to model evolution;

[0007] Further, in the real-time working condition perception module, the input torque and the output rotating speed are obtained by high-precision torque sensors and photoelectric encoders, and the sampling frequency is not less than 10 kHz, so as to ensure capturing transient impact load; the shell vibration signals are collected by a three-axis acceleration sensor array arranged at key support positions of the box, and the spatial resolution is not less than 4 measuring points per square meter, so as to identify structural resonance modes and stress propagation paths; the lubricating oil temperature and pressure are measured by distributed miniature PT100 temperature probes and piezoresistive pressure transmitters arranged at the oil inlet, the oil outlet and the tooth side injection area, and the response time is less than 50 ms, so as to ensure accurate characterization of the heat-flow coupling boundary conditions; the ambient temperature and humidity are measured by an external meteorological level sensor, and are used for heat exchange boundary condition compensation calculation; all sensor data are time-stamped and aligned by an industrial edge computing node, and abnormal values are filtered out, and a sliding window mean value method and wavelet denoising are combined for processing, so as to output stable and reliable working condition state vector sequences;

[0008] Preferably, in the multi-field coupling modeling module, the gear meshing stiffness time-varying model is established based on finite element contact analysis and Hertz theory, taking into account the influence of tooth surface modification parameters, manufacturing errors and elastic deformation, and the stiffness curve continuously changes with the meshing position and is input as an excitation source into the dynamic equation; the bearing support stiffness model distinguishes the contact angle changes of the rolling elements and the inner and outer rings, dynamically updates the support point stiffness matrix in different load directions, and supports the solution of the 6-degree-of-freedom displacement-force mapping relationship; the lubricating oil film thickness dynamic evolution model adopts a modified Reynolds equation, introduces a viscosity-density relationship formula that changes with temperature and pressure, and the boundary conditions are determined by the relative motion speed of the gears and the surface roughness, so as to accurately divide the elastohydrodynamic lubrication (EHL) region; the electromagnetic loss distribution model is suitable for integrated electric drive reducers, including stator copper loss, core eddy current and hysteresis loss, and rotor stray loss components, the spatial distribution of which is closely related to the magnetic field harmonics of the motor, and is calibrated through magnetic field-circuit coupling simulation; further, the four-field coupling solution adopts a loose coupling iteration strategy, the mechanical field dynamics equation outputs the load distribution to drive the boundary update of the thermal field and the flow field, the thermal field temperature field acts on the material properties and the lubricating oil viscosity, the flow field oil film pressure is fed back to the support stiffness and friction torque, and the electromagnetic field loss is added as a heat source to the temperature rise calculation, and the convergence tolerance of each iteration is set to be less than 3% of the relative change between fields.

[0009] Preferably, in the loss balancing decision module, the global objective function is defined as: minimizing , wherein represents the instantaneous power loss of the i-th grid cell, represents the power loss of the i-th grid cell in the j-th time interval, represents the power loss of the i-th grid cell in the j-th time interval, The adjustable weight coefficient is in the range of [0.8, 1.2] and [0.6, 1.0] respectively, and is automatically switched according to the working condition type; the regional cumulative loss smoothness is quantified by calculating the standard deviation of the loss integral in the adjacent space region, reflecting the thermal stress gradient risk; the multi-objective weighted optimization algorithm adopts an improved non-dominated sorting genetic algorithm (NSGA-II), with a population size of 100, a maximum evolution generation of 50, a crossover probability of 0.9, a mutation probability of 0.1, an optimization solution completed every 2 seconds, and the closest regulation combination to the ideal point in the Pareto frontier output; the regulation instruction set includes discrete-continuous hybrid variables such as lubricating oil distribution ratio, cooling intensity level, active lubrication trigger threshold and motor derating coefficient, which are converted into device executable commands through rule mapping;

[0010] Further, in the adaptive execution module, the lubricating oil pump adopts frequency control, with a flow regulation range of 30% to 100% of the rated value, a step size of 5%, and a response delay of less than 200ms; the cooling fan is configured with multiple speed regulating gears, which are dynamically switched according to the hot spot temperature rise rate, and the highest gear is started to force air cooling when the local temperature rise slope is detected to exceed 5°C / s; the active lubrication nozzle is independently controlled according to space zoning, with the injection timing synchronized with the gear meshing phase, and the oil injection pulse width determined by the current meshing area load size, ensuring that the heavy load gear surface is preferentially supplied with oil; the clutch engagement pressure regulation is used for load transfer control during gear shifting of the multi-gear reducer, avoiding local over-damage caused by instantaneous impact; the motor output characteristic parameter regulation includes torque slope limitation and adjustment of the intervention time of field weakening control, reducing the additional iron loss caused by high-frequency harmonic current; all actuators have a fault safety mode, which restores to the default energy-saving running state when communication is interrupted or the instruction is out of limit;

[0011] In addition, in the data feedback update module, the measured indirect indicators are obtained by non-invasive means, an infrared thermal imager is deployed on the outer wall of the reducer at a specific observation window, with a spatial resolution of not less than 320x240 pixels and a frame rate of 25fps, used to extract the average temperature and maximum temperature difference of each functional area; the vibration energy spectrum distribution is obtained by short-time Fourier transform of the acceleration signal, focusing on the energy aggregation degree in the frequency band of 1kHz to 8kHz as a criterion for friction state and oil film integrity; the deviation residual is defined as the normalized Euclidean distance between the predicted and measured indicators, and when the average residual in the last 5 sampling periods exceeds the preset threshold of 0.15, the model parameter online identification process is triggered; the online identification adopts recursive least squares method (RLS) combined with forgetting factor (λ=0.98) to estimate and update the key uncertainty parameters such as lubricating oil viscosity-temperature coefficient, heat dissipation coefficient and contact thermal resistance in real time, ensuring the long-term effectiveness of the model;

[0012] In another aspect, a multi-field loss balancing method for a reducer based on working condition adaptive feedback is provided. The method comprises the following steps: in step S110, multi-dimensional working condition data during the operation of the reducer is collected in real time by a distributed sensor network, including mechanical, thermal, fluid and electromagnetic parameters, and time synchronization and noise suppression processing are performed to generate a uniform format working condition state vector; in step S120, the working condition state vector is input into a multi-field coupled physical model, and a data-driven correction term is combined to solve the loss density distribution of key components such as gears, bearings, shafts and motors in space and time dimensions; in step S130, based on the loss distribution result, a multi-objective optimization problem with "loss balancing" as the target is constructed, and an optimal control instruction set that can simultaneously reduce peak loss, smooth regional cumulative loss and improve system total efficiency is obtained; in step S140, the control instruction set is sent to the actuator to adjust the lubricating oil supply strategy, cooling intensity, active lubrication logic or motor operating point, so as to realize dynamic reconstruction of the multi-physical field distribution inside the reducer; in step S150, real running state data after control is collected, indirect observable indicators are extracted and compared with model prediction results, deviation residuals are calculated, and when the residuals exceed the tolerance range, the model parameter online updating mechanism is started to complete self-correction and iterative evolution of the multi-field coupled model.

[0013] Preferably, in step S110, the multi-dimensional working condition data collection process adopts a time-triggered communication protocol to ensure data consistency, all sensors are marked with UTC time stamps at the sampling time, and the maximum allowed clock deviation is ±10μs; the noise suppression processing adopts an adaptive decomposition method based on empirical wavelet transform (EWT), selects the optimal modal segmentation method for different signal types, and removes random interference and periodic interference while retaining effective feature bands; the working condition state vector is packaged in the form of a fixed-length data frame, containing original values, unit identifiers and quality labels for subsequent module calls.

[0014] Further, in step S120, the multi-field coupled physical model has completed offline calibration in the full working condition range in the initialization stage, covering at least 100 typical load spectrum and temperature condition combinations; the data-driven correction term is composed of a deep residual network, the input is the current working condition state vector and the model prediction residual at the last time, and the output is the correction amount of the key parameters of each sub-model, such as mesh stiffness offset, convective heat transfer coefficient gain factor, etc., thereby improving the generalization ability of the model under unknown working conditions; the solving process runs on an embedded high-performance computing unit, and the time consumption of a single solving process is controlled within 80ms, meeting the real-time requirement.

[0015] Further, in the step S130, the constraint conditions of the multi-objective optimization problem include the physical limits of the actuator, the minimum lubrication guarantee requirement and the temperature rise safety boundary, and any control instruction cannot violate the above-mentioned hard limits; the selection process of the optimal control instruction set introduces a fuzzy decision mechanism, converts the operator preference (such as priority to life or priority to efficiency) into weight coefficient adjustment rules, and enhances the human-machine collaborative adaptability of the system.

[0016] Compared with the prior art, the present application has the following beneficial effects:

[0017] By constructing a mechanical-thermal-fluid-electromagnetic four-field coupling model and realizing closed-loop feedback updating, the fine and dynamic prediction of the internal multi-physical field loss distribution of the reducer is realized for the first time, and the prediction accuracy is improved by not less than 40% compared with the traditional single field model.

[0018] A multi-objective optimization strategy with "loss balance" as the core is proposed, which breaks through the limitation of only pursuing the maximum total efficiency in the past, effectively suppresses the formation of local hot spots, and reduces the maximum temperature difference of key components by more than 25%, significantly delaying the initiation of fatigue cracks.

[0019] An operating condition self-adaptive feedback mechanism is adopted to support online model correction and dynamic evolution of control strategy, and the system still maintains a prediction deviation of less than 10% after 500 hours of continuous operation, and the long-term stability is far superior to that of an open-loop control system.

[0020] The technical leap from passive bearing loss to active regulation loss is realized, and the overall energy efficiency of the machine is improved by 8%-15%, especially in heavy load and variable speed operating conditions.

[0021] The system is compatible with multiple types of reducer platforms, and is especially suitable for high dynamic demand scenarios such as wind power, engineering machinery and new energy vehicles, and has wide engineering promotion value. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 The figure is a schematic diagram of the overall technical scheme architecture of the reducer multi-field loss balancing method and system based on operating condition self-adaptive feedback according to the present application.

[0023] Figure 2 The figure is a schematic diagram of the core principle framework of the multi-field coupling modeling and loss balancing decision-making synergy in the present application. DETAILED DESCRIPTION

[0024] Please refer to Figure 1 and Figure 2 , in order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the specific embodiments, structures, features and effects of the present application will be described in detail below in combination with the drawings and preferred embodiments.

[0025] Embodiment one

[0026] This embodiment takes the working condition adaptive feedback multi-field loss balancing system applied to large wind turbine gearbox as the background, and describes the complete technical implementation path of the system in detail. Wind power equipment faces complex working conditions such as sudden change of wind speed, frequent start and stop, and extreme environmental temperature fluctuation during operation, which leads to drastic changes in mechanical stress inside the reducer, easy rupture of lubricating oil film, and significant local temperature rise. The traditional static optimization strategy is difficult to cope with such strong non-steady state dynamic process. Therefore, this embodiment constructs a complete system architecture integrating real-time sensing, multi-physical field coupling modeling, loss balancing decision, closed-loop execution and model online evolution.

[0027] The system first collects multi-dimensional working condition data through a distributed sensor network deployed at key positions of the reducer. The input end high-precision torque sensor is installed at the high-speed shaft inlet, designed with a strain gauge full-bridge circuit, with a range of -2000 Nm to +6000 Nm, a resolution of 0.1 Nm, and a sampling frequency of 12 kHz, ensuring accurate capture of instantaneous torque impact caused by wind shear. The output end optical encoder is installed at the end of the low-speed shaft, using incremental A / B / Z phase output mode, with a line number of 4096 PPR, and after four times frequency processing, the angle resolution can reach 0.022°, and the corresponding time resolution is 83.3μs, which can still ensure sufficient signal density in the low wind speed running state of less than 15 rpm per minute. Shell vibration monitoring uses a three-axis acceleration sensor array, with a total of 16 measuring points, located at the front and rear bearing seats, the middle part of the box reinforcement and the flange area of the base connection, the spatial distribution meets the requirement of 3.8 measuring points per square meter, and the sensor model is IEPE piezoelectric accelerometer, with a sensitivity of 100 mV / g and a frequency response range of 0.5 Hz to 10 kHz, used to capture typical fault characteristic frequencies such as gear meshing frequency and its sideband components, rolling element passing frequency, etc.

[0028] The lubricating oil system is equipped with three micro PT100 temperature probes, which are embedded in the oil pipeline (20 cm away from the pump outlet), the oil outlet collection tank, and the tooth surface projection area 5 mm downstream of the active nozzle. The probe diameter is not more than 4 mm, the response time is controlled within 45 ms, the measurement range is -40°C to 150°C, and the accuracy is ±0.3°C. The pressure detection uses a silicon piezoresistive transmitter installed in the T-shaped three-way node of the main oil supply pipeline, with a range of 0 to 1.6 MPa, an overload capacity of 2.5 MPa, an output of 4-20 mA standard signal, and a response time of less than 35 ms. The environmental temperature and humidity sensor is placed in the cabin near the reducer shell but not directly affected by the heat radiation. It uses a combination of capacitive humidity sensitive elements and platinum resistance structure, with a temperature measurement range of -40°C to 85°C, a relative humidity measurement range of 0% to 100% RH, and an accuracy of ±0.5°C and ±2% RH, respectively. It participates in the dynamic compensation calculation of the external natural convection heat exchange boundary conditions.

[0029] All sensor data is accessed by an industrial-grade edge computing node based on an ARM Cortex-A72 quad-core processor with a main frequency of 1.8 GHz, equipped with 4GB DDR4 memory and 32GB solid state storage, running a real-time Linux operating system (PREEMPT_RT patch) to ensure task scheduling certainty. Data collection uses a time-triggered communication protocol (TTCAN), with all devices synchronized to the UTC time reference, using a GPS module to provide nanosecond-level timing signals, with the maximum allowed clock deviation for each channel strictly controlled within ±8μs. The raw data frame contains fields: timestamp (UTC microsecond level), sensor ID, raw value, unit code (predefined enumeration value), quality label (0 for normal, 1 for abnormal to be filtered, 2 for failure). The data preprocessing process performs sliding window mean filtering (window length 100 points) and adaptive denoising based on empirical wavelet transform (EWT) in sequence. The EWT decomposition process automatically divides the scale interval according to the signal spectrum characteristics, prioritizing the preservation of IMF components within the 800Hz to 6kHz frequency band for vibration signals, and suppressing power grid harmonic interference; for temperature and pressure signals, focus on low-frequency trend items, and eliminate high-frequency burrs. Finally, a uniform format of working condition state vector is generated, with a dimension of 1x28, including: input torque, output speed, 16 vibration channel RMS values, 3 temperature readings, 2 pressure values, environmental temperature, and environmental humidity, all normalized to the [0,1] interval and with an effectiveness flag bit.

[0030] The working condition vector is transmitted in real time to the multi-field coupling modeling module, which is deployed in an embedded high-performance computing unit (equipped with NVIDIA Jetson AGX Xavier platform) with 32 TOPS INT8 computing power and supports model parallel solution. The modeling process is divided into two levels: the bottom layer is a four-field coupling model based on physical mechanism, and the top layer is a data-driven correction mechanism composed of deep residual network. The time-varying model of gear engagement stiffness is established based on three-dimensional finite element contact analysis, with fine meshing in the fillet area of the tooth root, a total of about 1.2 million degrees of freedom. The model takes into account the parameters of the tooth surface drum modification (modification amount 35 μm, modification length ratio 60%), cumulative error of pitch (±18 μm), tooth inclination error (±12 μm), and elastic deformation of shafting. In each meshing period, 120 discrete points are divided along the meshing line direction, and the contact pressure and deformation are calculated point by point through the Hertz contact theory, and then the instantaneous engagement stiffness curve is obtained by integration. The amplitude shows a nonlinear growth trend with the increase of load, and obvious fluctuations appear in the single-tooth and double-tooth alternating area. The stiffness sequence is input to the multi-body dynamics equation as an excitation source:

[0031]

[0032] wherein is the generalized mass matrix, is the damping matrix (including material internal friction and lubrication shear loss), is the nonlinear stiffness matrix, is the generalized coordinate vector, is the external load vector.

[0033] The bearing support stiffness model uses the Jones-Harris rolling bearing theoretical framework, which considers the inner and outer rings, retainers and rolling elements as independent rigid bodies, and considers the Hertz contact deformation between the balls and the raceway. The model distinguishes the contact angle changes under radial and axial load actions, and dynamically updates the support point stiffness matrix in six degrees of freedom under different working conditions. For example, when the axial force increases, the contact angle of the angular contact ball bearing increases from the initial 25° to 32°, resulting in an increase in axial stiffness and a decrease in radial stiffness. The displacement-force mapping relationship of each support point is output by the model as the boundary condition input for structural vibration analysis.

[0034] The dynamic evolution model of the lubricating oil film thickness is solved based on the two-dimensional transient Reynolds equation, and the Dowson-Higginson viscosity-density relationship is introduced:

[0035]

[0036] wherein, is the density under reference state, usually the initial density under normal pressure and room temperature, Reference pressure, usually atmospheric pressure, Reference viscosity, viscosity at atmospheric pressure and reference temperature, Reference temperature, usually ambient temperature, Density of lubricant, Kinematic viscosity, Local film pressure, Temperature, Viscosity-pressure exponent. The boundary conditions are determined by the relative velocity of the gear (up to 15 m / s) and the surface roughness (Ra=0.4 μm). The solution domain is divided into 200×150 grids, and the finite difference method is used for iterative solution to identify the spatial distribution of the EHL (elastohydrodynamic lubrication) region, the mixed lubrication region, and the boundary lubrication region. The minimum thickness of the oil film is predicted to be no less than 1.8 μm, and if it is lower than this threshold, it is determined that there is a risk of metal contact.

[0037] The electromagnetic loss distribution model is applicable to the electric drive reducer structure of the integrated permanent magnet synchronous motor. The stator copper loss is calculated by taking into account the increase in AC resistance caused by skin effect; the core loss adopts the Bertotti separation model, which is decomposed into three parts: eddy current loss, hysteresis loss, and residual stray loss, which are proportional to the square, first power, and 1.5 power of the magnetic flux density, respectively; the rotor eddy current loss is obtained through three-dimensional transient magnetic field simulation, focusing on the circulating current heating caused by end leakage. All electromagnetic losses are superimposed as spatial distribution heat sources to the right side of the heat conduction equation.

[0038] The four-field coupling solution adopts a loose coupling iterative strategy, and a complete solution cycle is executed every 20 ms. The current working condition vector is loaded at the initial time, the mechanical field is solved, the gear meshing force, bearing reaction force, and shaft deformation distribution are obtained; the above results are input as the boundary conditions of the thermal field and the flow field to solve the temperature field distribution based on the Fourier heat conduction law and the convective heat transfer equation; at the same time, the temperature field is fed back to the material property database to update the elastic modulus of the gear steel (linearly decreasing with temperature rise), the viscosity of the lubricating oil (exponential decay), and the contact thermal resistance (related to surface roughness and pressure); the flow field oil film pressure distribution acts on the support stiffness model, forming a lubrication-structure feedback loop; electromagnetic loss as an additional heat source participates in the temperature rise calculation. The relative change rate of the field variables (such as the maximum contact stress, hot spot temperature, and oil film pressure peak value) is judged at each iteration, and when the changes of all variables in three consecutive iterations are less than 2.8%, it is determined to be converged, and the spatial loss density distribution cloud map and time sequence curve at the current time are output. The average time consumption of the entire solution process is 76 ms, which meets the 80 ms period requirement of the control system.

[0039] The loss balancing decision module receives the loss distribution results from the modeling module and constructs a multi-objective optimization problem with "loss balancing" as the core. The objective function is defined as:

[0040]

[0041] wherein is the finite element mesh cell index, is the instantaneous power loss of the th cell; is the preset functional area (such as the first-stage gear pair meshing area, the intermediate shaft bearing area, the final-stage planet carrier support area, etc.), a total of 9 areas are divided; is the standard deviation of the area cumulative loss, reflecting the thermal stress gradient risk; is the system total efficiency, defined as the ratio of output power to input power; the weight coefficient and According to the working condition type, dynamically adjust: set =1.1, =0.7 in the starting and accelerating stage, focusing on suppressing peak loss; adjust to =0.9, =0.9 in the stable running stage, taking into account efficiency and uniformity; set =1.2, =0.6 in the emergency braking stage, giving priority to preventing local overheating.

[0042] The constraint conditions include: the minimum flow of the lubricating oil pump should not be less than 35% of the rated value to ensure the minimum lubrication requirement; the maximum power consumption of the cooling fan should not exceed 15% of the system auxiliary power; any control instruction should not make the predicted maximum temperature exceed the material allowable limit (gear steel is 180°C, sealing element is 120°C); the intervention time of motor field weakening control should not be earlier than 85% of the rated speed to avoid mechanical impact caused by sudden torque drop.

[0043] The optimization algorithm uses an improved NSGA-II, the population size is 100, the coding method is a mixed integer-real vector, and contains 5 decision variables: lubricating oil pump frequency set value (real number, 30%-100%), cooling fan gear position (integer, 1-5 gears), number of active lubrication nozzle openings (integer, 1-4 groups), clutch engagement pressure offset (real number, ±20% of the reference value), motor derating coefficient (real number, 0.8-1.0). The crossover operation uses simulated binary crossover (SBX) with a distribution index of 15; the mutation uses polynomial mutation with a distribution index of 20. Optimization solution is completed every 2 seconds, generating a Pareto frontier containing about 40 non-dominated solutions. The decision mechanism uses fuzzy comprehensive evaluation method to convert operator preferences into weight adjustment rules: if "life priority" is selected, then increase by 10%, decrease by 10%; if "efficiency priority" is selected, then the opposite. Finally, the ideal point (minimum , minimum , maximum ) the solution with the shortest Euclidean distance as the optimal control instruction set.

[0044] After receiving the control instruction set, the adaptive execution module maps it to specific execution commands. The lubricating oil pump is driven by a frequency converter, which adopts a V / F control strategy, with a frequency adjustment step of 0.5 Hz (corresponding to a flow step of about 5%). The actual response delay is measured to be 185 ms, which meets the design requirement of less than 200 ms. The cooling fan is configured with five speed stages, and is switched by PWM duty cycle. When the infrared thermal imager detects that the temperature rise rate of any area exceeds 4.8°C / s, it immediately switches to the fifth gear to force air cooling, and the air volume is increased to 140% of the rated value. The active lubrication nozzles are independently controlled according to the space partition, with each group corresponding to a gear pair. The injection timing is triggered by the encoder signal, and the rising edge of the pulse width modulation signal is synchronized with the gear entering the meshing area 1.2 ms in advance. The oil injection duration is linearly adjusted in the range of 2-8 ms according to the current load size, ensuring that the heavily loaded tooth surface is fully lubricated. The clutch engagement pressure is adjusted by an electro-hydraulic proportional valve, with a control signal of 0-10V analog quantity corresponding to a pressure range of 0-3.5MPa. During gear shifting, a ramp-type pressure loading is implemented, with a rise time set to 300ms to avoid load sudden change. The motor output characteristic parameters are transmitted to the driver through CAN bus, adjusting the torque slope limit value (originally 500 Nm / s, can be reduced to 200 Nm / s) and the weak magnetic control starting speed (originally 4500 rpm, can be delayed to 4800 rpm) to reduce the additional iron loss caused by high-frequency harmonic current.

[0045] All actuators are configured with safety monitoring logic. Once communication interruption, instruction overrun or hardware failure is detected, the fail-safe mode is triggered immediately: the lubricating oil pump resumes to 60% rated flow operation, the cooling fan maintains medium-speed operation at the third gear, the active lubrication system closes non-critical nozzles and only retains basic spraying, and the motor switches to constant power output mode. Although this mode is not optimal, it can ensure the safe operation of the system for at least 4 hours under abnormal conditions.

[0046] The data feedback update module continuously collects the actual running state data after control. The infrared thermal imager is installed on the observation window of the reducer outer wall, using a non-refrigerated vanadium oxide detector, with a resolution of 384x288 pixels, a spectral response range of 7.5-14μm, and a frame rate of 25fps, obtaining a temperature field image every 40ms. After distortion correction and radiation calibration, the average temperature and maximum temperature difference of 9 functional areas are extracted. The vibration energy spectrum distribution is obtained by short-time Fourier transform (STFT) on three-axis acceleration signals, with a window length of 512 points, an overlap rate of 75%, and a frequency spectrum resolution of 4.88Hz. The energy concentration degree in the frequency band of 1kHz to 8kHz is analyzed as a criterion for oil film integrity—if the energy proportion in this frequency band exceeds 65% of the total energy, it indicates that there is a risk of boundary lubrication.

[0047] The deviation residual is defined as the normalized Euclidean distance between the predicted index and the measured index:

[0048]

[0049] in, For the first Predicted values ​​for each variable, For the first The measured values ​​of each variable. For the first 10 observed variables (total) =12, including 9 regional temperatures and 3 vibration frequency band energy). Its standard deviation. Within 5 consecutive sampling periods (total 10 seconds) When the mean exceeds 0.145, the model is deemed inaccurate, and the online identification process is initiated. The identification algorithm employs recursive least squares (RLS) combined with a forgetting factor (λ=0.98). The parameters to be estimated include the lubricating oil viscosity-temperature coefficient (ISO VG320 baseline value corrected by ±15%), the shell convective heat transfer coefficient (baseline value ±30%), and the gear contact thermal resistance (baseline value ±25%). The parameter update cycle is 1 second, and the multi-field coupled model is reinitialized after each update to achieve self-correction and iterative evolution. Long-term operation tests show that after 520 hours of continuous variable load testing, the prediction deviation stabilizes within 9.7%, verifying the effectiveness of the model update mechanism.

[0050] Example 2

[0051] This embodiment focuses on the application scenario of a two-speed electric drive reducer in new energy vehicles, highlighting a loss balancing implementation method based on a different algorithm architecture. The core technical difference lies in the use of a surrogate model to replace the solution of some physical equations in the multi-field coupling modeling module, thereby significantly reducing the computational load and adapting to the resource-constrained characteristics of the vehicle controller. Compared to Embodiment 1, this embodiment does not simply change the application scenario, but rather reconstructs the core modeling logic while ensuring functional consistency, resulting in a substantial difference in technical approach.

[0052] In new energy vehicle operating conditions, the reducer needs to frequently respond to the driver's accelerator pedal commands, experiencing various transient processes such as rapid acceleration, regenerative braking, and idle start-stop. The system has extremely high real-time control requirements, and must complete the entire link response from sensing to execution within 20ms. Therefore, the high-fidelity but computationally intensive finite element-Reynolds equation joint solution scheme in Example 1 cannot be directly used. To address this, this example proposes a hybrid modeling paradigm of "physical guidance + data fitting".

[0053] The sensor configuration and data preprocessing process of the real-time operating condition perception module are basically the same as in Implementation Example 1, but the sampling strategy has been adjusted: to reduce the load on the vehicle CAN FD bus, vibration signals are sampled using event triggering. Full-speed acquisition at 10kHz is initiated when the detected acceleration peak exceeds 0.8g, and reduced to 1kHz for the remaining time. The sampling interval for temperature and pressure sensors is extended from 100ms to 200ms, and Kalman filtering is used to predict intermediate values. The operating condition state vector is compressed to a 1×20 dimension, removing some redundant vibration channels while retaining key position information.

[0054] The core innovation of the multi-field coupling modeling module lies in introducing three types of surrogate models to replace the traditional numerical solver. The first type is the gear meshing stiffness surrogate model, whose training data comes from an offline high-precision finite element simulation database. It covers a three-dimensional working condition mesh with input torque of 0-3000 Nm, speed of 0-8000 rpm, and oil temperature of -20°C to 100°C, generating a total of 120,000 samples. A deep neural network (DNN) is used for fitting, with a network structure of 6 fully connected layers (28-dimensional input layer, 512, 256, 128, 64, and 32 hidden layer nodes respectively, and a 120-dimensional output layer corresponding to the meshing stiffness sequence). The activation function is Swish, and the loss function is MAE. After training, the prediction error is less than 3.5%. This model directly receives the working condition state vector as input and outputs a complete time-varying meshing stiffness curve, with a computation time of only 8 ms.

[0055] The second type is the thermal field distribution surrogate model, constructed based on Gaussian process regression (GPR). Training data is obtained from three-dimensional transient heat conduction simulations, considering the temperature field response under different combinations of cooling airflow, oil flow distribution, and electromagnetic loss inputs. The first 20 modes are selected as basis functions, and the kernel function parameters (RBF kernel) are automatically selected using Bayesian optimization to achieve efficient approximation of complex nonlinear heat conduction behavior. The model input consists of historical temperature sequences (10 steps long) from six representative temperature measurement points, the current cooling intensity level, and lubricating oil flow rate. The output is a predicted temperature distribution across the entire enclosure surface, maintaining spatial resolution consistent with the infrared camera. A single inference takes 12 ms.

[0056] The third type is the lubrication state classification model, implemented using a lightweight convolutional neural network (CNN-Lite). The input is the energy spectrum matrix (64×64) of the vibration signal in the 1kHz-8kHz frequency band, and the output is the probability distribution of four lubrication states: complete elastohydrodynamic lubrication, mixed lubrication, boundary lubrication, and dry friction. The network contains only 3 convolutional layers (kernel size 3×3, channel numbers 16 / 32 / 64), 2 pooling layers, and 1 fully connected layer, with a total of less than 50,000 parameters, allowing it to run on MCU-level processors. This model replaces the traditional Reynolds equation solution, achieving rapid discrimination of lubrication states.

[0057] The four-field coupling relationship is achieved through a state transition diagram: mechanical field output load distribution → look-up table to obtain bearing stiffness variation → input to the dynamic model → generate vibration spectrum → send to the lubrication state classification model → output lubrication rating → feedback to the friction coefficient correction term → affect heat source input → input to the thermal field surrogate model → output temperature distribution → react on material properties and oil parameters → update to the mechanical field and lubrication model. The entire coupling solution period is controlled within 18 ms, meeting the real-time requirements of vehicle-mounted.

[0058] The objective function form of the wear equalization decision module remains unchanged, but the optimization algorithm is replaced by the fast approximate Pareto search algorithm (FAPS) designed for embedded platforms. This algorithm is based on a reference point-guided local search strategy, does not generate a complete Pareto front, but directly locates the feasible solution near the ideal point. The population size is reduced to 20, and a greedy initialization strategy is used, with only 10 iterations converging. The decision variable encoding is simplified to discrete: lubricating oil pump gear (1-5 levels), cooling fan gear (1-3 levels), active lubrication enable switch (0 / 1), motor operating mode (standard / energy saving / motion). Optimization is completed every 1.5 seconds, with power consumption only 23% of the original algorithm.

[0059] The adaptive execution module is adapted to the vehicle-mounted environment. The lubricating oil pump is replaced by a constant-displacement pump + proportional overflow valve combination structure, with flow control achieved by adjusting the valve opening, and the response time improved to 120 ms. The cooling fan is integrated into the motor cooling system, sharing the same PWM drive circuit, and the gear switching is coordinated by the vehicle controller. The active lubrication nozzle is only enabled when the high-speed gear is running, controlled by an electromagnetic valve, and the oil injection timing is linked to the clutch separation action, starting 150 ms in advance to pre-lubricate the meshing tooth surface. Motor output characteristic adjustment is achieved by modifying the torque mapping curve in the vehicle VCU, with the derating coefficient pre-stored in the form of a MAP table, dynamically called according to the battery SOC and motor temperature.

[0060] The data feedback update module uses an incremental model calibration mechanism. Due to the inability to carry an infrared thermal imager, a distributed NTC temperature sensor array (12 measurement points) is used to estimate the surface temperature gradient. The bias residual calculation is based on the weighted mean square error of the predicted temperature and the measured temperature of the surrogate model, with the weights allocated according to the importance of the measurement points (bearing seat weight 1.0, middle of the shell 0.6, far from the heat source area 0.3). When the residual continuously exceeds the standard, the surrogate model fine-tuning program is started: freeze the deep network weights, only update the last layer of fully connected layers online gradient descent, learning rate set to 0.001, each update only uses the latest 5 groups of data to avoid catastrophic forgetting. Tests show that this mechanism can reduce the thermal model prediction error from the initial 18% to 6.2% after the vehicle has traveled 1000 kilometers.

[0061] This embodiment successfully reduces the system computing demand to the range that can be borne by the vehicle-mounted ECU without sacrificing the core function by introducing an agent model system, which embodies the expansion of the engineering applicability of the application on resource-constrained platforms and forms a complementary technical solution with embodiment one.

[0062] Embodiment three

[0063] This embodiment proposes a wear balancing architecture based on multi-modal sensor fusion and reinforcement learning decision-making for the special needs of planetary reducers for engineering machinery in extreme heavy load and dust pollution environment. The essential difference lies in that the decision-making module abandons the traditional optimization algorithm and instead adopts a deep Q network (DQN) to realize end-to-end policy learning, solving the problems of difficult manual setting of multi-objective weights and complex and variable constraint conditions.

[0064] Engineering machinery reducers often work in harsh environments such as mines and tunnels, with strong randomness and suddenness of load spectrum, and long maintenance cycle, requiring the control system to have a high degree of autonomous decision-making capability. Traditional model-based optimization methods rely on accurate parameter calibration, and their performance declines sharply under the influence of uncertain factors such as oil deterioration and increased wear. Therefore, this embodiment constructs an intelligent decision-making framework centered on experience accumulation.

[0065] The perception module adds two new types of sensors on the original basis: one is a fiber Bragg grating (FBG) strain sensor embedded in the gear root to directly measure local stress changes with a sampling frequency of 5 kHz and an accuracy of ±5με; the other is an oil particle counter installed on the oil return line to monitor the ISO 4406 pollution level in real time, outputting data once a minute. These two types of signals are fused with the original data to form an expanded working condition state vector (1×32 dimension), which better reflects the internal damage evolution trend.

[0066] The multi-field coupling modeling module still uses the high-fidelity physical model of embodiment one, but its role has changed: it is no longer used as a basis for real-time decision-making, but as a digital twin environment for training reinforcement learning agents. Before shipment, the system uses historical operation data and virtual simulation to generate millions of working condition segments, covering dozens of degradation modes such as normal wear, poor lubrication, and local peeling, and runs complete four-field coupling solutions on high-performance servers to generate a "state-action-reward" triplet database. The state is the working condition vector, the action is the control instruction combination, and the reward is defined as:

[0067]

[0068] wherein, is the reward value at time step , and is the allowable temperature limit, For reference temperature difference, preset upper limit of safety temperature difference, used for normalization, For rated efficiency, design efficiency under ideal working condition, For maximum unit temperature, For adjacent area temperature difference, For efficiency, For execution cost (such as energy consumption, mechanical wear increment), To For task weight. The reward function inherently encodes the design philosophy of "loss balance".

[0069] The deep Q network adopts a double-flow convolutional structure: one branch processes the numerical type working condition vector (after standardization, input 4-layer fully connected layer), and the other branch processes the vibration spectrum graph (input 3-layer convolutional layer), and finally the features are spliced into the value evaluation layer. The action space is discretized into 36 combinations (6 gears of lubricating oil pump x 3 gears of cooling fan x 2 kinds of lubrication logic x 3 kinds of motor mode). The training process uses experience replay and target network freezing technology to ensure stable convergence. After training is completed, the network parameters are fixed to the vehicle-mounted AI chip (Horizon Journey 5), and the inference delay is less than 5ms.

[0070] When running online, the system no longer solves the optimization problem, but directly outputs the optimal action according to the current state by DQN. The state input network is collected every 5 seconds, the action instruction with the highest Q value is obtained and executed. At the same time, the newly generated (s, a, r, s') sequence is recorded and uploaded to the cloud for model retraining regularly, realizing group knowledge sharing.

[0071] The data feedback update module adds a health state evaluation function. Based on the peak-to-peak value and skewness index of the FBG stress signal, combined with the particle count growth rate, a wear degree fuzzy evaluation model is established to output a health index HI ∈ [0, 1]. When HI < 0.6, automatically trigger the model retraining request, and push the preventive maintenance suggestion to the operation and maintenance personnel. This mechanism not only enables the system to regulate loss, but also predicts the end of life, realizing the transition from "passive balance" to "active health management".

[0072] The above is only a preferred embodiment of the present application, not any form of limitation on the present application, although the present application has been disclosed as above with a preferred embodiment, however, it is not intended to limit the present application, any person skilled in the art, within the scope of the technical solution of the present application, can make some changes or modifications to the above disclosed technical content to make equivalent embodiments with equivalent changes, but as long as it does not deviate from the technical solution of the present application, any simplification, modification, equivalent change and modification of the above embodiments according to the technical essence of the present application, all still belong to the scope of the technical solution of the present application.

Claims

1. A method for multi-field loss equalization of a reducer based on adaptive feedback under operating conditions, characterized in that, include: Multi-dimensional operating condition data during the operation of the reducer are collected in real time through a distributed sensor network. The multi-dimensional operating condition data includes mechanical parameters, thermal parameters, fluid parameters, and electromagnetic parameters. The multidimensional operating condition data is subjected to time synchronization and noise suppression processing to generate a standardized operating condition state vector. The operating condition vector is input into the multi-field coupled physical model, and combined with the data-driven correction term, the multi-field loss density distribution of the key components of the reducer in the spatial and temporal dimensions is obtained. Based on the multi-field loss density distribution, a multi-objective optimization problem with loss balance as the core orientation is constructed, and the optimal control instruction set is obtained by solving the problem. The objective function of the multi-objective optimization problem comprehensively considers the suppression of maximum single-point loss, the smoothness of regional cumulative loss, and the improvement of the overall system efficiency. The optimal control instruction set is sent to the actuator to adjust the lubricating oil supply strategy, cooling intensity, active lubrication logic or motor operating point, so as to realize the dynamic reconstruction of the multi-physics field distribution inside the reducer. Collect actual operating status data after regulation, extract indirect observable indicators and compare them with the multi-field loss density distribution results predicted by the model, calculate the deviation residual, and when the deviation residual exceeds the preset tolerance range, start the online update mechanism of model parameters to correct the multi-field coupled physical model, and complete the self-correction and iterative evolution of the model. The loss balancing decision module receives the loss distribution results from the modeling module and constructs a multi-objective optimization problem with loss balancing as its core. The objective function is defined as: in For finite element mesh element index, For the first Instantaneous power loss of each unit; The system is divided into 9 pre-defined functional areas. The standard deviation of the cumulative loss in the region reflects the risk of thermal stress gradient. The overall system efficiency is defined as the ratio of output power to input power; weighting coefficients. and Dynamically adjust based on operating conditions: Set to [specific settings] during the startup acceleration phase. =1.1, =0.7, focusing on suppressing peak loss; Adjusted to during the stable operation phase =0.9, =0.9, balancing efficiency and uniformity; set to [value] during emergency braking. =1.2, =0.6, to prevent local overheating.

2. The method for multi-field loss equalization of a reducer based on adaptive feedback under operating conditions as described in claim 1, characterized in that, The multi-dimensional operating condition data includes input torque, output speed, housing vibration signal, lubricating oil temperature, lubricating oil pressure, ambient temperature and humidity, and drive motor current and voltage.

3. The method for multi-field loss equalization of a reducer based on adaptive feedback according to claim 1, characterized in that, The multi-dimensional operating condition data undergoes time synchronization and noise suppression processing to generate a standardized operating condition state vector, including: A time-triggered communication protocol is used to timestamp-align all sensor data, with the maximum allowed clock deviation being a preset threshold. An adaptive decomposition method based on empirical wavelet transform is used to suppress noise in different types of signals, preserving effective characteristic frequency bands while removing random and periodic interference. The processed data is encapsulated into a fixed-length data frame containing the original numerical value, unit identifier, and quality label, which serves as the operating condition state vector.

4. The method for multi-field loss equalization of a reducer based on adaptive feedback under operating conditions according to claim 1, characterized in that, The multi-field coupled physical model includes a time-varying model of gear meshing stiffness, a bearing support stiffness model, a dynamic evolution model of lubricating oil film thickness, and an electromagnetic loss distribution model. The multi-field coupled physical model achieves loosely coupled iterative solution of mechanical field, thermal field, fluid field, and electromagnetic field through the transfer of variables at the inter-field interface.

5. The method for multi-field loss equalization of a reducer based on adaptive feedback according to claim 4, characterized in that, The data-driven correction term is composed of a deep neural network. Its input is the current operating condition state vector and the model prediction residual at the previous moment. Its output is the correction amount of the key parameters of each sub-model in the multi-field coupled physical model.

6. The method for multi-field loss equalization of a reducer based on adaptive feedback according to claim 1, characterized in that, The constraints of the multi-objective optimization problem include the physical limits of the actuator, the minimum lubrication guarantee requirements, and the temperature rise safety boundary.

7. The method for multi-field loss equalization of a reducer based on adaptive feedback according to claim 1, characterized in that, The indirect observable indicators include the temperature difference between different sections of the outer wall of the reducer obtained by an infrared thermal imager, and the vibration energy spectrum distribution of a specific frequency band obtained by performing a short-time Fourier transform on the vibration signal of the casing.

8. The method for multi-field loss equalization of a reducer based on adaptive feedback according to claim 1, characterized in that, The online parameter update mechanism of the model adopts the recursive least squares method combined with the forgetting factor to estimate and update the lubricating oil viscosity-temperature coefficient, heat dissipation coefficient and contact thermal resistance parameters in real time.

9. A multi-field loss equalization system for a reducer based on adaptive feedback under operating conditions, employing the multi-field loss equalization method for a reducer based on adaptive feedback under operating conditions as described in any one of claims 1 to 8, characterized in that, include: The real-time operating condition perception module is used to collect multi-dimensional operating condition data during the operation of the reducer, and to perform time synchronization and preprocessing on the multi-dimensional operating condition data to generate a standardized operating condition state vector. The multi-field coupling modeling module is used to receive the working condition state vector, construct a multi-field coupling physical model based on a preset physical mechanism model and a data-driven correction mechanism, and solve the multi-field loss density distribution of the key components of the reducer. The loss equalization decision module is used to receive the multi-field loss density distribution, combine it with a preset global objective function oriented towards loss equalization, and use a multi-objective weighted optimization algorithm to generate the optimal control instruction set. An adaptive execution module is used to receive the optimal control instruction set and adjust at least one of the lubricating oil pump, cooling fan, active lubrication nozzle, clutch or motor to achieve closed-loop intervention on the multi-physical field distribution inside the reducer. The data feedback update module is used to collect actual operating status data after regulation, compare the predicted multi-field loss density distribution with the measured indirect observable indicators, calculate the deviation residual and use it to correct the parameters of the multi-field coupled physical model online, forming a complete feedback link from perception to execution to model evolution.

10. The multi-field loss equalization system for a reducer based on adaptive feedback according to claim 9, characterized in that, The real-time operating condition sensing module includes a high-precision torque sensor, an optical encoder, a triaxial accelerometer array, a distributed temperature probe, a piezoresistive pressure transmitter, and a meteorological environmental sensor; the data feedback and update module includes an infrared thermal imager deployed on the observation window of the reducer's outer wall and a spectrum analysis unit for processing vibration signals.

Images