Temperature drift self-compensation system of torque sensor

Abstract

The invention discloses a temperature drift self-compensation system of a torque sensor, and belongs to the technical field of sensor measurement and signal compensation. The system mainly solves the technical problem that the measurement precision of the torque sensor is reduced due to the change of environment temperature. The system comprises a multi-dimensional signal sensing module, an intelligent fusion processing module, a self-calibration management module and an integrated output interface module, and torque and temperature distribution signals are synchronously collected through a multi-sensor array arranged on a torque sensitive element; and carrying out real-time fusion calculation and nonlinear correction on the temperature drift error by using an embedded multivariable compensation algorithm, optimizing compensation parameters by combining online learning and offline calibration, and finally outputting standardized anti-interference torque data through high-density integrated packaging. According to the invention, dynamic self-compensation of temperature drift is realized, and the measurement precision and long-term stability of the torque sensor in a full temperature range are significantly improved.

Description

Technical Field

[0001] This invention discloses a temperature drift self-compensation system for a torque sensor, belonging to the field of sensor measurement and signal compensation technology. Background Technology

[0002] In existing technologies, torque sensors are susceptible to environmental temperature changes in practical applications, resulting in significant temperature drift, which is one of the core issues leading to decreased measurement accuracy. Current common hardware compensation methods rely on placing one or a few temperature sensing elements inside or near the sensor, measuring the temperature at a single point to infer the temperature state of the entire sensitive area. This method has inherent limitations: the temperature field of the torque-sensitive structure in actual operation is often non-uniformly distributed, and the temperature data from a single measurement point cannot accurately reflect the true thermal gradient causing the drift. This results in an inaccurate compensation model, leading to limited or even failed compensation under complex or rapidly changing temperature conditions.

[0003] Furthermore, most existing compensation systems employ relatively fixed compensation coefficients, which are typically calibrated at limited temperature points before shipment. This method struggles to adapt to changes in sensor drift characteristics caused by material aging, stress relaxation, and variations in installation environments during long-term use. While some technologies introduce online calibration, they often lack systematic self-learning and adaptive mechanisms, failing to maintain high accuracy throughout the sensor's entire lifespan. Simultaneously, the separate design of the sensing unit, processing circuitry, and interface not only increases the risk of interference during signal transmission but may also introduce new errors due to mismatched thermal expansion coefficients of various components, impacting the overall reliability, stability, and integration of the system. Therefore, a temperature drift self-compensation system for torque sensors is urgently needed to address these issues. Summary of the Invention

[0004] The purpose of this invention is to provide a temperature drift self-compensation system for a torque sensor. This system synchronously acquires the original torque signal and the spatially distributed temperature field signal through a multi-dimensional signal sensing module. An embedded multivariate compensation algorithm in the intelligent fusion processing module performs real-time fusion and nonlinear correction of the temperature drift error, generating a precise digital torque value. A self-calibration management module dynamically optimizes and updates the compensation algorithm parameters based on an external calibration protocol and an online learning mechanism. An integrated output interface module encapsulates the processed data and system status information and reliably outputs it. This constructs a complete closed loop from signal sensing, intelligent processing, parameter self-optimization to standard output, achieving adaptive and self-sustaining high-precision measurement of the torque sensor across the entire temperature range.

[0005] The objective of this invention can be achieved through the following technical solutions: A temperature drift self-compensation system for a torque sensor includes the following modules: a multi-dimensional signal sensing module, an intelligent fusion processing module, a self-calibration management module, and an integrated output interface module. The multidimensional signal sensing module is used to perform coordinated measurement and signal acquisition of torque-sensitive structures and thermal environment based on the principle of multi-physics coupling, and generate a mixed raw signal set containing uncompensated raw torque signal and distributed temperature distribution signal; the multidimensional signal sensing module includes: a synchronous signal acquisition unit and a front-end signal conditioning unit; The intelligent fusion processing module is used to perform real-time data fusion and nonlinear drift correction based on the mixed original signal set using an embedded multivariable compensation algorithm, and generate a precise torque digital quantity after temperature drift compensation; the intelligent fusion processing module includes: a digital fusion and drift calculation unit and a precise torque synthesis unit; The self-calibration management module is used to refer to the precise torque digital quantity and, according to a preset external calibration protocol, complete the online learning and offline calibration optimization of the embedded multivariable compensation algorithm parameters, and output the updated dynamic compensation model parameter set. The integrated output interface module is used to perform protocol encapsulation and drive output of multi-source data by using the precise torque digital quantity and combining it with high-density system-level encapsulation technology, so as to obtain a standardized anti-interference torque data stream and complete system status information.

[0006] Preferably, the multidimensional signal sensing module includes a synchronization signal pickup unit and a front-end signal conditioning module, specifically: The synchronous signal acquisition unit is used to synchronously acquire the original torque signal and the spatial temperature field distribution by relying on the multi-sensor array arranged at the key temperature change points of the torque-sensitive element, and generate a synchronous original signal set containing the uncompensated torque signal and the multi-channel temperature signal. The front-end signal conditioning unit is used to amplify, filter, and impedance match the weak differential voltage signal and sensor resistance change based on the synchronous original signal set, so as to obtain a conditioned standard signal set that conforms to the input range of the analog-to-digital converter.

[0007] Preferably, the synchronization signal pickup unit includes: By relying on a thin-film platinum resistance temperature sensor and a magnetoelastic torque sensing ring, an integrated measurement of the stress distribution on the surface of the torsion bar and the near-field thermal field is performed, generating a synchronous raw signal set with strong anti-electromagnetic interference characteristics. The thin-film platinum resistance temperature sensor is arranged symmetrically at the edge of the stress-sensitive region of the magnetoelastic torque sensing ring to synchronously capture the radial and axial temperature gradients of the torque-sensitive region.

[0008] Preferably, the front-end signal conditioning unit includes: Based on the characteristics of the original synchronous signal set, an instrumentation amplifier and a multi-stage active filter are used to amplify the weak differential voltage signal with high common-mode rejection ratio and filter out noise in a specific frequency band. The sensor resistance change is excited and measured by a precision constant current source drive and a four-wire measurement method to eliminate the influence of lead resistance temperature drift and obtain a highly stable conditioned standard signal set.

[0009] Preferably, the intelligent fusion processing module includes a digital fusion and drift calculation unit and a precise torque synthesis unit, specifically as follows: The digital fusion and drift calculation unit is used to call the conditioned torque and temperature signals in the mixed original signal set, and perform real-time calculation on the temperature drift error under the current working condition based on the embedded dynamic compensation model to generate the real-time temperature drift compensation amount. The precision torque synthesis unit applies the real-time temperature drift compensation amount to perform subtraction to eliminate drift in the uncompensated original torque signal, and outputs a precise digital torque quantity that has been compensated for across the entire temperature range.

[0010] Preferably, the digital fusion and drift calculation unit includes: The digital fusion and drift calculation unit includes: a model calling unit, a multivariate fusion unit, and a compensation quantity generation unit; The model calling unit is used to access non-volatile memory, load and initialize pre-stored dynamic compensation model parameters, and prepare a real-time calculation environment. The multivariate fusion unit is used to simultaneously read in the conditioned torque signal and multi-channel temperature signal, and perform real-time calculation based on polynomial regression on the nonlinear relationship between multiple input variables and the target drift amount. The compensation amount generation unit is used to calculate and output the comprehensive temperature drift error amount under the current operating condition based on the real-time calculated nonlinear relationship, and generate the real-time temperature drift compensation amount.

[0011] Preferably, the precise torque synthesis unit includes: The precise torque synthesis unit includes: a drift elimination unit, a dimensional restoration and standardization unit, and a final output processing unit; The drift elimination unit is used to apply the real-time temperature drift compensation to perform algebraic subtraction on the uncompensated original torque digital signal to initially eliminate the system error introduced by temperature. The dimensional restoration and standardization unit is used to perform physical dimensional restoration and normalization processing on the drift-eliminated torque digital signal according to the sensitivity coefficient calibrated by the sensor at the factory, to obtain a standard physical quantity value. The final output processing unit is used to perform amplitude limiting filtering and synchronous control of the output refresh rate on the standard physical quantity value to generate a stable and continuous accurate torque digital quantity.

[0012] Preferably, the self-calibration management module includes online learning and offline calibration optimization of the embedded multivariable compensation algorithm parameters by referring to the precise torque digital quantity and according to a preset external calibration protocol, specifically as follows: Based on preset external calibration instructions and temperature control environment, the raw output of the sensor is automatically collected at multiple stable temperature points to generate an offline calibration dataset covering the entire temperature range; relying on real-time identification of zero torque and stable load conditions, the residual of the precise torque digital quantity is called to perform small-step iterative optimization on the key parameters of the compensation model, and the online adaptive update of the model parameters is completed. The offline calibration dataset and online update results are processed, and the parameter set of the embedded multivariate compensation algorithm is optimized and verified through fitting algorithm and rationality check. Finally, the verified parameter set is written into non-volatile memory, and the updated dynamic compensation model parameter set is output.

[0013] Preferably, the integrated output interface module includes the use of the precise torque digital quantity combined with high-density system-level packaging technology to perform protocol encapsulation and drive output of multi-source data, specifically as follows: By integrating the precise torque digital quantity, key temperature sensing unit data and system self-test status, the multi-source information is formatted, timestamp synchronized and cyclic redundancy check code added to form a standardized data packet with a complete frame header and check structure. Based on the standardized data packets and the preset communication protocol, frame assembly, bit stuffing and physical layer encoding are performed on the data payload and protocol control information to generate protocol data frames that conform to the industrial bus standard. Relying on the physical layer driver chip and protection circuit integrated in a high-density system-in-package, the protocol data frame is subjected to level conversion, differential driving and real-time transmission, so as to complete the reliable external transmission of the standardized anti-interference torque data stream and complete system status information.

[0014] The beneficial effects of this invention are: This invention achieves high-precision suppression of drift caused by complex temperature fields through distributed temperature sensing and multivariate fusion compensation. The system relies on a spatially arranged multi-point temperature sensor array to synchronously capture the non-uniform temperature gradient distribution in the torque-sensitive area. Based on this, an embedded intelligent algorithm is used to construct a dynamic compensation model with multiple input variables. This model can accurately calculate and compensate for measurement errors caused by thermal gradients and nonlinear temperature changes in real time, thereby significantly improving the accuracy and stability of torque measurement across the entire temperature range and under rapid temperature change conditions.

[0015] This invention leverages online self-learning and system-level integrated design to enhance long-term reliability and maintenance-free operation. Through a self-calibration management module, the system intelligently identifies operating conditions and automatically optimizes compensation parameters during operation, enabling the sensor to continuously adapt to changes in its own state and environmental drift. Simultaneously, high-density integrated packaging integrates sensing, processing, and interface units, reducing signal link interference and localized thermal stress, improving the system's overall anti-interference capability, structural reliability, and long-term operational consistency, while lowering the need for calibration and maintenance during use. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the temperature drift self-compensation system for a torque sensor according to the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] Example: Figure 1 As shown, a temperature drift self-compensation system for a torque sensor includes the following components: a multi-dimensional signal sensing module, an intelligent fusion processing module, a self-calibration management module, and an integrated output interface module. The multi-dimensional signal sensing module is used to perform coordinated measurement and signal acquisition of torque-sensitive structures and thermal environment based on the principle of multi-physics coupling, and generate a mixed raw signal set containing uncompensated raw torque signal and distributed temperature distribution signal. The intelligent fusion processing module is used to perform real-time data fusion and nonlinear drift correction based on the mixed original signal set using an embedded multivariable compensation algorithm, and generate a precise torque digital quantity after temperature drift compensation. The self-calibration management module is used to refer to the precise torque digital quantity and, according to a preset external calibration protocol, complete the online learning and offline calibration optimization of the embedded multivariable compensation algorithm parameters, and output the updated dynamic compensation model parameter set. The integrated output interface module is used to perform protocol encapsulation and drive output of multi-source data by using the precise torque digital quantity and combining it with high-density system-level encapsulation technology, so as to obtain a standardized anti-interference torque data stream and complete system status information.

[0019] In this embodiment, based on the principle of multiphysics coupling, collaborative measurement and signal acquisition of the torque-sensitive structure and the thermal environment are performed to generate a mixed raw signal set containing uncompensated raw torque signal and distributed temperature distribution signal. The specific implementation method is as follows: The implementation of the multidimensional signal sensing module focuses on overcoming the limitations of traditional single-point temperature measurement. Through innovative sensor array layout and high-fidelity signal chain design, it achieves synchronous capture of the spatial temperature field distribution in the torque-sensitive area and correlates it with the original torque signal. This outputs a set of mixed original signals that can comprehensively and accurately reflect the current thermodynamic state of the sensor, providing an unprecedented high-dimensional data foundation for subsequent intelligent compensation.

[0020] The synchronization signal acquisition unit is the core of this physical information conversion process. This embodiment employs an integrated design scheme of a heterogeneous micro-sensor array. The array is based on a non-contact magnetoelastic torque sensing ring, which directly senses the shear stress field generated on the surface of the torsion bar due to the transmitted torque. This sensing ring outputs two differential induced voltage signals. and Theoretical analysis and calibration show that the measured torque T has a definite functional relationship with the normalized difference between the two signals, and its core expression can be characterized as: ,in, It is a proportionality coefficient related to the properties of the magnetoelastic sensitive material, the number of coil turns, and the structural dimensions. This formula reveals the fundamental principle of torque measurement, but its... ,even and Each of them will drift with temperature changes.

[0021] To accurately characterize the temperature field causing the aforementioned drift, this unit innovatively mounts four high-precision thin-film platinum resistance temperature sensors symmetrically and securely at 90-degree circumferential intervals along the edge of the stress-sensitive area of ​​the magnetoelastic sensing ring. This symmetrical arrangement is not random; its purpose is to actively measure the temperature gradient rather than a single-point absolute temperature. Specifically, by measuring the resistance of the platinum resistors at diagonal positions, an approximate temperature gradient along the radial direction of the torsion bar can be calculated. Radial; by comparing the average temperature of platinum resistance thermometers at different axial positions, the axial temperature gradient can be obtained. Axial direction. This gradient information is key to distinguishing between uniform temperature rise and non-uniform localized heating, and the non-uniform thermal field is the main cause of additional thermal stress interference and complex nonlinear drift. Therefore, the original synchronous signal set output by this unit is a six-element vector: Each platinum resistance value Rather than the temperature of the patch The relationship follows the international standard function, and its quadratic approximation expression is: In the formula, The nominal resistance of the platinum resistance thermometer at 0°C is 100.00Ω. and These are the first and second order temperature resistivity of platinum, respectively. This formula is the basis for converting resistance measurements into precise temperature values.

[0022] The front-end signal conditioning unit converts the aforementioned weak and easily contaminated raw analog signal into a stable, clean standard signal with amplitude adapted to the digital system. This unit consists of two parallel and independently optimized high-precision analog signal channels. This embodiment uses a high common-mode rejection ratio instrumentation amplifier as the input stage, and its output... The relationship with the input is ,in The gain is programmable. The signal passes through a precisely designed multi-stage active bandpass filter with a cutoff frequency, designed to retain all effective spectral components of the torque signal while completely filtering out high-frequency switching noise and extremely low-frequency drift disturbances.

[0023] For four temperature signal channels, the key to conditioning accuracy lies in achieving error-free resistance-to-voltage conversion and suppressing temperature drift in the measurement link itself. This embodiment configures each platinum resistance thermometer with an independent, bandgap-referenced precision constant current source to provide a highly stable excitation current. To completely eliminate the resistance of the connecting wires To mitigate the systematic errors caused by temperature drift, a strict four-wire Kelvin measurement method was employed: two "force lines" provided constant current. The other two "sensing wires" directly measure the voltage drop across the platinum resistance solder joint. Therefore, the resistance value of a platinum resistance thermometer can be accurately calculated using the following formula, and it is independent of the resistance of the wires: This voltage signal It is then fed into a low-drift, low-noise operational amplifier for moderate amplification, and passes through an anti-aliasing low-pass filter to finally form a standardized voltage signal that corresponds one-to-one with temperature.

[0024] Through the above specific implementation, the multi-dimensional signal sensing module successfully materialized the abstract system error of "temperature drift" into a multi-channel electrical signal dataset that can be synchronously sampled and has clear physical meaning. This module not only provides the original torque signal to be compensated... , More importantly, it acquires spatial temperature gradient information through a distributed array. radial, Axial axis, and the absolute temperature of each measuring point pass Precise quantification. This information collectively constitutes the complete input vector necessary for subsequent intelligent fusion processing algorithms to perform multivariate and nonlinear compensation, ensuring from the data source that the entire self-compensation system can approach its theoretical optimal performance.

[0025] In this embodiment, based on the mixed original signal set, an embedded multivariate compensation algorithm is used to perform real-time data fusion and nonlinear drift correction to generate a precise digital torque quantity after temperature drift compensation. The specific implementation method is as follows: The core of this module lies in its internally running embedded multivariate compensation algorithm. This algorithm is not a simple linear offset correction, but an intelligent mathematical model that can dynamically fuse the original torque value, multi-point absolute temperature, and spatial temperature gradient information, and solve complex nonlinear drift errors in real time. Its implementation goal is to establish a digital twin corrector that can accurately predict and offset all systematic output deviations of physical sensors caused by temperature changes, thereby outputting a precise digital quantity that is independent of temperature and reflects only the true mechanical torque.

[0026] The digital fusion and drift calculation unit is the computational engine of this algorithm model. Its implementation begins with the model calling unit loading a pre-generated optimal compensation model parameter set from non-volatile memory. This parameter set is obtained before the sensor leaves the factory by applying known torque over a wide temperature range and collecting massive amounts of data, then training it using an advanced fitting algorithm. The model itself is a multivariate function with temperature as its core variable. A basic but effective implementation is an extended multinomial regression model. This model considers not only individual temperature measurement points... The absolute temperature is used, and more importantly, a gradient term reflecting the non-uniformity of the temperature field is introduced. It calculates the temperature drift compensation amount under the current operating conditions. The core formula can be expressed as: in, The absolute temperature measured by the i-th platinum resistance thermometer can be obtained through... The inverse solution yields the result; radial and The axial gradient is calculated from symmetrically arranged temperature sensors. It is the normalized value of the uncompensated original torque voltage signal; It is the average temperature; These are the model parameters loaded from memory. The formula introduces... , The cross term is designed to compensate for the nonlinear characteristics of torque signal sensitivity as it changes with temperature, which is key to achieving high-precision compensation across the entire temperature range.

[0027] The multivariate fusion unit is responsible for performing the above model calculations in real time. It synchronously reads in the signals after they have been digitized by the ADC: the normalized raw torque and voltage digital values, and the temperature digital values ​​for each channel. This unit runs in the embedded processor, cyclically calculating the above polynomials at a fixed sampling period. The compensation quantity generation unit is responsible for outputting the calculation results. This value is a digital quantity, and its physical meaning is equivalent to the error value that the torque reading should be corrected for under the current temperature field. The unit is the same as that of torque.

[0028] The precision torque synthesis unit receives the raw torque digital signal from the front end. and compensation amount from the drift calculation unit The final correction and standardization are then implemented. The drift elimination unit performs the most direct algebraic operations, and its core formula is extremely simple yet crucial: This step, in principle, directly eliminates the systematic error introduced by temperature. However, at this point... It remains an internal digital quantity related to the sensor's specific sensitivity and ADC gain. The role of the dimensional reduction and normalization unit is to convert it into an engineering value with universal physical meaning. It calls the sensor calibration sensitivity coefficient S stored in memory and performs the reduction using the following formula: Where S itself may be a slowly varying function with temperature, this calibration is performed and updated by the self-calibration management module. Final output processing unit. Apply necessary post-processing. This includes using moving average filtering and median filtering based on historical data to suppress random noise; output limiting to prevent outliers caused by transient interference; and latching and buffering the final result according to the preset communication protocol refresh rate to ensure that the output to the interface module is a series of stable, continuous, and time-accurate precise torque digital quantities.

[0029] In this embodiment, referring to the precise digital torque quantity and according to a preset external calibration protocol, the online learning and offline calibration optimization of the embedded multivariable compensation algorithm parameters are completed, and the updated dynamic compensation model parameter set is output. The specific implementation method is as follows: The self-calibration management module works collaboratively through offline calibration and online learning modes to jointly construct and continuously maintain the core of the embedded multivariate compensation algorithm in the intelligent fusion processing module—the dynamic compensation model parameter set. Its fundamental purpose is to overcome the time-varying drift characteristics of sensors caused by individual manufacturing differences, natural material aging, and long-term stress relaxation. This allows the system to not only achieve high accuracy at the time of manufacture but also autonomously adapt to slow changes throughout its entire lifecycle, achieving "self-sustaining" accuracy. The core of its implementation lies in designing a complete and automated data-driven parameter update pipeline. This pipeline uses precise measurement data as input, optimizes algorithms, and ultimately produces validated and reliable model parameters.

[0030] Offline calibration is a crucial process for establishing an initial high-precision compensation model for sensors and performing periodic comprehensive recalibration in a controlled, precision environment. Its implementation relies on an automated platform comprised of a high- and low-temperature control chamber, a standard torque calibrator, and host computer calibration software. The specific process is as follows: the host computer issues a sequence of calibration commands through the system's dedicated calibration interface. Upon receiving the commands, the self-calibration management module first controls the sensor to enter calibration mode. Subsequently, the control chamber, according to the commands, calibrates from low to high temperatures at multiple preset stable temperature points. The process is repeated cyclically. After thermal equilibrium is reached at each temperature point, the calibrator applies a series of known standard torque values. The module then synchronously acquires and records all the corresponding raw signals at this time: including the raw digital torque value without any compensation. and readings from all temperature sensors. This generates a structured offline calibration dataset covering the entire operating temperature range and torque measurement range. After obtaining the dataset, the module calls the built-in batch parameter fitting algorithm. Taking the aforementioned polynomial compensation model as an example, its goal is to find a set of parameters. This ensures that, for all calibration data, the model predicts the compensated torque. With standard torque This minimizes the error. This is typically transformed into a least squares problem with the objective function: ,in This is the formula for calculating the compensation amount executed by the drift calculation unit. By solving this normal equation, the optimal set of initial parameters can be obtained. .

[0031] Online learning refers to the system's ability to perform seamless fine-tuning under specific operating conditions in a real-world environment, aiming to address slow time drift and environmental differences. Its implementation relies on intelligent perception of operating conditions. An embedded operating condition identifier continuously monitors torque output, speed, and signal noise characteristics. When a clear "zero torque" state and "known stable load" state are identified, it is considered a valid learning window. In the zero torque state, theoretically, the precise torque digital value... It should always be zero. Therefore, any non-zero output can be considered as the comprehensive residual under the current temperature field. This residual E includes the errors that the model failed to fully compensate for, as well as newly generated drift. Online learning algorithms use this residual to update the model parameters in small, incremental steps. A classic and effective implementation is to use the recursive least squares method. Its core iterative formula can be simplified as follows: , , In this set of formulas, It is the parameter vector estimate at time t; It is based on the current temperature at time t. The calculated regression vector is the vector composed of the variable terms multiplied by the parameters in the formula; It is the residual currently observed; It is the gain matrix; It is the covariance matrix; This is the forgetting factor, where 0 < λ ≤ 1, used to assign higher weights to new data, enabling the algorithm to track slow parameter changes. Through this iteration, the model parameters are continuously fine-tuned, gradually approaching the optimal state of the current sensor.

[0032] Parameter processing, validation, and persistence are the final hurdles to ensure the safety and reliability of the learning process. This applies to new parameters obtained through offline fitting. Or are the parameters gradually updated through online learning? Before being written to the non-volatile memory that determines the system's behavior, all parameters must undergo rigorous validity checks. These checks include: whether the parameter values ​​are within a preset physical reasonable range; whether the fitting residuals of the updated model on historical datasets and the latest data segment have significantly decreased; and whether the step size of the online learning update is too drastic. After successful validation, the module performs an atomic write operation, safely replacing the old parameter set with the updated dynamic compensation model parameter set. Simultaneously, the module updates the parameter version number and reports this version information as part of the system status through the integrated output interface module, thus forming a complete closed loop from data acquisition, optimization calculation, security verification to persistent storage and version management.

[0033] In this embodiment, the precise digital torque quantity is used in conjunction with high-density system-level packaging technology to perform protocol encapsulation and drive output of multi-source data, thereby obtaining a standardized anti-interference torque data stream and complete system status information. The specific implementation method is as follows: The integrated output interface module receives precise digital torque values ​​from the intelligent fusion processing module and integrates other key system status information. Through a complete processing chain from data encapsulation and protocol construction to physical drive, it transforms this information into a standardized data stream capable of stable transmission in complex industrial electromagnetic environments. This module deeply integrates high-density system-in-package technology with an industrial-grade communication protocol stack. Its core objective is to ensure that the high-precision torque value, obtained through complex compensation, is delivered to the host controller with minimal distortion, maximum reliability, and deterministic timing, thereby truly realizing its measurement value.

[0034] Multi-source information integration and standardized data packet construction are the first steps in the data processing of this module. The key to its implementation lies in creating a unified, self-describing data structure. Internally, the module maintains an application-layer data object, which is continuously updated and contains the following core fields: 32-bit floating-point precise torque digital quantity, unit N·m. 16-bit integer, processor core temperature, to : 16-bit integer, measured values ​​of each platinum resistance temperature sensor, A 16-bit status word containing self-calibration activation flags, model parameter versions, power supply voltage monitoring flags, and hardware self-test results. A 32-bit unsigned incrementing sequence number is used to detect packet loss. In each fixed output cycle, the module performs data packetization. It generates a precise timestamp for the data object, derived from the processor's hardware timer, ensuring the data has an accurate time base. Subsequently, it serializes all fields according to a preset byte order, generating a raw data payload. To ensure data integrity during transmission, the module must calculate a cyclic redundancy check (CRC) code for this payload. This embodiment uses the widely adopted CRC-16-CCITT standard. The calculation process can be briefly described as follows: the raw data payload is treated as a binary polynomial with the most significant bit first. Combine it with the generator polynomial The remainder polynomial obtained by performing modulo 2 division. The coefficient is the 2-byte CRC value. This operation is performed in real time by a hardware CRC accelerator and an optimized software algorithm, and can be expressed as: The module adds a fixed frame to the header of the payload and appends a calculated CRC value to the tail, thereby assembling a complete standardized data packet with strong error detection capabilities.

[0035] Industrial protocol frame assembly and physical layer encoding are crucial steps in adapting standardized data packets to specific industrial networks. The module integrates a CAN FD controller IP core. Its workflow is as follows: the standardized data packets obtained in the previous step are loaded as a data field into the controller's transmit buffer. A configuration identifier is set; this ID defines the logical address and message priority of the sensor. For CAN FD frames, the data field length and a higher data transmission bit rate must also be set. Before the controller initiates transmission, the module's protocol processing logic automatically performs bit stuffing according to the CAN FD specification: that is, after five consecutive bits of the same polarity, an inverse polarity bit is automatically inserted. This technique ensures sufficient level transitions, facilitating clock synchronization by the receiving node and enhancing anti-interference capabilities. The entire bit stream to be transmitted, including the frame start, arbitration field, control field, data field, CRC field, acknowledgment field, and frame end, constitutes a complete protocol data frame. From an information theory perspective, the raw torque information... Through this process, it is encoded into a signal sequence with specific physical meaning and temporal structure. Its reliability is far higher than that of directly transmitting analog voltage.

[0036] High-density integrated packaging and physical layer driven transmission are the hardware foundation for realizing the aforementioned communication functions and also the physical embodiment of the anti-interference design of this module. This embodiment employs system-level packaging technology to integrate heterogeneous components such as the magnetoelastic sensing ring, analog conditioning chip, microprocessor (including all aforementioned algorithm modules), CAN FD controller, and physical layer transceiver into a compact metal housing only slightly larger than the sensing ring itself, using embedded substrate technology. This high degree of integration brings multiple advantages: extremely short high-frequency signal traces significantly reduce spatial radiation and lower sensitivity to interference; the unified metal housing provides excellent electromagnetic shielding and heat dissipation paths; and the power supplies for sensor signals and communication signals are precisely isolated and filtered through internal multi-layer boards. The physical layer transceiver is responsible for transmitting the digital bit stream output by the controller. This is converted into a differential voltage signal capable of long-distance transmission over twisted-pair cables. It employs differential drive, mapping logic "0" and "1" to a defined voltage difference between CAN_H and CAN_L, such as a dominant bit corresponding to approximately 2V and a recessive bit corresponding to approximately 0V. The drive process can be abstracted as a controlled voltage source model. Simultaneously, the transceiver integrates protection circuitry, such as electrostatic discharge protection diodes, common-mode chokes, and transient voltage suppressors, to jointly resist various surges and pulse interferences on the bus. In each output cycle, a pair of differential electrical signals, representing precise torque, multiple temperature readings, and the overall health status of the system, and rigorously protected, is reliably emitted from the high-density integrated package of this sensor, forming the standardized, interference-resistant torque data stream.

[0037] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0038] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A temperature drift self-compensation system for a torque sensor, characterized in that, It includes the following modules: multi-dimensional signal sensing module, intelligent fusion processing module, self-calibration management module, and integrated output interface module: The multi-dimensional signal sensing module is used to perform coordinated measurement and signal acquisition of torque-sensitive structures and thermal environment based on the principle of multi-physics coupling, and generate a mixed raw signal set containing uncompensated raw torque signal and distributed temperature distribution signal. The multidimensional signal sensing module includes: a synchronous signal pickup unit and a front-end signal conditioning unit; The intelligent fusion processing module is used to perform real-time data fusion and nonlinear drift correction based on the mixed original signal set using an embedded multivariable compensation algorithm, and generate a precise torque digital quantity after temperature drift compensation; the intelligent fusion processing module includes: a digital fusion and drift calculation unit and a precise torque synthesis unit; The self-calibration management module is used to refer to the precise torque digital quantity and, according to a preset external calibration protocol, complete the online learning and offline calibration optimization of the embedded multivariable compensation algorithm parameters, and output the updated dynamic compensation model parameter set. The integrated output interface module is used to perform protocol encapsulation and drive output of multi-source data by using the precise torque digital quantity and combining it with high-density system-level encapsulation technology, so as to obtain a standardized anti-interference torque data stream and complete system status information.

2. The temperature drift self-compensation system for a torque sensor according to claim 1, characterized in that, The multidimensional signal sensing module includes a synchronization signal acquisition unit and a front-end signal conditioning module, specifically: The synchronous signal acquisition unit is used to synchronously acquire the original torque signal and the spatial temperature field distribution by relying on the multi-sensor array arranged at the key temperature change points of the torque-sensitive element, and generate a synchronous original signal set containing the uncompensated torque signal and the multi-channel temperature signal. The front-end signal conditioning unit is used to amplify, filter, and impedance match the weak differential voltage signal and sensor resistance change based on the synchronous original signal set, so as to obtain a conditioned standard signal set that conforms to the input range of the analog-to-digital converter.

3. The temperature drift self-compensation system for a torque sensor according to claim 2, characterized in that, The synchronization signal pickup unit includes: By relying on a thin-film platinum resistance temperature sensor and a magnetoelastic torque sensing ring, an integrated measurement of the stress distribution on the surface of the torsion bar and the near-field thermal field is performed, generating a synchronous raw signal set with strong anti-electromagnetic interference characteristics. The thin-film platinum resistance temperature sensor is arranged symmetrically at the edge of the stress-sensitive region of the magnetoelastic torque sensing ring to synchronously capture the radial and axial temperature gradients of the torque-sensitive region.

4. The temperature drift self-compensation system for a torque sensor according to claim 2, characterized in that, The front-end signal conditioning unit includes: Based on the characteristics of the original synchronous signal set, an instrumentation amplifier and a multi-stage active filter are used to amplify the weak differential voltage signal with high common-mode rejection ratio and filter out noise in a specific frequency band. The sensor resistance change is excited and measured by a precision constant current source drive and a four-wire measurement method to eliminate the influence of lead resistance temperature drift and obtain a highly stable conditioned standard signal set.

5. The temperature drift self-compensation system for a torque sensor according to claim 1, characterized in that, The intelligent fusion processing module includes a digital fusion and drift calculation unit and a precise torque synthesis unit, specifically as follows: The digital fusion and drift calculation unit is used to call the conditioned torque and temperature signals in the mixed original signal set, and perform real-time calculation on the temperature drift error under the current working condition based on the embedded dynamic compensation model to generate the real-time temperature drift compensation amount. The precision torque synthesis unit applies the real-time temperature drift compensation amount to perform subtraction to eliminate drift in the uncompensated original torque signal, and outputs a precise digital torque quantity that has been compensated for across the entire temperature range.

6. The temperature drift self-compensation system for a torque sensor according to claim 5, characterized in that, The digital fusion and drift calculation unit includes: The digital fusion and drift calculation unit includes: a model calling unit, a multivariate fusion unit, and a compensation quantity generation unit; The model calling unit is used to access non-volatile memory, load and initialize pre-stored dynamic compensation model parameters, and prepare a real-time calculation environment. The multivariate fusion unit is used to simultaneously read in the conditioned torque signal and multi-channel temperature signal, and perform real-time calculation based on polynomial regression on the nonlinear relationship between multiple input variables and the target drift amount. The compensation amount generation unit is used to calculate and output the comprehensive temperature drift error amount under the current operating condition based on the real-time calculated nonlinear relationship, and generate the real-time temperature drift compensation amount.

7. The temperature drift self-compensation system for a torque sensor according to claim 5, characterized in that, The precise torque synthesis unit includes: The precise torque synthesis unit includes: a drift elimination unit, a dimensional restoration and standardization unit, and a final output processing unit; The drift elimination unit is used to apply the real-time temperature drift compensation to perform algebraic subtraction on the uncompensated original torque digital signal to initially eliminate the system error introduced by temperature. The dimensional restoration and standardization unit is used to perform physical dimensional restoration and normalization processing on the drift-eliminated torque digital signal according to the sensitivity coefficient calibrated by the sensor at the factory, to obtain a standard physical quantity value. The final output processing unit is used to perform amplitude limiting filtering and synchronous control of the output refresh rate on the standard physical quantity value to generate a stable and continuous accurate torque digital quantity.

8. The temperature drift self-compensation system for a torque sensor according to claim 1, characterized in that, The self-calibration management module includes online learning and offline calibration optimization of the embedded multivariable compensation algorithm parameters by referencing the precise torque digital quantity and according to a preset external calibration protocol. Specifically, it includes: Based on preset external calibration instructions and temperature control environment, the raw output of the sensor is automatically collected at multiple stable temperature points to generate an offline calibration dataset covering the entire temperature range; relying on real-time identification of zero torque and stable load conditions, the residual of the precise torque digital quantity is called to perform small-step iterative optimization on the key parameters of the compensation model to complete the online adaptive update of the model parameters. The offline calibration dataset and online update results are processed, and the parameter set of the embedded multivariate compensation algorithm is optimized and verified through fitting algorithm and rationality check. Finally, the verified parameter set is written into non-volatile memory, and the updated dynamic compensation model parameter set is output.

9. The temperature drift self-compensation system for a torque sensor according to claim 1, characterized in that, The integrated output interface module includes the use of precise torque digital quantity and high-density system-level packaging technology to carry out protocol encapsulation and drive output of multi-source data, specifically as follows: By integrating the precise torque digital quantity, key temperature sensing unit data and system self-test status, the multi-source information is formatted, timestamp synchronized and cyclic redundancy check code added to form a standardized data packet with a complete frame header and check structure. Based on the standardized data packet and the preset communication protocol, frame assembly, bit stuffing and physical layer encoding are performed on the data payload and protocol control information to generate a protocol data frame that conforms to the industrial bus standard. Relying on the physical layer driver chip and protection circuit integrated in a high-density system-in-package, the protocol data frame is subjected to level conversion, differential driving and real-time transmission, so as to complete the reliable external transmission of the standardized anti-interference torque data stream and complete system status information.

Images