Inertial measurement unit temperature compensation fault diagnosis method and system

Abstract

This invention discloses a method and system for diagnosing temperature compensation faults in inertial measurement units (IMUs). The method achieves highly reliable fault detection by introducing a redundant temperature source T2 and a functional safety-oriented diagnostic logic. Instead of directly comparing temperature signals, the method substitutes the primary temperature T1 and the redundant temperature T2 into the IMU error compensation model to calculate two sets of compensation results B. compT1 and B compT2 The difference between residual offset Only if the difference is residual offset A functional failure is only identified when the threshold is exceeded by reverse engineering based on application safety requirements. This method effectively solves the shortcomings of traditional diagnostic methods, such as high cost, cold start failure, and high false alarm rate, and meets functional safety requirements.

Description

Technical Field

[0001] This invention belongs to the field of inertial sensor technology, specifically relating to a fault diagnosis method and system for a high-precision inertial measurement unit (IMU). Background Technology

[0002] Inertial measurement units (IMUs), especially microelectromechanical systems (MEMS)-based IMUs, are indispensable components in modern autonomous driving systems, drones, and high-precision industrial equipment. The performance of MEMS sensors, particularly their bias and scale factor, is highly sensitive to operating temperature. To maintain high accuracy over a wide temperature range (e.g., -40°C to +85°C), the IMU must measure its internal temperature (T1) in real time and use sophisticated compensation models to correct for temperature-induced drift.

[0003] Serious consequences of T1 failure: If this critical internal temperature sensor T1 malfunctions, or its heat conduction path changes, the compensation model will input incorrect information, causing the IMU to output incorrect data.

[0004] For example, in automotive dead reckoning (DR), a +20°C temperature signal error could cause a gyroscope compensation error, resulting in a 0.1 dps zero bias. After traveling 80 kph over 60 seconds, this could lead to a lateral positioning error of over 40 meters. Similarly, it could cause an accelerometer zero bias of 3 mg, resulting in a longitudinal positioning error of approximately 50 meters.

[0005] In autonomous driving scenarios, this level of error is unacceptable, therefore functional safety standards must be met. Background of functional safety standard (ISO 26262): The automotive functional safety standard ISO 26262 introduces the concept of ASIL (Automotive Safety Integrity Level) to assess the potential risks posed by system failure. The ASIL level is determined based on the following three factors: Severity (S): The consequences of failure (e.g., whether it causes personal injury).

[0006] Exposure (E): The probability of a failure scenario occurring.

[0007] Controllability (C): The ability of the driver or system to avoid accidents. The functional safety requirement for Level 2 autonomous driving systems is typically ASIL-B. If the IMU temperature signal, which is a core input, does not meet the ASIL requirement, its failure will directly lead to the final output signal not meeting the ASIL-B requirement, thus posing a potential risk to the entire system.

[0008] Existing technology and its shortcomings: Currently, the mainstream technologies for solving IMU fault tolerance problems have the following drawbacks: 1. System-level hardware redundancy: This is the most traditional high-reliability solution, such as using dual-mode or triple-mode redundancy architectures. This involves installing two or three complete IMU units simultaneously, using a voting mechanism or mutual comparison to detect and isolate faulty units.

[0009] Disadvantages: Cost, size, weight, and power consumption increase exponentially, making it too expensive for cost- and space-constrained automotive or drone applications.

[0010] 2. Pure software / algorithm redundancy: This type of method does not add hardware, but relies on sensor fusion algorithms (such as extended Kalman filter EKF) to detect anomalies by comparing the IMU's solution with the differences from external reference sources (such as GNSS, wheel speedometer, visual odometry).

[0011] Defect 1: The diagnostic method fails when the external reference source (such as GNSS) is lost in a tunnel or urban canyon.

[0012] Defect 2 (critical): Extremely poor performance during the IMU "cold start" phase. External GNSS reference sources take tens of seconds or even minutes (100 to 2000 seconds) to reach a stable state after startup. During this period, algorithms such as EKF have difficulty converging, which can easily lead to false alarms or missed alarms.

[0013] Simple temperature signal comparison (comparing IMU internal temperature T1 and external ambient temperature T2): Drawback: It's too simplistic, and the temperature threshold cannot account for the differences in compensation characteristics between different products. For example, some IMU units have very small temperature compensation corrections, so even if the temperature difference between T1 and T2 is large, the impact on the final output is within an acceptable range. Conversely, some units have very significant temperature compensation corrections, so even a small temperature difference can cause serious deviations in the output. Therefore, simple temperature difference comparisons cannot distinguish between "harmful temperature differences" and "harmless temperature differences," easily leading to false alarms. Summary of the Invention

[0014] The purpose of this invention is to disclose a method and system for diagnosing temperature compensation faults in inertial measurement units (IMUs), particularly a method for diagnosing faults in the temperature compensation process of key IMU parameters (such as zero bias and scaling factor) using component-level redundant temperature information. This method is implemented by introducing component-level redundant temperature sources and employing a functional safety-oriented multi-stage diagnostic algorithm. Addressing the deficiencies in the aforementioned background technology, this invention aims to solve the following technical problems: 1. Achieve reliable diagnosis of critical thermal faults in IMUs at extremely low hardware costs, avoiding the use of multiple sets of IMU hardware redundancy; 2. Provide a diagnostic method that does not rely on external signals (such as GNSS) to ensure effective diagnostic capability during the warm-up phase of IMU startup and when external signals are lost; 3. Provide a diagnostic method that can intelligently distinguish between "normal physical temperature difference" and "abnormal fault temperature difference" to solve the false alarm problem caused by simple temperature comparison; 4. Provide a design that meets functional safety objectives, which diagnoses not the temperature value itself, but whether the erroneous temperature value has caused harm beyond the safety threshold to the final compensation result of the IMU (zero bias and scaling factor compensation) (functional orientation).

[0015] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a method for diagnosing temperature compensation faults in an inertial measurement unit (IMU), used to diagnose whether a fault in the first temperature sensor T1 in the IMU causes functional impairment to the IMU's error compensation results. The method runs on a processing unit and includes the following steps: Step 1: Independent rationality check; obtain the main temperature value T1 provided by the first temperature sensor inside the IMU chip, and the reference temperature value T2 provided by the auxiliary temperature source on the IMU mounting assembly, and check whether the readings of T1 and T2 are within the preset physical operating range (e.g., -40°C to +85°C); Step Two: Dynamic Temperature Difference Comparison Process; a. Calculate the actual difference between T1 and T2, based on the IMU power consumption P. diss and / or ambient reference temperature T amb The expected temperature difference ΔT between T1 and T2 is calculated using a pre-calibrated thermal model. model ; b. Calculate the residual temperature difference. T = |T1-T2| - ΔT model ; c. When residual T ≤ Thresh T When the temperature difference conforms to the multi-order RC network thermal model, it is determined to be fault-free; when the residual T Thresh T At this point, the next step of the functional check is performed; this avoids the false alarms caused by harmless temperature differences resulting from the "simple temperature signal comparison" in Background Art 3. residual T It is the difference ΔT between the actual temperature difference |T1-T2| output by the two temperature sensors and the temperature difference calculated by the temperature model. model The difference is Thresh TThis is a pre-set dynamic temperature tolerance threshold, which can be set according to the sensor's accuracy, for example, ±5℃. If residual T If the value exceeds the threshold, it is determined to be a fault, and the next stage of diagnosis is initiated. If the value is within the threshold range, it is determined to be a non-fault, and the diagnosis ends.

[0016] Step 3: Static functional comparison process (zero bias); a. The T1 and T2 and their respective rates of change d T1 / d t d T2 / d t Substituting each into the pre-calibrated IMU error compensation model, two sets of zero-bias compensation vectors B are calculated. compT1 and B compT2 ; b. Calculate the difference between the two sets of compensation results. offset = || B compT1 - B compT2 ||2; c. The difference residual offset Compared with the preset functional safety threshold Thresh offset Compare; d. When residual offset Thresh offset When this occurs, it is determined to be a static functional fault, indicating that the fault in T1 has caused harm to the final output of the IMU beyond the safety requirements.

[0017] As a further improvement to the present invention, in step three, when residual offset ≤ Thresh offset This also includes a dynamic functional comparison process; a. Substitute T1 and T2 into the complete compensation algorithm (zero bias + scaling factor) (this compensation algorithm is a method for calibrating and compensating the IMU, usually a high-order polynomial or neural network model), based on the IMU's raw readings (Acc). Raw Gyr Raw This yields two sets of final output data Acc. CorT1 and Acc CorT2 and Gyr CorT1 and Gyr CorT2 ; b. Calculate the final output difference: residual acc = ||Acc CorT1 -Acc CorT2 ||2 and residual gyr = ||GyrCorT1 -Gyr CorT2 ||2; c. If residual acc Thresh Acc_Cor or residual gyr Thresh Gyr_Cor If so, it is determined to be a dynamic functional failure.

[0018] As a further improvement of the present invention, in step one, if T1 is out of range, it is determined to be a serious fault; if T2 is out of range, the diagnostic task is terminated.

[0019] As a further improvement of the present invention, in step three, the functional safety threshold Thresh... offset It is derived by reverse reasoning through the error propagation model based on the safety requirements of the application (such as the maximum positioning error requirement in vehicle dead reckoning). For example, in the background technology introduction, in the application of vehicle dead reckoning (DR), the lateral and longitudinal errors of DR reckoning over 1KM are required to be less than 1.5‰, then the corresponding zero bias threshold of the gyroscope is 3.8mdps and the zero bias threshold of the accelerometer is 0.15mg.

[0020] As a further improvement of the present invention, in step two, the thermal model is a description of T1, T2 and IMU power consumption P. diss A multi-order RC thermal network model of dynamic thermal relationships between the components uses several resistors (Rth, representing thermal resistance) and capacitor components (Cth, representing thermal capacity) to simulate the transient and steady-state process of heat conduction from the IMU chip T1 through different paths (such as package, PCB) to the auxiliary temperature source T2.

[0021] As a further improvement of the present invention, the IMU error compensation model is a high-order polynomial or neural network model used to calculate the compensation value B of the IMU zero bias based on temperature and its rate of change. comp Compensation value K of the scaling factor comp .

[0022] The IMU's bias and scale factor are nonlinear temperature functions, and the model is a pre-calibrated function mapping f(T) imu → Bias err &Scale err In this invention, instead of creating or modifying an error compensation model to correct the output, the calibrated model is used to assess the severity of the fault. That is, "run the compensation model once for each of the two temperature sets and see how much the results differ."

[0023] Based on the above-mentioned IMU temperature compensation fault diagnosis method, the present invention also proposes an IMU diagnostic system, characterized in that it includes: a. An IMU chip with an integrated first temperature sensor to provide a master temperature value T1, which is the input signal for subsequent compensation, i.e. the object to be diagnosed; b. An auxiliary temperature source to provide an independent reference temperature value T2; c. A processing unit, connected to the IMU chip and the auxiliary temperature source, for storing the pre-calibrated (offline calibrated) IMU error compensation model and the multi-order RC thermal network model, and running the diagnostic method described in any one of the above.

[0024] The processing unit is responsible for real-time calculations, running a periodic task (e.g., every 10ms) and executing the branch logic of each stage, from temperature signal acquisition to data preprocessing (for the acquired T1, T2...). amb Perform low-pass filtering (noise reduction), call the RC model to calculate the IMU model temperature, calculate the temperature difference, call the temperature compensation model to calculate the two temperatures (T1, T2) as the zero bias (B) input. comp ) and scaling factor (K) comp The task involves calculating and comparing the difference between the two compensation values, performing operations and conditional logic judgments, and so on.

[0025] Furthermore, the processing unit runs a multi-stage diagnostic algorithm in real time. This multi-stage diagnostic algorithm is a progressive diagnostic strategy that moves from assessing the reasonableness of the temperature signal to assessing the safety hazards of the fault. The expected causes of the fault are as follows: ; The process comprises stages 1 through 4, prioritizing computational efficiency and avoiding false alarms. Firstly, stages 1 and 2 are relatively fast (simple logical judgments and algebraic operations), while stages 3 and 4 require complex floating-point operations within the compensation model. Stage 4, in addition to the static zero-bias compensation value calculation, also involves dynamic scaling factor calculations, consuming more computational power. Therefore, if stage 2 passes (physical consistency), it indicates a good match between T1 and T2, directly determining no fault and omitting the subsequent complex calculations in stages 3 and 4, saving computational resources for the processing unit. Secondly, stage 2 is "physical anomaly detection," while stages 3 and 4 are "hazard confirmation." Physical anomalies do not necessarily mean the device is unusable. For example, if the actual temperature difference is slightly larger, but the IMU zero bias is not sensitive to temperature within that temperature range (flat area), stage 3 will still pass, preventing a fault report that would render the function unusable. Therefore, maintaining the order of these four stages constitutes a complete diagnostic link from electrical integrity → physical consistency → fault hazard assessment.

[0026] As a further improvement of the present invention, the auxiliary temperature source includes: a. A device for measuring the ambient reference temperature T amb Physical temperature sensors, such as NTC thermistors mounted on the PCB board adjacent to the IMU; b. A virtual sensor: based on IMU power consumption P diss and ambient reference temperature T amb A multi-stage RC thermal network model is used to construct the master temperature value T1, the reference temperature value T2, and the power consumption P. diss The dynamic physical relationship between them.

[0027] Because physical sensors cannot be directly integrated into the IMU and placed flush with the MEMS structure, the physical distance between the sensor and the heat-sensitive components, which are located on the PCB, creates a temperature difference. On the other hand, under steady-state conditions, the sensor temperature may be close to the MEMS core temperature; however, during high-power operation (such as ASIC self-heating) or drastic environmental changes (cold start), a temperature difference of several degrees, or even tens of degrees Celsius, may occur. To eliminate this heat transfer difference, a mathematical model is established, using measurable boundary conditions (PCB temperature, system power consumption) to calculate the unmeasurable internal state of the IMU. This is the task of the RC thermal network model. Its core function is based on the law of conservation of energy, constructing the master temperature value (T1), reference temperature value (T2), and power consumption (P). diss The dynamic physical correlation between T1 and T2 is used to predict in real time the temperature difference that should exist between T1 and T2 under current power consumption and environmental conditions, thus providing a physical benchmark for subsequent diagnostics.

[0028] Taking a second-order RC thermal model consisting of two independent nodes (Node 1: the IMU chip itself; Node 2: the auxiliary temperature source and surrounding PCB area) as an example, when the chip generates heat during operation, Node 1 heats up first, and the heat is conducted through the thermal resistance R. After a period of time, Node 2 begins to heat up. The specific workflow is as follows: Input collection: P diss (t): Real-time self-heating power consumption of the IMU chip (calculated from the IMU operating current I × V); T amb : Ambient reference temperature measured by NTC; T1[n-1], T2[n-1]: The calculated model temperatures of the IMU chip and auxiliary temperature source at the previous cycle time. At the initial moment of the diagnostic program startup (n=0), the system reads the sensor measured data to initialize the state of the thermal network model. T1[0] and T2[0] are respectively taken from T1 and T2[n-1]. amb The actual measured value of the sensor.

[0029] Mathematical model (system of differential equations): Based on the thermo-electric analogy (current = heat flow, voltage = temperature, resistance = thermal resistance, capacitance = thermal capacity). ; Establish the nodal heat balance equation: Equation 1: ; left( ): The heat stored in the chip itself (represented by the temperature change T1).

[0030] The first item on the right (P) diss ): Heat source. The self-heating power consumption generated by the IMU chip during operation.

[0031] The second item on the right ( ): Heat flowing to the PCB. Heat is conducted to node 2 through the chip pins or the bottom thermal adhesive.

[0032] The third item on the right ( Heat flowing into the environment. The chip dissipates heat directly into the environment through its package surface.

[0033] Equation 2: ; left( ): The heat stored in the PCB node (represented by the temperature change of T2).

[0034] The first item on the right ( ): Heat from the chip. This is the energy source for this node (note: it is equal in magnitude and opposite in sign to the corresponding term in Equation 1, which conforms to the law of conservation of energy).

[0035] The second item on the right ( Heat flowing into the environment. The PCB dissipates heat to the air or the casing through its surface.

[0036] Algorithm implementation (discretization solution): In subsequent cycles, the model state value from the previous cycle (e.g., 10ms) is combined with the current power consumption input processing unit to perform discrete iterative calculations. The current temperature = temperature of the previous cycle + (self-heating - conducted heat - heat dissipation) × (time / heat capacity). Equation 3: ; The final output is the predicted temperature difference for the current cycle: ΔT model [n] = T1[n] - T2[n].

[0037] Compared with the prior art, the beneficial effects of the present invention are: (1) Extremely low cost: This invention only requires the addition of an extremely low-cost NTC thermistor (or a pure software thermal model) to achieve high coverage of critical thermal faults, which greatly reduces the threshold for achieving functional safety. (2) Solving the diagnostic challenges during the startup phase: The diagnostic method of this invention uses a base ambient temperature combined with a multi-order RC thermal model that can describe transient thermal characteristics. This enables the method to have diagnostic capabilities even during the initial startup phase of the system, and it does not depend on any external reference source (such as GNSS); (3) Avoids false alarms and has high reliability: The core advantage of this invention lies in its "function-oriented" diagnostic logic (stages three and four), which can distinguish between "benign temperature differences" and "malicious faults." Even if there are differences between T1 and T2 that exceed the expectations of the thermal model (residual T Thresh T However, the results obtained after substituting both into the compensation model are similar (i.e., residual). offset (If the temperature difference is very small), this method will determine it as "no fault", thus effectively avoiding false alarms caused by simple temperature difference comparison; (4) Complies with functional safety (ISO26262) design principles: The threshold Thresh of this method offset It is derived by reverse calculation through a rigorous error propagation model based on the security requirements of the application layer. This ensures that the diagnostic system is only triggered when the fault is about to "endanger" the system security, giving the diagnostic sensitivity a clear, traceable physical and security significance. Attached Figure Description

[0038] Figure 1 This is a flowchart of the multi-stage diagnostic process in this invention; Figure 2 This is a flowchart illustrating the computation of the multi-order RC thermal network model in this invention. Figure 2 In the model, T1_node represents the temperature of the IMU chip. It is the first critical node. T2_node: Represents the auxiliary temperature source and the surrounding PCB area. It is the second key node of the model; P_diss: The self-heating power consumption of the IMU is modeled as a "current source" that continuously injects "heat" into the T1 node; R_th_12: The main thermal resistance between T1 and T2 (e.g., conduction through PCB copper). C_th_1 and C_th_2: The heat capacities of T1 and T2, respectively. They determine the "inertia" (i.e., thermal delay) of heating and cooling at the two nodes. R_th_1a and R_th_2a: The thermal resistance of each node to the "environmental baseline" Tamb, representing their heat dissipation paths to the external environment through packaging, air convection, etc. Detailed Implementation

[0039] The present invention will now be described in detail with reference to the embodiments shown in the accompanying drawings. However, it should be noted that these embodiments are not intended to limit the present invention. Equivalent changes or substitutions in function, method, or structure made by those skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0040] The method disclosed in this invention provides a flexible and low-cost solution while ensuring high reliability. It can be widely applied to IMU application scenarios that need to meet functional safety requirements, such as automotive, aerospace and high-precision industrial robots.

[0041] This invention discloses a fault diagnosis method for temperature compensation in an inertial measurement unit (IMU), used to diagnose whether a fault in the first temperature sensor T1 in the IMU causes functional harm to the error compensation result of the IMU. The method runs on a processing unit and includes the following steps: Step 1: Independent rationality check; obtain the main temperature value T1 provided by the first temperature sensor inside the IMU chip, and the reference temperature value T2 provided by the auxiliary temperature source on the IMU mounting component, and check whether the readings of T1 and T2 are within the preset physical operating range; if T1 is out of range, it is determined to be a serious fault; if T2 is out of range, the diagnostic task is terminated.

[0042] Step Two: Dynamic Temperature Difference Comparison Process; a. Calculate the actual difference between T1 and T2, based on the IMU power consumption P. diss and / or ambient reference temperature T amb The expected temperature difference ΔT between T1 and T2 is calculated using a pre-calibrated thermal model. model ; b. Calculate the residual temperature difference. T = |T1-T2| - ΔT model ; c. When residual T ≤ Thresh T When no fault is found, the residual value is determined to be fault-free; when the residual value is zero, the residual value is determined to be fault-free. T Thresh T Then proceed to the next step of functional testing; The thermal model is a description of the power consumption P of T1, T2, and IMU. dissA multi-stage RC thermal network model of the dynamic thermal relationship between the two uses several resistor and capacitor components to simulate the transient and steady-state process of heat conduction from IMU chip T1 to auxiliary temperature source T2 through different paths.

[0043] Step 3: Static functional comparison process; static functional diagnosis corresponds to long-distance straight travel (sensitive to zero bias); a. The T1 and T2 and their respective rates of change d T1 / d t d T2 / d t Substituting each into the pre-calibrated IMU error compensation model, two sets of zero-bias compensation vectors B are calculated. compT1 and B compT2 ; b. Calculate the difference between the two sets of compensation results. offset = || B compT1 - B compT2 ||2; c. The difference residual offset Compared with the preset functional safety threshold Thresh offset Compare; d. When residual offset Thresh offset When this occurs, it is determined to be a static functional fault, indicating that the fault in T1 has caused harm to the final output of the IMU beyond the safety requirements.

[0044] In step three, when residual offset ≤ Thresh offset This also includes a dynamic functional comparison process; a. Substitute T1 and T2 into the complete compensation algorithm (zero bias + scaling factor), based on the IMU's raw readings (Acc). Raw Gyr Raw This yields two sets of final output data Acc. CorT1 and Acc CorT2 and Gyr CorT1 and Gyr CorT2 ; b. Calculate the final output difference: residual acc = ||Acc CorT1 -Acc CorT2 ||2 and residual gyr = ||Gyr CorT1 -Gyr CorT2 ||2; c. If residual acc ThreshAcc_Cor or residual gyr Thresh Gyr_Cor If so, it is determined to be a dynamic functional failure.

[0045] Among them, Thresh Acc_Cor The acceleration dynamic consistency threshold is used to diagnose whether the accelerometer's scale factor is affected by temperature and exceeds the limit.

[0046] Its physical meaning is: when a vehicle undergoes large dynamic acceleration and deceleration (such as emergency braking), if the temperature difference between T1 and T2 causes inconsistent scaling factor compensation, the system's allowed "maximum error in acceleration calculation".

[0047] Thresh Gyr_Cor The dynamic consistency threshold for angular velocity is used to diagnose whether the scale factor of the gyroscope is affected by temperature and exceeds the limit.

[0048] Its physical meaning is: when a vehicle makes a sharp turn (such as a turnaround or lane change), if the temperature difference between T1 and T2 causes inconsistent scaling factor compensation, the system's maximum allowable error in angular velocity calculation.

[0049] In step three, the functional safety threshold Thresh offset It is derived from the application's security requirements through reverse derivation using an error propagation model.

[0050] In step three, the IMU error compensation model is a high-order polynomial or neural network model used to calculate the compensation value B for the IMU zero bias based on temperature and its rate of change. comp Compensation value K of the scaling factor comp .

[0051] Based on the above-mentioned IMU temperature compensation fault diagnosis method, this invention also proposes an IMU diagnostic system, including: a. An IMU chip with an integrated first temperature sensor to provide the main temperature value T1; b. An auxiliary temperature source to provide an independent reference temperature value T2; c. A processing unit connected to the IMU chip and an auxiliary temperature source for storing a pre-calibrated IMU error compensation model and a multi-order RC thermal network model, and for running the diagnostic method according to any one of claims 1-6.

[0052] The auxiliary temperature source includes: a. A device for measuring the ambient reference temperature T amb Physical temperature sensor; b. A virtual sensor: based on IMU power consumption P diss and ambient reference temperature T amb A multi-stage RC thermal network model is used to construct the master temperature value T1, the reference temperature value T2, and the power consumption P. diss The dynamic physical relationship between them.

[0053] Example 1 This embodiment takes a dead reckoning (DR) module for an automotive ADAS system as an example. The goal of this module is to meet the functional safety requirements of ASIL B level.

[0054] This embodiment is for a dead reckoning system for vehicles equipped with Level 2 autonomous driving functions, which requires the IMU to meet ASIL B level functional safety.

[0055] I. System Construction and Offline Calibration: (1) Hardware: Select an automotive-grade IMU, whose ASIC integrates the main temperature sensor (T1). On the PCB, a high-precision NTC thermistor is placed next to the IMU as a reference temperature sensor (Tamb); (2) Calibration: During the calibration stage before mass production, the PCB module is placed in a high and low temperature chamber. Thermocouples are placed at key points to detect the actual real-time temperatures T1 and T2; (3) NTC calibration: Measure the NTC resistance at three temperature points: -40 ℃, +25 ℃, and +85 ℃, and use the Steinhart-Hart equation 1 / T=A+Bln(R)+C(ln(R)) 3 Solve for coefficients A, B, and C, and make corrections considering their self-heating effect; (4) Thermal model calibration, such as Figure 2 As shown: Temperature cycling was performed in the chamber at different rates (e.g., 1 ℃ / min and 4 ℃ / min), while T1, T2 and IMU power consumption P were recorded. diss A second-order RC thermal network model (R) was fitted. th,1 C th,1 ,R th,2 C th,2 ); (5) IMU Compensation Model Calibration: During the above temperature cycle, the raw output of the six axes of the IMU is recorded simultaneously. A calibration model is fitted that includes temperature (T), temperature change rate (dT / dt), and temperature accumulation term ΔT. accumulation Higher-order polynomials are used as IMU error compensation models. ; (6) Diagnostic threshold setting: Based on application requirements (DR lateral and longitudinal errors L of 1KM constant speed straight driving distance). err,max(≤ 1.5 m, vehicle speed V = 80 km / h), derived by reverse derivation of the formula: Longitudinal velocity: V x ≈22.22m / s; Travel time: T≈45s; True angular velocity: ω=0 (straight-line travel); True acceleration: a=0 (uniform speed); Error requirements: Lateral ΔY≤1.5m, longitudinal ΔD≤1.5m.

[0056] II. Derivation of Lateral Error-Gyroscope Diagnostic Threshold: Error model: The main source of error is the zero bias Δω b； The heading angle error accumulates over time: Δθ(t) = Δω b ×t; Lateral velocity error: V y (t)=V x ×sin(Δθ)≈V x ×Δω b ×t; Lateral position error ΔY is V y Integral over time: ; Reverse derivation: ; calculate: ; Unit conversion: 0.0000666 rad / s × (180 / π) ≈ 3.8 mdps To ensure a lateral error of less than 1.5 meters (0.15%) while traveling at a constant speed of 80 kph for 1 km, the zero bias of the Z-axis gyroscope must be better than 3.8 mdps.

[0057] III. Derivation of Longitudinal Error-Accelerometer Diagnostic Threshold: Error model: The primary source of error is zero bias: Δa b； The DR system incorrectly assumes the vehicle is moving at a speed of Δa. b accelerate; The longitudinal position error ΔD is Δa b Second integral: ; Reverse derivation: ; calculate: ; Unit conversion (m / s² to mg): (0.00148m / s 2 / 9.8m / s 2 )×1000≈0.15mg; Under the condition of traveling at a constant speed of 80 kph for 1 km, in order to ensure that the longitudinal error is less than 1.5 meters (0.15%), the zero bias of the X-axis accelerometer must be better than 0.15 mg.

[0058] Example 2 The difference from Example 1 is that while Example 1 continues to use "functional safety target: lateral and longitudinal position error limits" for reverse calculation, the operating condition has been changed from "uniform linear motion" to "high dynamic maneuver," and the following content has been added: A. Accelerometer dynamic threshold Thresh Acc_Cor Derivation: Operating condition: The vehicle is undergoing emergency braking on a highway; Braking deceleration: a brake =5m / s 2 (Approximately 0.5g); Braking duration: t = 4s (deceleration from 72kph to 0); Safety objective: During this process, the longitudinal positioning deviation ΔD caused by IMU error must not exceed 1.5 meters; Error model: Position error: ; Here Δa err This is the threshold we require: Thresh Acc_Cor； calculate: ; Unit conversion: 0.1875 m / s 2 ≈19mg; Conclusion: Thresh, the dynamic consistency threshold of accelerometers Acc_Cor The setting is 19mg. This means that under the two temperature compensation systems T1 and T2, if the difference in acceleration output exceeds 19mg, a dynamic functional fault is determined.

[0059] B. Gyroscope dynamic threshold (Thresh) Gyr_Cor Derivation: Operating condition: The vehicle passes through a high-curve ramp (Curve Traversal); Vehicle turning angular velocity: ω turn =30° / s (for sharper curves); Time taken: t = 10s (continuous turning); Safety objective: During this process, the lateral deviation ΔY caused by IMU error must not exceed 1.5 meters; Error model: This is the scaling factor error ω turn The resulting deviation in heading angle leads to a deviation in lateral position. Simplified estimation formula (small angle approximation): ; Assuming the vehicle speed is V x =10m / s (36kph, ramp speed limit); Here Δω err That is, the threshold Thresh Gyr_Cor； calculate: ; ; Unit conversion: ; Conclusion: The dynamic consistency threshold setting for gyroscopes is Thresh. Gyr_Cor It is 0.17 dps.

[0060] As attached Figure 1 As shown, the processing unit executes the following multi-stage diagnostic process in real time: Phase 1: Independent rationality check; Obtain the main temperature value T1 provided by the first temperature sensor inside the IMU chip, and the reference temperature value T2 provided by the auxiliary temperature source on the IMU mounting assembly, and check whether the readings of T1 and T2 are within the preset physical operating range. If T1 is out of range, mark the IMU as faulty; if T2 is out of range, stop the diagnosis. Phase Two: Dynamic Temperature Difference Comparison Process; a. Calculate the actual difference between T1 and T2, based on the IMU power consumption P. diss And / or ambient reference temperature, using a pre-calibrated thermal model to calculate the expected temperature difference ΔT between T1 and T2. model ; b. Calculate the residual temperature difference. T =|T1-T2|-ΔT model ; c. When residual T ≤Thresh T When no fault is found, the residual value is determined to be fault-free; when the residual value is zero, the residual value is determined to be fault-free. T Thresh T Then proceed to the next step of functional testing; Phase Three: Static Functional Comparison Process; a. The T1 and T2 and their respective rates of change dT1 / d t d T2 / d t Substituting each into the pre-calibrated IMU error compensation model, two sets of zero-bias compensation vectors B are calculated. compT1 and B compT2 ; b. Calculate the difference between the two sets of compensation results. offset =||B compT1 -B compT2 ||2; c. The difference residual offset Compared with the preset functional safety threshold Thresh offset Compare; d. When residual offset Thresh offset When this occurs, it is determined to be a static functional fault, indicating that the fault in T1 has caused harm to the final output of the IMU beyond the safety requirements.

[0061] Phase Four: Dynamic Functional Comparison Process; When residual offset ≤Thresh offset If so, proceed with the following steps; a. Substitute T1 and T2 into the complete compensation algorithm (zero bias + scaling factor), based on the IMU's raw readings (Acc). Raw Gyr Raw This yields two sets of final output data Acc. CorT1 and Acc CorT2 and Gyr CorT1 and Gyr CorT2 ; b. Calculate the final output difference: residual acc =||Acc CorT1 -Acc CorT2 ||2 and residual gyr =||Gyr CorT1 -Gyr CorT2 ||2; c. If residual acc Thresh Acc_Cor or residual gyr Thresh Gyr_Cor If the error is found, it is determined to be a dynamic functional failure, that is, the faulty T1 signal causes the zero bias and / or scale factor correction to exceed the safety requirements. If all checks pass, the diagnosis is successful and there is no fault.

[0062] Example 3 This embodiment is applied to the end effector of a high-precision industrial robot that is extremely sensitive to cost and has limited space.

[0063] I. System Configuration and Model Calibration: Hardware configuration: Main temperature (T1): The temperature sensor integrated inside the IMU chip; Auxiliary temperature (T2 - virtual): In this embodiment, no physical sensor is used; it is implemented purely in software. T2 is defined as the "virtual theoretical temperature" calculated by the processing unit. Virtual T2 calculation: T2 = T model =g(P diss ,T amb ,R th C th ); Thermal model input: At this point, the RC thermal network model receives real-time power consumption P. diss and ambient temperature T amb As input, where T amb The ambient temperature is read from the robot's surroundings. Advantages: Zero increase in hardware costs.

[0064] II. Setting the diagnostic threshold: Because industrial robots may require control precision in a shorter time, such as an end effector positioning error of ≤1.0mm within 5 seconds, Operating condition: Time taken for a single action T = 5s; Safety target: End-effector position deviation ΔX ≤ 1.0 mm. Structural parameters: The lever arm length from the IMU to the end effector of the robot arm is set to R=1m. Threshold derivation: Based on the formula ΔX≈R·(Bias) gyro ·T) is derived by reverse deduction: ; In the formula, Bias gyro The gyroscope has zero bias; the threshold value needs to be determined. Substitute the values: Bias gyro ≤(1.0×10 -3 ) / (1×5)=2×10 -4 rad / s=0.0002 rad / s≈11.5mdps, Thresh (Static Functional Thresh) offset Set to 11.5 mdps.

[0065] III. Diagnostic Logic Phase Two (Physical Consistency): Compare the measured value T1 with the virtual value T2 (T model The difference between ) If |T1-ΔT model |>Thresh T The physical layer is determined to be faulty (possibly due to IMUT1 corruption), and the process proceeds to the next diagnostic stage. Phase 3 (Static Functionality): Although deviations were detected at the physical layer, further hazard assessment is required. T1 and T model Substitute each into the IMU error compensation model and calculate B. compT1 and B compT_model， Difference calculation: residual offset =||B compT1 -B compT_model || 2， Judgment: If residual offset If the speed exceeds 11.5 mdps, a fault is detected and the robot stops moving. Advantages: This embodiment overcomes space limitations and cost constraints, while utilizing a software thermal model to simulate the expected behavior of T1, achieving redundancy, and is particularly suitable for diagnosis during the IMU cold start phase.

[0066] The following advantages can be achieved by using the inertial measurement unit temperature compensation fault diagnosis method and system provided by this invention: (1) Extremely low cost: Only one extremely low-cost NTC thermistor (or pure software thermal model) needs to be added to achieve high coverage of critical thermal faults, which greatly reduces the threshold for achieving functional safety. (2) Solve the diagnostic challenges in the startup phase: By combining the basic ambient temperature with a multi-order RC thermal model that can describe transient thermal characteristics, this method can also have diagnostic capabilities in the preheating phase of the system startup, and does not depend on any external reference source (such as GNSS). (3) Avoids false alarms and has high reliability: This method is "function-oriented" and has "diagnostic logic" (such as...). Figure 1 The core advantage lies in the middle stages (three and four). Even if the temperature difference between T1 and T2 is large, the results obtained after substituting them into the compensation model are similar, and this method will still determine that there is "no fault", effectively avoiding false alarms caused by simple temperature difference comparison; (4) Complies with functional safety (ISO26262) design principles: The diagnostic threshold is calculated in reverse through a strict error propagation model based on the safety requirements of the application layer, ensuring that the diagnostic system is only triggered when the fault is about to "endanger" the system safety.

[0067] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for diagnosing temperature compensation faults in an inertial measurement unit (IMU), used to diagnose whether a fault in the first temperature sensor T1 within the IMU causes functional harm to the error compensation results of the IMU, characterized in that... The method operates on a processing unit and includes the following steps: Step 1: Independent rationality check; obtain the main temperature value T1 provided by the first temperature sensor inside the IMU chip, and the reference temperature value T2 provided by the auxiliary temperature source on the IMU mounting assembly, and check whether the readings of T1 and T2 are within the preset physical operating range; Step Two: Dynamic Temperature Difference Comparison Process; a. Calculate the actual difference between T1 and T2, based on the IMU power consumption P. diss and / or ambient reference temperature T amb The expected temperature difference ΔT between T1 and T2 is calculated using a pre-calibrated thermal model. model ; b. Calculate the residual temperature difference. T = |T1-T2| - △T model ; c. When residual T ≤ Thresh T When no fault is found, the residual value is determined to be fault-free; when the residual value is zero, the residual value is determined to be fault-free. T Thresh T Then proceed to the next step of functional testing; Step 3: Static functional comparison process; a. The T1 and T2 and their respective rates of change d T1 / d t d T2 / d t Substituting each into the pre-calibrated IMU error compensation model, two sets of zero-bias compensation vectors B are calculated. compT1 and B compT2 ; b. Calculate the difference between the two sets of compensation results. offset = || B compT1 - B compT2 ||2; c. The difference residual offset Compared with the preset functional safety threshold Thresh offset Compare; d. When residual offset Thresh offset When this occurs, it is determined to be a static functional fault, indicating that the fault in T1 has caused harm to the final output of the IMU beyond the safety requirements.

2. The method according to claim 1, characterized in that, In step three, when residual offset ≤Thresh offset This also includes a dynamic functional comparison process; a. Substitute T1 and T2 into the complete compensation algorithm, based on the IMU's raw readings (Acc). Raw Gyr Raw This yields two sets of final output data Acc. CorT1 and Acc CorT2 and Gyr CorT1 and Gyr CorT2 ; b. Calculate the final output difference: residual acc = ||Acc CorT1 -Acc CorT2 ||2 and residual gyr = ||Gyr CorT1 -Gyr CorT2 ||2; c. If residual acc Thresh Acc_Cor or residual gyr Thresh Gyr_Cor If so, it is determined to be a dynamic functional failure.

3. The method according to claim 1, characterized in that, In step one, if T1 is out of range, it is determined to be a serious fault; if T2 is out of range, the diagnostic task is terminated.

4. The method according to claim 1, characterized in that, In step three, the functional safety threshold Thresh offset It is derived from the application's security requirements through reverse derivation using an error propagation model.

5. The method according to claim 1, characterized in that, In step two, the thermal model is a description of T1, T2, and IMU power consumption P. diss A multi-stage RC thermal network model of the dynamic thermal relationship between the two uses several resistor and capacitor components to simulate the transient and steady-state process of heat conduction from IMU chip T1 to auxiliary temperature source T2 through different paths.

6. The method according to claim 1, characterized in that, In step three, the IMU error compensation model is a high-order polynomial or neural network model used to calculate the compensation value B for the IMU zero bias based on temperature and its rate of change. comp Compensation value K of the scaling factor comp .

7. An IMU diagnostic system implementing the IMU temperature compensation fault diagnosis method according to any one of claims 1 to 6, characterized in that, include: a. An IMU chip with an integrated first temperature sensor to provide the main temperature value T1; b. An auxiliary temperature source to provide an independent reference temperature value T2; c. A processing unit connected to the IMU chip and an auxiliary temperature source for storing a pre-calibrated IMU error compensation model and a multi-order RC thermal network model, and for running the diagnostic method according to any one of claims 1-6.

8. The system according to claim 7, characterized in that, The auxiliary temperature source includes: a. A device for measuring the ambient reference temperature T amb Physical temperature sensor; b. A virtual sensor: based on IMU power consumption P diss and ambient reference temperature T amb A multi-stage RC thermal network model is used to construct the master temperature value T1, the reference temperature value T2, and the power consumption P. diss The dynamic physical relationship between them.

Images