Static calibration system and calibration method of fuel pressure sensor based on CAN communication

Abstract

The invention discloses a static calibration system and method for a fuel pressure sensor based on CAN communication, and the system is characterized in that the system comprises a fuel pressure sensor, a temperature control cabin, a pressure servo system, a multi-environment collection module, and a six-degree-of-freedom mechanical arm, the temperature control cabin is used for simulating the temperature change, the pressure servo system is used for simulating the pressure change, and the multi-environment collection module is used for collecting the pressure change. The multi-environment acquisition unit integrates a temperature sensor, a stress detection unit and a CAN communication unit, the six-degree-of-freedom mechanical arm is used for simulating mechanical stress changes, and the calibration method comprises three-dimensional environment parameter calibration space construction, multi-algorithm cascade compensation model and intelligent data compression technology. According to the invention, the precision, the efficiency and the storage space utilization rate of the fuel pressure sensor in the static calibration stage process are improved, and a new technical development direction is provided for the development of the fuel sensor.

Description

Technical Field

[0001] The present invention relates to the technical field of sensor calibration, and in particular to a static calibration system and a calibration method for a fuel pressure sensor based on CAN communication. Background Art

[0002] For motorcycle riders, being able to monitor their range anytime, anywhere is crucial for safe riding. Range is directly determined by the remaining fuel level. Currently, the ECU (Electronic Control Unit) collects real-time fuel level data from a fuel sensor and uses a corresponding algorithm to calculate the remaining range. The accuracy of the fuel sensor and its algorithm directly determines the accuracy of the motorcycle's range display.

[0003] However, most existing static calibration methods and data processing methods for fuel pressure sensors have many limitations;

[0004] 1. Single static calibration algorithm: poor resistance to environmental interference

[0005] Most existing calibration technologies rely on single linear regression or polynomial fitting models for static calibration, without considering the coupling interference of multiple physical fields:

[0006] Temperature changes cause sensor drift, without dynamic compensation: The fuel sensor's pressure-voltage characteristic changes nonlinearly with temperature (temperature drift coefficient > 0.05% / °C), but traditional fuel sensor calibration is only performed at room temperature. This results in significant pressure signal offset during high-temperature riding (measured error reaches ±2.5% FS).

[0007] Insufficient power supply noise suppression: The high-frequency noise (10-100kHz) of the motorcycle ignition system affects the sensor's power supply circuit. However, traditional calibration algorithms do not embed a noise separation module, and static calibration parameters become invalid during actual vehicle operation (signal glitches cause transient errors).

[0008] 2. Missing calibration data dimensions: Nonlinear error superposition

[0009] Most existing calibration processes only collect data on the relationship between pressure and voltage, ignoring the impact of other factors on the sensor output value:

[0010] Failure to consider mechanical stress interference with the sensor: Improper sensor installation can affect the sensor's accuracy due to engine vibration (20-500Hz) and frame deformation, or residual stress in the metal diaphragm can cause the sensor's zero-point drift. This is because traditional calibration algorithms don't incorporate stress compensation parameters. Such fuel sensors can experience baseline drift after long-term use.

[0011] Rigid adaptation of medium characteristics: Differences in fuel components (such as ethanol gasoline and pure gasoline) lead to density changes. Existing calibration parameters are not associated with the fuel dielectric constant, and pressure measurement accuracy cannot be guaranteed in mixed fuel scenarios.

[0012] 3. Poor adaptability to dynamic working conditions: the algorithm is out of touch with the physical scene

[0013] Traditional calibration algorithms do not integrate dynamic riding data, resulting in a mismatch between calibration results and actual operating conditions:

[0014] Sudden acceleration / braking signal distortion: High-frequency fluctuations in fuel pressure (e.g., pressure transient rate >10 MPa / s during sudden acceleration) can cause signal delays or peak clipping due to fixed-cutoff frequency filtering algorithms, leading to the ECU misjudging a fuel supply failure (false alarm rate ≥12%).

[0015] Lack of correlation between tilt angle and pressure: When the motorcycle turns, the fuel tank liquid level tilts and the fuel pressure distribution changes, but the calibration algorithm does not associate the tilt sensor data. As a result, the pressure-fuel volume mapping relationship deviates from the calibration curve in the tilted state.

[0016] 4. Calibration results are not reproducible: lack of self-verification and iteration mechanisms

[0017] Existing technologies fix parameters after calibration and are unable to cope with sensor aging or environmental changes:

[0018] Poor long-term stability: The sensor's sensitive components creep or the packaging glue ages, resulting in decreased accuracy. Traditional calibration parameters lack periodic self-correction functions, and the full-scale error exceeds ±1.5% after six months of use.

[0019] Lack of cross-validation: The calibration process is not cross-checked with independent data sources such as fuel flow meters and oxygen sensors. Single data calibration cannot identify sensor nonlinear faults (such as local sensitivity attenuation caused by diaphragm fatigue cracks).

[0020] 5. Low calibration efficiency: high reliance on manual labor

[0021] The existing calibration process relies on manual multi-point calibration and empirical parameter debugging:

[0022] Low degree of automation: Each sensor needs to be pressurized and data recorded point by point in a constant temperature chamber. Calibration of a single sensor takes more than 30 minutes, which is difficult to meet the needs of motorcycle mass production.

[0023] Poor calibration consistency: Manual adjustment of filter thresholds or compensation coefficients can result in output dispersion of up to ±2% among sensors in the same batch.

[0024] Existing fuel pressure sensor calibration technology is limited by a single algorithm, missing data dimensions, and insufficient adaptability to dynamic scenarios. This results in large pressure measurement errors and low reliability of calculated range in motorcycle riding environments. A calibration method that integrates multi-dimensional data, dynamic environment compensation, and a self-verification mechanism is urgently needed to improve sensor accuracy and stability under complex operating conditions. Summary of the Invention

[0025] In view of the above-mentioned deficiencies in the prior art, the present invention provides a static calibration system and calibration method for a fuel pressure sensor based on CAN communication, which realizes high-precision and high-efficiency static calibration of the fuel sensor.

[0026] In order to achieve the above object, the present invention adopts the following technical solutions:

[0027] A static calibration system for a fuel pressure sensor based on CAN communication, characterized by comprising a fuel pressure sensor, a temperature control cabin, a pressure servo system, a multi-environment acquisition module, and a six-degree-of-freedom robotic arm.

[0028] The fuel pressure sensor is integrated with a CAN transceiver for transmitting pressure data to a host computer;

[0029] The temperature control chamber adopts liquid nitrogen and electric heating wire dual-mode temperature control to simulate temperature changes. The temperature adjustment range is -40℃-150℃, and the adjustment accuracy is 0.3℃.

[0030] The pressure servo system is used to simulate pressure changes, with a pressure range of 50kPa-500kPa and an adjustment accuracy of 1kPa;

[0031] A six-degree-of-freedom robotic arm is used to simulate installation stress changes, with a stress range of 0Mpa-10Mpa and an error of less than ±0.5%;

[0032] The multi-environment acquisition module includes a CAN transceiver unit, a temperature sensor, and a stress detection unit, and is used to collect the temperature and stress of the fuel pressure sensor and transmit the data to the host computer through the CAN transceiver unit.

[0033] A static calibration method for a fuel pressure sensor based on CAN communication, characterized in that it includes the following steps:

[0034] S1. Environmental parameter coupling calibration space is constructed. The calibration space is constructed by temperature, pressure, and mechanical stress, and the total number of calibration points N is calculated by the following formula:

[0035]

[0036] Where T max 、T minis the highest and lowest temperature during the calibration process, △T is the discretization temperature step; P max 、P min is the maximum and minimum pressure during the calibration process, △P is the discretized pressure step, σ max and σ min are the maximum and minimum mechanical stresses, △σ is the discretized mechanical stress step size;

[0037] S2, adopts a multi-algorithm cascade compensation model and a three-level compensation architecture, with inputs: original output value Raw, temperature T, pressure P, and stress σ;

[0038] First layer: third-order polynomial basis,

[0039] Ba es=a0+a1*T+a2*P+a3*σ+a4*T 2 +a5*P 2 +...+a 18 *σ 2 *P+a 19 *T*P*σ

[0040] Base is the polynomial basis of the reference pressure value output by the polynomial model, a0~a 19 are polynomial coefficients obtained by fitting calibration data using the least squares method and are used to describe the linear and nonlinear coupling relationships among temperature, pressure, and stress;

[0041] The second layer: RBF neural network nonlinear fitting,

[0042]

[0043] RBF_OUT is the pressure compensation value output by the RBF network, X represents the input vector, including temperature T, pressure P and stress σ,C i represents the center vector of the i-th hidden layer node, which is extracted from N sets of calibration data through K-means clustering, wi represents the weight coefficient of the i-th hidden layer node, which is obtained through supervised learning training, σ i represents the width parameter of the i-th Gaussian kernel function, which is calculated by the distance between cluster centers, ||XC i ||Euclidean distance between the input vector and the center;

[0044] Third layer: Random forest residual correction

[0045] Residual=RandomForest.predict([T,P,σ,Base,RBF_OUT])

[0046] RandomForest.predict() is a random forest model. The original parameters and the output of the previous model are input in (), the residual is predicted and corrected, and Residual is the pressure value output by the random forest model;

[0047] The final formula is:

[0048] Final=Base+RBF_OUT+Residual

[0049] Final is the pressure value of the final calibration output;

[0050] S3. Intelligent compression of calibration data, based on sparse coding algorithm, reduces the dimension to 8 principal components through principal component analysis method; adopts Huffman coding compression, with a compression rate of ≥97%; stores the parameter table in the 0x1A00~0x1BFF address segment in FLASH.

[0051] The beneficial effects of the present invention include: significantly improving the static calibration accuracy and efficiency of sensors through multi-parameter coupling calibration technology; the multi-algorithm cascade compensation model breaks through the accuracy bottleneck brought about by the use of a single algorithm in traditional sensors; and the intelligent data compression technology breaks through the limitations of on-board storage resources and is suitable for the calibration of various fuel sensors. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the connection structure of the calibration system of the present invention;

[0053] Figure 2 It is a multi-algorithm cascade calibration flow chart of the present invention;

[0054] Figure 3 It is a diagram of the calibration parameter storage structure of the present invention. DETAILED DESCRIPTION

[0055] The present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0056] A kind of Figure 1 The static calibration system of the fuel pressure sensor based on CAN communication shown in the figure includes a fuel pressure sensor, a temperature control cabin, a pressure servo system, a multi-environment acquisition module, and a six-degree-of-freedom robotic arm.

[0057] The fuel pressure sensor uses a piezoresistive pressure sensitive element with a nominal pressure of 1.26 kPa, a maximum overload of 20 kPa, and a breaking pressure of no less than 30 kPa. It also integrates a CAN transceiver for transmitting pressure data to the host computer and for connecting to the multi-environment acquisition module.

[0058] The temperature control chamber uses liquid nitrogen and electric heating wire dual-mode temperature control to simulate temperature changes. The temperature adjustment range is -40℃-150℃, and the adjustment accuracy is 0.3℃;

[0059] Pressure servo system, used to simulate pressure changes, with a pressure range of 50kPa-500kPa and an adjustment accuracy of 1kPa;

[0060] Six-degree-of-freedom robotic arm, used to simulate installation stress changes, with a stress range of 0Mpa-10Mpa, an error of less than ±0.5%, and a repeatability of ±0.01mm;

[0061] The multi-environment acquisition module includes a CAN transceiver unit, a temperature sensor, and a stress detection unit. The temperature sensor has a sensing range of -40°C to 150°C. The stress detection unit includes a strain gauge and a vibration sensor, which are used to collect the temperature and stress of the fuel pressure sensor. The CAN transceiver unit uses a CAN FD communication interface: it supports the SAE J1939 protocol and has a transmission rate of 5Mbps, which is used to transmit data to the host computer.

[0062] like Figure 2 The static calibration algorithm of the fuel pressure sensor based on CAN communication shown includes the following steps:

[0063] 1. Three-dimensional parameter space calibration

[0064] Calibration point generation:

[0065] Temperature (T), pressure (P), and stress (σ) are divided by gradient:

[0066] T∈{T min , T min +△T,...,T max -△T,T max},P∈{P min , P min +△P,...,P max -△P,P max},σ∈{σ min ,σ min +△σ,...,σ max -△σ,σ max}.

[0067] T min ,T max They are the minimum and maximum values ​​within the temperature calibration range, △T is the temperature step value, the size of this value affects the number of temperature calibration points, P min ,P max are the minimum and maximum values ​​within the pressure calibration range, △P is the pressure step value, the size of this value affects the number of pressure calibration points, σmin ,σ max are the minimum and maximum values ​​within the stress calibration range, respectively; △σ is the stress step value, the size of which affects the number of stress calibration points.

[0068] Temperature (T), pressure (P), and mechanical stress (σ) constitute the calibration space, covering extreme operating conditions (such as -40°C cold start, low air pressure in the plateau, and excessive installation stress);

[0069] The total number of calibration points N is calculated using the following formula:

[0070]

[0071] 2. Multi-algorithm cascade compensation model

[0072] Three-level compensation structure (such as the attached Figure 2 shown):

[0073] Polynomial basis model (3rd order expansion):

[0074] P Base =a0+a1*T+a2*P+a3*σ+a4*T 2 +a5*P 2 +...+a 18 *σ 2 *P+a 19 *T*P*σ

[0075] P Base is the reference pressure value (polynomial basis) output by the polynomial model, a0~a 19 are polynomial coefficients obtained by fitting the calibration data using the least squares method and are used to describe the linear and nonlinear coupling relationships among temperature, pressure, and stress.

[0076] RBF neural network nonlinear fitting:

[0077]

[0078] P RBF_OUT is the pressure compensation value output by the RBF network, X represents the input vector, including temperature T, pressure P and stress σ,C i represents the center vector of the i-th hidden layer node, which is extracted from N (the total number of calibration points) groups of calibration data through K-means clustering. wi represents the weight coefficient of the i-th hidden layer node, which is obtained through supervised learning (such as gradient descent) training. σ i represents the width parameter of the i-th Gaussian kernel function, which is calculated by the distance between cluster centers, ||XC i ||Euclidean distance between the input vector and the center.

[0079] Random Forest Residual Correction:

[0080] P Final =P Base +P RBF_OUT +Residual(T,P,σ,P Base P RBF )

[0081] Residual() is a random forest model. In (), the original parameters and the pressure value predicted by the previous model output (prediction residual and correction) are input. Final It is the pressure value of the final calibration output.

[0082] 3. Intelligent compression of calibration data

[0083] Sparse coding process:

[0084] (1) Dimensionality reduction to 8 principal components (retaining 99% variance information) through principal component analysis (PCA);

[0085] (2) Huffman coding compression, compression rate ≥ 97%;

[0086] (3) The parameter table is stored in the address segment 0x1A00 to 0x1BFF in FLASH (512 bytes).

[0087] Calibration data storage and CAN communication protocol

[0088] like Figure 3 The calibration parameter table structure shown

[0089] 1. Calibration parameter storage

[0090] Address 0x1A00~0x1A1F: Polynomial coefficients (a0~a 19 , 32 bytes);

[0091] Address 0x1A20~0x1A5F: RBF network parameters (center point C i , weight wi, 64 bytes);

[0092] Address 0x1A60~0x1A7F: Random Forest feature importance table (32 bytes);

[0093] Address 0x1A80~0x1BFF: check code (CRC32) and reserved area.

[0094] Post-calibration verification and exception handling

[0095] 1. Calibration accuracy verification

[0096] Full parameter space test: 10% calibration points are randomly selected for verification, with a maximum error of ≤0.8% FS;

[0097] Temperature drift suppression: At extreme temperatures of -40°C and 150°C, zero point drift is less than ±0.3% FS (traditional method ±2% FS).

[0098] 2. Abnormal calibration processing

[0099] Residual out-of-limit detection: If the residual of a calibration point is greater than 1.2% FS, it triggers iterative recalibration:

[0100] 1). Re-collect the data of this point (the number of sampling times is increased to 500 times);

[0101] 2). Locally update the RBF network weights;

[0102] 3). Recalculate the CRC32 checksum and write it into FLASH.

[0103] Summary of implementation effects

[0104] By constructing a three-dimensional environmental parameter calibration space, a multi-algorithm cascade compensation model, and intelligent data compression technology, the present invention improves the accuracy, efficiency, and storage space utilization of the fuel pressure sensor during the static calibration phase, providing a new technical development direction for the development of fuel sensors.

[0105] The technical solutions provided by the embodiments of the present invention are introduced in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the embodiments of the present invention. The description of the above embodiments is only applicable to help understand the principles of the embodiments of the present invention. At the same time, for those skilled in the art, according to the embodiments of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A static calibration system for a fuel pressure sensor based on CAN communication, characterized by: Including fuel pressure sensor, temperature control cabin, pressure servo system, multi-environment acquisition module, six-degree-of-freedom robotic arm, The fuel pressure sensor is integrated with a CAN transceiver for transmitting pressure data to a host computer; The temperature control chamber adopts liquid nitrogen and electric heating wire dual-mode temperature control to simulate temperature changes. The temperature adjustment range is -40℃-150℃, and the adjustment accuracy is 0.3℃. The pressure servo system is used to simulate pressure changes, with a pressure range of 50kPa-500kPa and an adjustment accuracy of 1kPa; A six-degree-of-freedom robotic arm is used to simulate installation stress changes, with a stress range of 0Mpa-10Mpa and an error of less than ±0.5%; The multi-environment acquisition module includes a CAN transceiver unit, a temperature sensor, and a stress detection unit, and is used to collect the temperature and stress of the fuel pressure sensor and transmit the data to the host computer through the CAN transceiver unit.

2. A static calibration method for a fuel pressure sensor based on CAN communication, characterized by: The following steps are included: S1. Environmental parameter coupling calibration space is constructed. The calibration space is constructed by temperature, pressure, and mechanical stress, and the total number of calibration points N is calculated by the following formula: Where T max 、T min is the highest and lowest temperature during the calibration process, △T is the discretization temperature step; P max 、P min is the maximum and minimum pressure during the calibration process, △P is the discretized pressure step, σ max and σ min are the maximum and minimum mechanical stresses, △σ is the discretized mechanical stress step size; S2, adopts a multi-algorithm cascade compensation model and a three-level compensation architecture, with inputs: original output value Raw, temperature T, pressure P, and stress σ; First layer: third-order polynomial basis, For es=a0+a1*T+a2*P+a3*σ+a4*T 2 +a5*P 2 +...+a 18 *σ 2 *P+a 19 *T*P*σ Base is the polynomial basis of the reference pressure value output by the polynomial model, a0~a 19 are polynomial coefficients obtained by fitting calibration data using the least squares method and are used to describe the linear and nonlinear coupling relationships among temperature, pressure, and stress; The second layer: RBF neural network nonlinear fitting, RBF_OUT is the pressure compensation value output by the RBF network, X represents the input vector, including temperature T, pressure P and stress σ,C i represents the center vector of the i-th hidden layer node, which is extracted from N sets of calibration data through K-means clustering, wi represents the weight coefficient of the i-th hidden layer node, which is obtained through supervised learning training, σ i represents the width parameter of the i-th Gaussian kernel function, which is calculated by the distance between cluster centers, ||XC i ||Euclidean distance between the input vector and the center; Third layer: Random forest residual correction Residual=RandomForest.predict([T,P,σ,Base,RBF_OUT]) RandomForest.predict() is a random forest model. The original parameters and the output of the previous model are input in (), the residual is predicted and corrected, and Residual is the pressure value output by the random forest model; The final formula is: Final=Base+RBF_OUT+Residual Final is the pressure value of the final calibration output; S3. Intelligent compression of calibration data, based on sparse coding algorithm, reduces the dimension to 8 principal components through principal component analysis method; adopts Huffman coding compression, with a compression rate of ≥97%; stores the parameter table in the 0x1A00~0x1BFF address segment in FLASH.

Images