A machine learning-based pressure sensor error calibration method and system

Abstract

The application relates to the technical field of instrument calibration, and particularly discloses a pressure sensor error calibration method and system based on machine learning, which comprises the following steps: acquiring a training set; wherein the training set comprises multiple groups of sample data; normalizing the training set to obtain a normalized training set; establishing an error calibration model for pressure sensor error calibration; training the error calibration model according to the normalized training set to obtain a trained error calibration model; inputting a current working temperature value and a current measured pressure value into the trained error calibration model for error calibration to output a current calibration pressure value corresponding to the current measured pressure value. The pressure sensor error calibration method based on machine learning can quickly and accurately calibrate the measurement error of a pressure sensor and improve the overall performance of the pressure sensor.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of instrument calibration, in particular to a pressure sensor error calibration method and system based on machine learning. BACKGROUND

[0002] The pressure sensor is a device for measuring and converting fluid pressure into standard electrical signals, which is widely used in aerospace, oil drilling, boiler pressure measurement and other industries. Due to the manufacturing process error of the pressure sensor and the temperature drift problem of the direct output signal of the sensor, the output result of the instrument is not accurate enough, it is difficult to provide reliable data for the production process, and the output data of the pressure sensor needs to be calibrated to meet the needs of high-precision industry, so the high-precision error calibration method is very important. SUMMARY

[0003] In view of the existing error calibration needs, the present application provides a pressure sensor error calibration method and system based on machine learning, which can quickly and accurately calibrate the pressure sensor error.

[0004] As a first aspect of the present application, a pressure sensor error calibration method based on machine learning is provided, the pressure sensor is placed in a high-low temperature test box, the temperature in the high-low temperature test box is controlled within a preset temperature range, so that the working temperature of the pressure sensor is within the preset temperature range, a plurality of working temperature values within the preset temperature range are set, and a high-precision pressure source is used to apply different pressures to the pressure sensor at each working temperature value to obtain a plurality of measured pressure values and their corresponding input pressure values output by the pressure sensor at each working temperature value; the pressure sensor error calibration method based on machine learning comprises:

[0005] Step S1: obtaining a training set; wherein the training set includes a plurality of groups of sample data, each group of sample data includes a working temperature value, an input pressure value at the working temperature value and a measured pressure value corresponding to the input pressure value;

[0006] Step S2: normalizing the training set to obtain a normalized training set;

[0007] Step S3: establishing an error calibration model for pressure sensor error calibration;

[0008] Step S4: training the error calibration model according to the normalized training set to obtain a trained error calibration model;

[0009] Step S5: inputting the current working temperature value and the current measured pressure value into the trained error calibration model for error calibration to output the current calibration pressure value corresponding to the current measured pressure value.

[0010] Further, the normalization of the training set to obtain the normalized training set further comprises:

[0011] (1) Normalizing the working temperature values in the plurality of sets of sample data, according to the following formula:

[0012]

[0013] wherein, represents the normalized value of the working temperature value T, T min represents the minimum working temperature value of the pressure sensor, T max represents the maximum working temperature value of the pressure sensor;

[0014] (2) Normalizing the input pressure values in the plurality of sets of sample data, according to the following formula:

[0015]

[0016] wherein, represents the input pressure value normalized value, represents the minimum input pressure value in the plurality of sets of sample data, represents the maximum input pressure value in the plurality of sets of sample data;

[0017] (3) Normalizing the measured pressure values in the plurality of sets of sample data, according to the following formula:

[0018]

[0019] wherein, represents the normalized value of the measured pressure value x, x min represents the minimum measured pressure value in the plurality of sets of sample data, x max represents the maximum measured pressure value in the plurality of sets of sample data.

[0020] Further, the establishment of the error calibration model for the error calibration of the pressure sensor further comprises:

[0021] (1) Calibrating the manufacturing error of the pressure sensor using the Gaussian process regression of the machine learning algorithm:

[0022] μ * = K(x * , X) · K(X, X) -1 · y

[0023]

[0024] wherein, μ* representing the output value of the Gaussian process regression, K(x * representing the covariance matrix, k rbf representing the kernel function of the covariance matrix K(x * representing the kernel function of the covariance matrix K(x f representing the predicted output range, l representing the function smoothness, X representing the working temperature values and the measured pressure values in the training set, y representing the input pressure values in the training set, x * representing the normalized value corresponding to the new input pressure value;

[0025] (2) According to the output characteristic curve change amount DT of the pressure sensor at different temperatures, the temperature drift of the pressure sensor is calibrated:

[0026]

[0027] wherein, T represents the current working temperature value of the pressure sensor, T i represents the previous working temperature value of the current working temperature value T in the training set, T i+1 represents the next working temperature value of the current working temperature value T in the training set, represents the pressure sensor output characteristic curve at the working temperature value T i +1 represents the pressure sensor output characteristic curve at the working temperature value T i+1 i+1 and T i , T c represents the reference temperature value;

[0028] (3) The calibration is performed by linear interpolation method combined with the manufacturing error and the temperature drift of the pressure sensor:

[0029]

[0030] wherein, P' represents the final pressure sensor calibration result, represents the output value of the Gaussian process regression corresponding to the working temperature value T i represents the output value of the Gaussian process regression corresponding to the working temperature value T i+1 , FullScale represents the measurement range of the pressure sensor.

[0031] Further, in the step S4 and the step S5, the symmetric positive definite covariance matrix K is decomposed into the product of two lower triangular matrices L and L T by Cholesky decomposition:

[0032] ​​​K = L·L T

[0033] wherein L and L T represent two lower triangular matrices;

[0034] Substitute α = L -T ·(L -1 ·y) into the formula μ * = K(x * , X)·K(X, X) -1 ·y to obtain

[0035] As another aspect of the present application, a machine learning-based pressure sensor error calibration device is provided for implementing the machine learning-based pressure sensor error calibration method described above, and the machine learning-based pressure sensor error calibration device comprises:

[0036] An acquisition module is configured to acquire a training set, wherein the training set comprises a plurality of groups of sample data, and each group of sample data comprises a working temperature value, an input pressure value at the working temperature value, and a measured pressure value corresponding to the input pressure value.

[0037] A normalization module is configured to normalize the training set to obtain a normalized training set.

[0038] An establishment module is configured to establish an error calibration model for pressure sensor error calibration.

[0039] A training module is configured to train the error calibration model according to the normalized training set to obtain a trained error calibration model.

[0040] A calibration module is configured to input a current working temperature value and a current measured pressure value into the trained error calibration model for error calibration to output a current calibration pressure value corresponding to the current measured pressure value.

[0041] As another aspect of the present application, a machine learning-based pressure sensor error calibration system is provided, comprising: a computer, a pressure sensor signal acquisition device, a high-low temperature test box and a high-precision pressure source, the computer comprising the machine learning-based pressure sensor error calibration device described above, the computer being connected with the pressure sensor signal acquisition device and the high-low temperature test box respectively, the pressure sensor signal acquisition device and the high-low temperature test box being connected, the computer controlling the temperature in the high-low temperature test box within a preset temperature range, the high-precision pressure source being used to sequentially apply different pressures to the pressure sensor in the high-low temperature test box at each working temperature value, so that the pressure sensor signal acquisition device can collect a plurality of measured pressure values output by the pressure sensor, the pressure sensor signal acquisition device uploading the collected plurality of measured pressure values to the computer, and the computer being capable of completing the machine learning-based pressure sensor error calibration process.

[0042] The machine learning-based pressure sensor error calibration method provided by the present application has the following beneficial effects: the error calibration method is essentially a software compensation method, which starts from the physical source of measurement error of the pressure sensor, combines the current advanced machine learning algorithm, and can well solve the data calibration problem of various types of pressure sensors, thereby improving the overall performance of the pressure sensor. BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, and together with the following detailed description, serve to explain the present application, but do not constitute a limitation thereof.

[0044] Figure 1 The flowchart of the machine learning-based pressure sensor error calibration method provided by the present application.

[0045] Figure 2 The flowchart of the specific embodiment of the machine learning-based pressure sensor error calibration method provided by the present application.

[0046] Figure 3 The structure diagram of the error calibration model provided by the present application.

[0047] Figure 4 The structure diagram of the machine learning-based pressure sensor error calibration system provided by the present application. DETAILED DESCRIPTION

[0048] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0049] In order to make the technical personnel in the art better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by the ordinary skilled in the art without creative labor should belong to the scope of protection of the present application.

[0050] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0051] In the present embodiment, a machine learning-based pressure sensor error calibration method is provided, as shown in Figure 4 The pressure sensor is placed in a high-low temperature test chamber, the temperature in the high-low temperature test chamber is controlled within a preset temperature range, so that the working temperature of the pressure sensor is within the preset temperature range, a plurality of working temperature values within the preset temperature range are set, and a high-precision pressure source is used to sequentially apply different pressures to the pressure sensor at each working temperature value to obtain a plurality of measured pressure values output by the pressure sensor at each working temperature value and the corresponding input pressure values; as shown in Figure 1 The machine learning-based pressure sensor error calibration method includes:

[0052] Step S1: obtaining a training set; wherein the training set includes a plurality of groups of sample data, each group of sample data including a working temperature value, an input pressure value at the working temperature value, and a measured pressure value corresponding to the input pressure value;

[0053] Specifically, in step S1, when obtaining a plurality of groups of sample data, starting from the lowest or highest temperature point of the pressure sensor working range, every certain temperature interval in the working range, the pressure sensor is subjected to one-time voltage rise from zero to full scale or voltage drop from full scale to zero. Every sampling point interval a certain pressure during the voltage rise and drop process, the temperature condition, input pressure value and sensor output signal value are recorded.

[0054] Step S2: normalizing the training set to obtain a normalized training set; the normalization processing can enhance the universality and stability of the data calibration method;

[0055] Preferably, the normalization of the training set to obtain a normalized training set further comprises:

[0056] (1) normalizing the working temperature value in the plurality of sets of sample data, and the formula is as follows:

[0057]

[0058] wherein, represents the normalized value of the working temperature value T, T min represents the minimum working temperature value of the pressure sensor, T max represents the maximum working temperature value of the pressure sensor;

[0059] (2) normalizing the input pressure value in the plurality of sets of sample data, and the formula is as follows:

[0060]

[0061] wherein, represents the input pressure value normalized value, represents the minimum input pressure value in the plurality of sets of sample data, represents the maximum input pressure value in the plurality of sets of sample data;

[0062] (3) normalizing the measured pressure value in the plurality of sets of sample data, and the formula is as follows:

[0063]

[0064] wherein, represents the normalized value of the measured pressure value x, x min represents the minimum measured pressure value in the plurality of sets of sample data, x max represents the maximum measured pressure value in the plurality of sets of sample data.

[0065] Step S3: establishing an error calibration model for error calibration of the pressure sensor according to the physical characteristics of the pressure sensor by using a machine learning algorithm;

[0066] It should be noted that the error of the pressure sensor mainly has two sources of manufacturing process and environmental temperature influence. The output characteristic curve of the pressure sensor after manufacturing is completed. Due to the error of the manufacturing process, the actual structure parameters of the sensor cannot be consistent with the theoretical design, and there is an inherent nonlinear error. This nonlinear error is more unpredictable at different working temperatures. These two types of errors are determined after the sensor is manufactured, and each sensor has unique physical characteristics.

[0067] Preferably, as shown in Figures 2-3 The error calibration model for calibrating the error of the pressure sensor further comprises:

[0068] (1) The manufacturing error of the pressure sensor is calibrated by using a Gaussian process regression (Gaussian Process Regression) of a machine learning algorithm:

[0069] μ * = K(x * ,X)·K(X,X) -1 ·y

[0070]

[0071] Wherein, μ * represents the output value of the Gaussian process regression, K(x * ,X) represents the covariance matrix, k rbf represents the kernel function of the covariance matrix K(x * ,X), σ f represents the predicted output range, l represents the function smoothness, X represents the working temperature value and the measured pressure value in the training set, y represents the input pressure value in the training set, x * represents the normalized value corresponding to the new input pressure value;

[0072] (2) The temperature drift of the pressure sensor is calibrated according to the output characteristic curve change DT of the pressure sensor at different temperatures:

[0073]

[0074] Wherein, T represents the current working temperature value of the pressure sensor, T i represents the previous working temperature value of the current working temperature value T in the training set, T i+1 represents the next working temperature value of the current working temperature value T in the training set, represents the output characteristic curve of the pressure sensor at the working temperature value T i , represents the output characteristic curve of the pressure sensor at the working temperature value T i+1the pressure sensor output characteristic curve under the temperature distribution, DT represents the change amount of the pressure sensor output characteristic curve, T i+1 and T i the corresponding change amount of the pressure sensor output characteristic curve, T c represents the reference temperature value, usually 20℃;

[0075] (3) Calibration by linear interpolation combined with the manufacturing error and temperature drift of the pressure sensor:

[0076]

[0077] wherein P' represents the final pressure sensor calibration result, represents the working temperature value T i the corresponding output value of the Gaussian process regression, represents the working temperature value T i+1 the corresponding output value of the Gaussian process regression, FullScale represents the measurement range of the pressure sensor.

[0078] It should be noted that in the calibration process, the temperature drift trend is introduced by linear interpolation to compensate for the error caused by the change of the working temperature. According to the temperature distribution, the temperature drift trend between the two reference curves is introduced by linear interpolation, the influence of the environment temperature on the sensor is restored, and the final calibration result is obtained.

[0079] Step S4: training the error calibration model according to the normalized training set to obtain a trained error calibration model; that is, the trained error calibration model can be used to calibrate the error of the pressure sensor, so that it realizes high-precision output under different working environments.

[0080] Preferably, in the step S3 and step S4, after the error calibration model is established, according to the working temperature range of the pressure sensor, a plurality of pressure sensor output characteristic curves D under different working temperatures are trained, and the change amount DT thereof under different temperatures is calculated, the input of the error calibration model is the working temperature value and the measured pressure value, and the output of the error calibration model is the pressure sensor calibration result.

[0081] Step S5: inputting the current working temperature value and the current measured pressure value into the trained error calibration model for error calibration to output the current calibration pressure value corresponding to the current measured pressure value.

[0082] Preferably, in the step S4 and step S5, the symmetric positive definite covariance matrix K is decomposed into the product of two lower triangular matrices L and L T by Cholesky decomposition, avoiding the situation that the matrix is close to singular, and ensuring the stability of calculation:

[0083] K=L·LT

[0084] wherein L and L T represent two lower triangular matrices;

[0085] Substitute α = L -T · (L -1 · y) into the formula μ * = K(x * , X) · K(X, X) -1 · y to obtain Simplify the calculation of forward calibration.

[0086] Specifically, in the step S5, two pre-trained sensor output characteristic curves under adjacent working temperatures and

[0087] Then use α to perform forward calibration to avoid covariance matrix inversion, and perform Gaussian process regression to obtain the result after manufacturing error calibration and

[0088]

[0089] Finally, the temperature drift amount is introduced by linear interpolation,

[0090] Finally, the output of the error calibration model forward calibration is: P' = FullScale * μ', and FullScale is the measurement range of the pressure sensor.

[0091] As another embodiment of the present application, a pressure sensor error calibration device based on machine learning is provided for implementing the machine learning-based pressure sensor error calibration method described above, and the pressure sensor error calibration device based on machine learning comprises:

[0092] An acquisition module is configured to acquire a training set; wherein the training set comprises a plurality of groups of sample data, and each group of sample data comprises a working temperature value, an input pressure value at the working temperature value, and a measured pressure value corresponding to the input pressure value.

[0093] A normalization module is configured to normalize the training set to obtain a normalized training set.

[0094] An establishment module is configured to establish an error calibration model for pressure sensor error calibration.

[0095] The training module is used to train the error calibration model based on the normalized training set to obtain the trained error calibration model.

[0096] The calibration module is used to input the current operating temperature value and the current measured pressure value into the trained error calibration model for error calibration, so as to output the current calibration pressure value corresponding to the current measured pressure value.

[0097] As another embodiment of the present invention, a pressure sensor error calibration system based on machine learning is provided, such as Figure 4 As shown, the system includes: a computer, a pressure sensor signal acquisition device, a high and low temperature test chamber, and a high-precision pressure source. The computer includes the pressure sensor error calibration device based on machine learning described above. The computer is connected to both the pressure sensor signal acquisition device and the high and low temperature test chamber. The pressure sensor signal acquisition device is connected to the high and low temperature test chamber. The computer controls the temperature inside the high and low temperature test chamber within a preset temperature range. The high-precision pressure source applies different pressures to the pressure sensor in the high and low temperature test chamber at each operating temperature value, enabling the pressure sensor signal acquisition device to acquire multiple measured pressure values ​​output by the pressure sensor. The pressure sensor signal acquisition device uploads the acquired multiple measured pressure values ​​to the computer, which then completes the pressure sensor error calibration process based on machine learning.

[0098] Specifically, a standard pressure source with higher accuracy than the pressure sensor is used to provide a reference for calibration, and a pressure sensor signal acquisition device is used to acquire the output signal of the pressure sensor in the high and low temperature test chamber. The high-precision pressure source, the pressure sensor signal acquisition device, and the high and low temperature test chamber are all automatically controlled by a computer.

[0099] The following is a specific implementation example, which provides a detailed explanation of the specific implementation process of the machine learning-based pressure sensor error calibration method provided by the present invention.

[0100] 1. Taking the operating temperature range of the pressure sensor as -40-85℃ as an example, the full scale of the pressure sensor is 130kPa. The pressure sensor's operating temperature values ​​T are collected at different operating temperatures: -40℃, -10℃, 10℃, 20℃, 50℃, and 85℃. i The output measured pressure value is obtained by applying the measured pressure P to the pressure sensor through a high-precision pressure source under each temperature condition. j The pressure sensor range was increased by 5% to 100% from 0% to a total of 21 pressure points, and data was collected at each (T) i ,P j The measured pressure value P output by the combined pressure sensor n The data is divided into 6 groups based on temperature conditions, D(T)i ,P j ,P n ), wherein i ∈ (-40℃, -10℃, 10℃, 20℃, 50℃, 85℃); j ∈ 1, 2, 3…, 21; n ∈ 1, 2, 3…, 21; the pressure sensor output characteristic curve at [-10, 50]℃ is selected as the training set.

[0101] 2. Normalize the working temperature value T i , the input pressure value P j and the measured pressure value Pn, respectively;

[0102] For example, the normalization formula of the working temperature value T i is as follows:

[0103]

[0104] 3. According to the two error sources of the pressure sensor, an error calibration model is established by using a machine learning algorithm, and then the error calibration model is trained according to the pressure sensor output characteristic curves sampled at [-40, 10, 20, 85]℃. The forward calculation is performed by using α = L -T ·(L -1 ·y) to avoid matrix inversion.

[0105] 4. After the error calibration model is trained, the output value P' of the error calibration model is:

[0106]

[0107] 5. The current measured pressure value and the current working temperature value output by the pressure sensor are input into the trained error calibration model for error calibration; wherein the current measured pressure value and the current working temperature value are normalized and standardized based on the pressure sensor output characteristic curve sampled at 20℃, simplified as (μ, T); then two pressure sensor output characteristic curves D and DT at adjacent temperatures are selected according to the working temperature value T in (μ, T).

[0108] If the to-be-calibrated data is sampled at temperature T = -10℃, the model selects the pressure sensor output characteristic curve D -40 at -40℃ and the pressure sensor output characteristic curve D 10 at 10℃.

[0109] D -40 and D 10 are respectively subjected to Gaussian process regression prediction with μ, and the output value μ -40and the output value μ of the Gaussian process regression at 10℃ 10 ;

[0110] According to the actual temperature distribution, the linear interpolation method is performed according to the pressure sensor output characteristic curve variation DT corresponding to -40℃ and 10℃ -40～10 μ' is calculated;

[0111]

[0112] Wherein, A is a proportional coefficient;

[0113] μ' is mapped back to the range of the pressure sensor, that is, the calibrated pressure value P' is obtained, P' = 130kPa * μ'; P' is the current calibrated pressure value corresponding to the current measured pressure value.

[0114] The machine learning-based pressure sensor error calibration method provided by the application is essentially a software compensation method, which starts from the physical source of the measurement error of the pressure sensor, combines the current advanced machine learning algorithm, and can well solve the data calibration problem of various types of pressure sensors, and improves the overall performance of the pressure sensor.

[0115] It can be understood that the above embodiments are only exemplary embodiments for illustrating the principles of the application, but the application is not limited thereto. Various modifications and improvements can be made by those skilled in the art without departing from the spirit and essence of the application, and these modifications and improvements are also regarded as the protection scope of the application.

Claims

1. A machine learning based method for calibrating errors of a pressure sensor, the method comprising: The pressure sensor is placed in a high-low temperature test chamber, the temperature in the high-low temperature test chamber is controlled in a preset temperature range, so that the working temperature of the pressure sensor is in the preset temperature range, a plurality of working temperature values in the preset temperature range are set, different pressures are sequentially applied to the pressure sensor by a high-precision pressure source at each working temperature value, so as to obtain a plurality of measured pressure values output by the pressure sensor at each working temperature value and corresponding input pressure values; the pressure sensor error calibration method based on machine learning comprises: Step S1: obtaining a training set; wherein the training set comprises a plurality of groups of sample data, each group of sample data comprising a working temperature value, an input pressure value at the working temperature value, and a measured pressure value corresponding to the input pressure value; Step S2: normalizing the training set to obtain a normalized training set; Step S3: establishing an error calibration model for pressure sensor error calibration; Step S4: training the error calibration model according to the normalized training set to obtain a trained error calibration model; Step S5: inputting a current working temperature value and a current measured pressure value into the trained error calibration model for error calibration to output a current calibrated pressure value corresponding to the current measured pressure value; Wherein, in the establishment of the error calibration model for pressure sensor error calibration, it further comprises: (1) using Gaussian process regression of machine learning algorithm to calibrate the manufacturing error of the pressure sensor: ， ， ， wherein, represents an output value of a Gaussian process regression, represents a covariance matrix, represents a covariance matrix a kernel function, represents a predicted output range, represents a function smoothness, represents a set of feature vectors composed of working temperature values and measured pressure values in the training set, represents an input pressure value in the training set, represents a normalized value corresponding to a new input pressure value; (2) The output characteristic curve change amount of the pressure sensor at different temperatures Calibrating the temperature drift of the pressure sensor: wherein a current operating temperature value of the pressure sensor, a previous operating temperature value of the pressure sensor, a next operating temperature value of the pressure sensor, a previous operating temperature value of the pressure sensor, a next operating temperature value of the pressure sensor, a pressure sensor output characteristic curve at an operating temperature value, a pressure sensor output characteristic curve at an operating temperature value, a pressure sensor output characteristic curve at an operating temperature value, a pressure sensor output characteristic curve at an operating temperature value, a pressure sensor output characteristic curve at an operating temperature value, a pressure sensor output characteristic curve at an operating temperature value, a pressure sensor output characteristic curve at an operating temperature value, a reference temperature value; (3) calibrating by linear interpolation method combined with the manufacturing error and temperature drift of the pressure sensor: wherein, representing the final pressure sensor calibration result, representing the working temperature value the output value of the corresponding Gaussian process regression, representing the working temperature value the output value of the corresponding Gaussian process regression, representing the measurement range of the pressure sensor.

2. The machine learning based pressure sensor error calibration method of claim 1, wherein, The normalization of the training set to obtain the normalized training set further comprises: (1) normalizing the working temperature value in the plurality of groups of sample data, the formula being as follows: wherein, representative of a working temperature value normalized value, representative of a minimum working temperature value of the pressure sensor, representative of a maximum working temperature value of the pressure sensor; (2) normalizing the input pressure value in the plurality of groups of sample data, the formula being as follows: wherein, representing an input pressure value a normalized value, representing a minimum input pressure value in the plurality of sets of sample data, representing a maximum input pressure value in the plurality of sets of sample data; (3) normalizing the measured pressure value in the plurality of groups of sample data, the formula being as follows: wherein, representing a measured pressure value a normalized value, representing a minimum measured pressure value in the plurality of sets of sample data, representing a maximum measured pressure value in the plurality of sets of sample data.

3. The machine learning based pressure sensor error calibration method of claim 1, wherein, In said step S4 and step S5, the symmetric positive definite covariance matrix is decomposed into the product of two lower triangular matrices and ​ wherein and represent two lower triangular matrices; to obtain .

4. A machine learning based pressure sensor error calibration apparatus for implementing the machine learning based pressure sensor error calibration method of any one of claims 1 to 3, characterized by, The pressure sensor error calibration device based on machine learning comprises: An acquisition module for acquiring a training set; wherein the training set comprises a plurality of groups of sample data, each group of sample data comprising a working temperature value, an input pressure value at the working temperature value, and a measured pressure value corresponding to the input pressure value; A normalization module for normalizing the training set to obtain a normalized training set; An establishment module for establishing an error calibration model for pressure sensor error calibration; A training module for training the error calibration model according to the normalized training set to obtain a trained error calibration model; A calibration module for inputting a current working temperature value and a current measured pressure value into the trained error calibration model for error calibration to output a current calibrated pressure value corresponding to the current measured pressure value.

5. A machine learning based pressure sensor error calibration system, comprising: Comprise: The computer, pressure sensor signal acquisition equipment, high and low temperature test box and high precision pressure source, the computer includes the pressure sensor error calibration device based on machine learning in claim 4, the computer is connected with the pressure sensor signal acquisition equipment and the high and low temperature test box respectively, the pressure sensor signal acquisition equipment and the high and low temperature test box are connected, the computer controls the temperature in the high and low temperature test box in the preset temperature range, the high precision pressure source is used to sequentially apply different pressures to the pressure sensor in the high and low temperature test box at each working temperature value, so that the pressure sensor signal acquisition equipment can collect multiple measured pressure values output by the pressure sensor, the pressure sensor signal acquisition equipment uploads the multiple measured pressure values collected to the computer, and the computer can complete the pressure sensor error calibration process based on machine learning.

Images