Method and system for detecting performance of pressure sensor

Abstract

The embodiment of the present application relates to the technical field of data analysis, in particular to a pressure sensor performance detection method and system, the embodiment of the present application can classify and identify various performance detection data based on the structure of target multivariate decision tree detection attribute identification branch, and ensure the accuracy and pertinence of data acquisition. The third performance detection response information is obtained by determining the confidence degree relationship list with the aid of the target mapping relationship network matched with the target identification unit, the complex relationship between various detection information can be fully considered, and the accuracy of evaluation is improved. The fourth performance detection response information is obtained by processing the second performance detection response information by the correlation identification unit, and the overall analysis of different performance data can be realized by the cooperation of multiple units. The performance index evaluation data is determined based on the third and fourth performance detection response information, which can provide comprehensive, accurate and scientific detection evaluation results for the performance of the pressure sensor.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and more particularly, to a pressure sensor performance detection method and system. BACKGROUND

[0002] In the field of pressure sensor performance detection, traditional detection methods often have many problems. When facing multi-dimensional sensor performance detection data, the traditional detection means lack effective classification and recognition mechanisms, and cannot accurately obtain key information from complex data. Moreover, when analyzing and evaluating, the complex correlation between information is not considered, resulting in inaccurate evaluation results. In addition, there is a lack of multi-unit collaborative analysis method, which cannot comprehensively and deeply analyze the performance of the pressure sensor in various aspects, so as to realize comprehensive and scientific performance index evaluation. SUMMARY

[0003] Therefore, the present application provides a pressure sensor performance detection method and system.

[0004] The embodiment of the present application provides a pressure sensor performance detection method, which is applied to a pressure sensor performance detection system, and the method comprises the following steps: acquiring first performance detection response information in multi-dimensional sensor performance detection data through a target recognition unit, wherein the target recognition unit is one detection attribute recognition branch in a target multi-element decision tree, and the detection attribute recognition branch comprises the target recognition unit and at least one associated recognition unit, and at least one associated recognition unit is used for processing at least one second performance detection response information determined based on the multi-dimensional sensor performance detection data; determining a confidence relationship list paired with the first performance detection response information based on at least one target mapping relationship network paired with the target recognition unit, and determining third performance detection response information based on the first performance detection response information and the confidence relationship list, wherein the relationship network size of the target mapping relationship network matches the detection response vector size of the first performance detection response information; acquiring at least one fourth performance detection response information obtained by processing the second performance detection response information corresponding to each associated recognition unit by at least one associated recognition unit; and determining performance index evaluation data paired with the multi-dimensional sensor performance detection data based on the third performance detection response information and at least one fourth performance detection response information.

[0005] The present application also provides a pressure sensor performance detection system, comprising: a memory for storing program instructions and data; a processor coupled with the memory for executing instructions in the memory to implement the method as described above.

[0006] The application further provides a computer storage medium comprising instructions which, when executed on a processor, implement the method described above.

[0007] The embodiment of the application can classify and identify various performance detection data based on the structure of the target multi-element decision tree detection attribute identification branch, ensure the accuracy and pertinence of data acquisition, determine the confidence degree relationship list by means of the target mapping relationship network matched with the target identification unit to obtain the third performance detection response information, fully consider the complex relationship between various detection information, and improve the accuracy of evaluation. The fourth performance detection response information is obtained by processing the second performance detection response information by the correlation identification unit, and the comprehensive analysis of different performance data can be realized by the multi-unit cooperation. The performance index evaluation data is determined based on the third and fourth performance detection response information, which can provide comprehensive, accurate and scientific detection evaluation results for the performance of the pressure sensor. BRIEF DESCRIPTION OF DRAWINGS

[0008] The accompanying drawings, which are incorporated herein and constitute part of the specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the principles of the application.

[0009] Figure 1 A step flowchart of a pressure sensor performance detection method provided by the embodiment of the application.

[0010] Figure 2 A structural block diagram of a pressure sensor performance detection system provided by the embodiment of the application. DETAILED DESCRIPTION

[0011] The technical solutions in the application will be described below with reference to the drawings.

[0012] The following description refers to the accompanying drawings. Unless otherwise noted, the same numbers in different drawings indicate the same or similar components. The following examples of embodiments are described in detail with reference to the drawings. They are not representative of all embodiments consistent with the application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the application. Note that the terms "first", "second", and the like in the specification and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence.

[0013] The method provided by the embodiment of the present application can be executed in a pressure sensor performance detection system, a computer device or similar operation device. Taking the case of being executed in the pressure sensor performance detection system, the pressure sensor performance detection system can include one or more processors (the processor can include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory for storing data. Optionally, the pressure sensor performance detection system can further include a transmission device for communication function. Those skilled in the art can understand that the above structure is only schematic and does not limit the structure of the pressure sensor performance detection system. For example, the pressure sensor performance detection system can further include more or less components than those shown above or have a different configuration from that shown above.

[0014] The memory can be used to store computer programs, for example, software programs of application software and modules, such as a computer program corresponding to the pressure sensor performance detection method of the embodiment of the present application. The processor executes various functions and data processing by running the computer program stored in the memory, that is, implements the above method. The memory can include a high-speed random access memory and can further include a non-volatile memory, such as one or more magnetic storage devices, a flash memory or other non-volatile solid-state memory. In some examples, the memory can further include a memory remotely arranged with respect to the processor, which can be connected to the pressure sensor performance detection system through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0015] The transmission device is used to receive or send data via a network. Specific examples of the above network can include a wireless network provided by a communication provider of the pressure sensor performance detection system. In one example, the transmission device includes a network adapter (Network Interface Controller, NIC) which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (Radio Frequency, RF) module which is used to communicate with the Internet in a wireless manner.

[0016] In combination with the above, please refer to Figure 1 , Figure 1 is a flowchart of a pressure sensor performance detection method provided by the embodiment of the present application. The method is applied to a pressure sensor performance detection system and can further include steps 110-140.

[0017] Step 110: obtaining first performance detection response information in multi-dimensional sensor performance detection data through a target identification unit.

[0018] The target recognition unit is one of the detection attribute recognition units in a detection attribute recognition branch of a target multi-element decision tree, and the detection attribute recognition branch includes the target recognition unit and at least one associated recognition unit, and the at least one associated recognition unit is used to process at least one second performance detection response information determined based on the multi-dimensional sensing performance detection data. Further, the multi-dimensional sensing performance detection data includes static performance detection data, dynamic performance detection data, environmental adaptability detection data, and durability stability detection data of the target pressure sensor.

[0019] Step 120: determining a confidence relationship list paired with the first performance detection response information based on at least one target mapping relationship network paired with the target recognition unit, and determining third performance detection response information based on the first performance detection response information and the confidence relationship list.

[0020] The relationship network size of the target mapping relationship network matches the detection response vector size of the first performance detection response information.

[0021] Step 130: obtaining at least one fourth performance detection response information obtained by processing the respective second performance detection response information by the at least one associated recognition unit.

[0022] Step 140: determining performance index evaluation data paired with the multi-dimensional sensing performance detection data based on the third performance detection response information and the at least one fourth performance detection response information.

[0023] It can be understood that the pressure sensor performance detection system is the execution subject of the embodiment of the present application, and is used to implement the above steps 110-140 in the whole detection and evaluation process.

[0024] Firstly, in step 110, the target recognition unit starts to acquire the first performance detection response information in the multi-dimensional sensor performance detection data. The multi-dimensional sensor performance detection data in the embodiment of the present application covers very comprehensive content, including static performance detection data, dynamic performance detection data, environmental adaptability detection data and durability stability detection data for the target pressure sensor. For example, in terms of static performance detection data, it may contain measurement error data in accuracy detection, such as the deviation of the measured value of a certain pressure sensor under a certain pressure from the true value, for example, the measured value is 100.5 kPa under 100 kPa pressure, and the true value is 100 kPa. The deviation data can be part of the content in accuracy detection; it may also contain the dispersion of output values when measuring the same pressure multiple times in repeatability detection, such as measuring 50 kPa pressure 10 times repeatedly, the output value of each time may have a certain fluctuation range. In dynamic performance detection data, there will be frequency response detection related data, such as the amplitude-frequency characteristic data of the pressure sensor to different frequency pressure signals, which may be in the frequency range of 100 Hz-1000 Hz, the output amplitude changes with the frequency, and the output amplitude of some sensors decreases obviously at 500 Hz compared with low frequency; and the data such as rise time and overshoot in step response detection, for example, the rise time of a certain sensor is 2 ms, and the overshoot is 15%. The environmental adaptability detection data involves temperature drift data under temperature influence detection, such as zero point temperature drift of ±0.3 mV / °C and sensitivity temperature drift of ±0.05% / °C in the range of-20°C to +60°C; humidity influence coefficient data under humidity influence detection, for example, the ratio of the change amount of sensor output to full-scale output when the humidity changes from 30% RH to 80% RH; and vibration stability data in vibration and impact influence detection, such as the ratio of the root mean square value of output fluctuation to full-scale output under 10 Hz-1000 Hz, 1g amplitude sinusoidal vibration, which is 0.5%. The durability stability detection data contains aging rate data in aging test, for example, after 1500 hours of aging test, the sensor sensitivity changes from 1.2 mV / Pa to 1.18 mV / Pa, and the aging rate is 1.67%. And the target recognition unit is a detection attribute recognition unit in the detection attribute recognition branch of the target multi-element decision tree. In addition to the target recognition unit, at least one associated recognition unit is included in the detection attribute recognition branch, and these associated recognition units will process at least one second performance detection response information determined based on the multi-dimensional sensor performance detection data respectively.

[0025] Then, step 120 is entered. Based on the at least one target mapping relationship network matched with the target recognition unit, a confidence relationship list matched with the first performance detection response information is determined, and then the third performance detection response information is determined based on the first performance detection response information and the confidence relationship list. The relationship network size of the target mapping relationship network in the embodiment of the present application matches the detection response vector size of the first performance detection response information. For example, the detection response vector of the first performance detection response information contains 5-dimensional data, such as the relative error in accuracy detection, the standard deviation in repeatability detection, the amplitude variation of the amplitude-frequency characteristic at a specific frequency in dynamic performance detection, the temperature drift in environmental adaptability detection, and the aging rate in durability stability detection. Then the target mapping relationship network will also have a corresponding relationship structure to match the 5-dimensional data. This target mapping relationship network can be constructed based on a large amount of experimental data and empirical data. For example, when a plurality of pressure sensors were tested previously, data of sensors of different brands and different models under various performance detections were collected, the data were sorted and analyzed, the correlation between the performance detection response information and the performance was found out, and the target mapping relationship network was constructed. Through the target mapping relationship network, the confidence relationship list corresponding to the first performance detection response information can be obtained. For example, if the relative error in the first performance detection response information is 2%, the corresponding confidence relationship list in the target mapping relationship network can show that the confidence of the sensor performance at a medium level is 0.6, the confidence at a good level is 0.3, and the confidence at a poor level is 0.1 under this relative error. Then the third performance detection response information is further determined according to the confidence relationship list and the first performance detection response information. The third performance detection response information is more comprehensive and more accurate information reflecting the sensor performance after comprehensively considering the first performance detection response information and the corresponding confidence relationship.

[0026] Then, in step 130, at least one fourth performance detection response information is obtained by at least one correlation identification unit processing the respective corresponding second performance detection response information. The correlation identification unit in the embodiment of the present application also processes the second performance detection response information based on the multi-dimensional sensing performance detection data. For example, the correlation identification unit can be specially designed to process the step response detection part in the dynamic performance detection data. For example, in the step response detection, the rise time, overshoot and settling time of the pressure sensor are the second performance detection response information, and the correlation identification unit can analyze and process these data according to the pre-set algorithm or rule. For example, for the rise time, the correlation identification unit can compare it with the average rise time of the same type of sensor, and if the rise time of the sensor is 30% longer than the average rise time, it will be marked as a slow rise time, and the fourth performance detection response information will be generated according to the result; for the overshoot, if the overshoot exceeds the set threshold (such as 20%), the correlation identification unit will determine that the overshoot is too large, and this will be reflected in the fourth performance detection response information. Similarly, for the humidity influence detection part in the environmental adaptability detection data, the correlation identification unit can analyze the change trend of the sensor output under different humidity, and if the change amount of the sensor output exceeds 10% of the full-scale output when the humidity rises from 40% RH to 70% RH, the sensor will be recorded in the fourth performance detection response information as being sensitive to humidity.

[0027] Finally, in step 140, based on the third performance detection response information and the at least one fourth performance detection response information, the performance index evaluation data paired with the multi-dimensional sensing performance detection data is determined. For example, the third performance detection response information contains the preliminary evaluation result of the sensor in the comprehensive performance according to the first performance detection response information and the confidence relationship list, which may be that the overall performance is at a slightly higher level. The fourth performance detection response information may contain the problems of slow rise time, large overshoot in the step response of the dynamic performance of the sensor, and sensitivity to humidity in the humidity influence of the environmental adaptability. By integrating these information, the performance index evaluation data will comprehensively and accurately evaluate the performance of the pressure sensor. For example, the performance index evaluation data may conclude that although the pressure sensor performs well in static performance and part of dynamic performance, the overall performance can only reach the medium level due to the problems in step response and humidity influence, and attention should be paid to the control of humidity environment or compensation for step response in actual application. In this way, the complete process from multi-dimensional sensing performance detection data to final performance index evaluation data based on the pressure sensor performance detection system is completed, which can provide comprehensive, scientific and accurate basis for the performance evaluation of the pressure sensor.

[0028] From another perspective, the process can be viewed. In step 110, when the first performance detection response information is acquired, for the linearity part of the static performance detection data, taking a certain pressure sensor as an example, after fitting a straight line through the data measured at multiple pressure points, the maximum deviation of the actual output-input characteristic curve from the fitted straight line is calculated to obtain the linearity. For example, in the pressure range of 0-1000 kPa, the maximum deviation is 5 kPa, and the full-scale output value is 10V, so the linearity is ±0.5%. This linearity data is part of the first performance detection response information acquired by the target identification unit. In step 120, the target mapping relationship network may have a corresponding association for such linearity data. For example, in the relationship network, linearity within ±0.5% may be associated with a higher performance confidence, and may be displayed in the confidence relationship list that under such linearity, the confidence of the sensor performance is 0.8, the medium is 0.2, and the poor is 0. According to such confidence relationship list and linearity data, the third performance detection response information determined will contain more accurate performance evaluation tendency. In step 130, when the correlation identification unit processes the frequency response detection data in the dynamic performance detection data as the second performance detection response information, for example, the amplitude-frequency characteristic curve of a certain sensor drops rapidly in the frequency range of 1000-2000 Hz, the correlation identification unit may mark this frequency range as the sensitive frequency area of the sensor, and reflect it in the fourth performance detection response information. Finally, in step 140, by comprehensively considering the performance evaluation tendency of linearity in the third performance detection response information and the situation of the sensitive frequency area in the fourth performance detection response information, the performance index evaluation data can conclude that the sensor has good performance in the low frequency band, but may have problems in the high frequency band, which needs to be paid attention to in specific application scenarios.

[0029] In processing environmental adaptability detection data, such as zero point temperature drift in temperature influence detection. For example, a certain pressure sensor has a zero point temperature drift of ±0.2 mV / °C in the temperature range of -10°C to +40°C. In step 110, this zero point temperature drift data is part of the multi-dimensional sensor performance detection data, which can be acquired by the target identification unit as the first performance detection response information. In step 120, the target mapping relationship network can show that the sensor has better performance in temperature stability according to this zero point temperature drift data in the confidence relationship list, for example, the performance of good confidence is 0.7, the medium is 0.3, and the poor is 0. The corresponding third performance detection response information will consider this confidence relationship. In step 130, when the correlation identification unit processes humidity influence detection data, if it is found that the sensor has a humidity influence coefficient of 0.03 (i.e. the ratio of the change in sensor output to the full-scale output when the humidity changes by a certain value) when the humidity changes from 20% RH to 60% RH, the sensor has a certain anti-interference ability to humidity. Finally, in step 140, the performance index evaluation data concludes that the sensor performs better in temperature adaptability and has certain ability in humidity adaptability, but may need further optimization.

[0030] For the aging rate in durable stability detection data, for example, the sensitivity of a certain pressure sensor changes from 1.1 mV / Pa to 1.08 mV / Pa after 2000 hours of aging test, and the aging rate is 1.82%. In step 110, this aging rate data is acquired by the target identification unit. In step 120, the target mapping relationship network displays that the sensor is at a medium level in durable stability according to the aging rate in the confidence relationship list, for example, the performance of good confidence is 0.3, the medium is 0.5, and the poor is 0.2. The third performance detection response information will contain this confidence-based evaluation. In step 130, if there is no other correlation processing (no other special processing) specifically for the aging rate, the fourth performance detection response information is mainly the result of other correlation identification units. Finally, in step 140, the performance index evaluation data combines the medium level evaluation of durable stability in the third performance detection response information and other results in the fourth performance detection response information to comprehensively conclude the overall performance evaluation of the sensor, for example, the overall performance is at a medium level, and needs to be calibrated regularly to ensure measurement accuracy in long-term use.

[0031] In the step response detection part of the dynamic performance detection data, in addition to the rise time and overshoot mentioned above, the settling time is also an important indicator. For example, the settling time of a certain pressure sensor is 10 ms, and this settling time data is part of the multi-dimensional sensor performance detection data obtained by the target recognition unit in step 110. In step 120, the target mapping relationship network displays that the sensor is at a normal level in the settling time of the step response according to the settling time data in the confidence level list, for example, the confidence level of good performance is 0.5, the confidence level of medium performance is 0.4, and the confidence level of poor performance is 0.1. The third performance detection response information takes into account this confidence level. In step 130, when the correlation recognition unit processes the settling time data and other step response data, it may comprehensively judge the overall performance of the sensor in the step response, which is reflected in the fourth performance detection response information. Finally, in step 140, the performance index evaluation data synthesizes the third performance detection response information and the fourth performance detection response information, and concludes that although the settling time of the sensor in the step response is normal, the overall step response performance needs to be improved due to problems in the rise time or overshoot.

[0032] In the whole process, the pressure sensor performance detection system detects and evaluates the performance of the pressure sensor from multiple dimensions through the cooperative work of various units and links, so that the evaluation result is comprehensive and accurate.

[0033] Further, the first performance detection response information is obtained by the target recognition unit from the multi-dimensional sensor performance detection data. The multi-dimensional sensor performance detection data includes static performance detection data of the pressure sensor (such as accuracy, repeatability, resolution, etc.), dynamic performance detection data (such as frequency response, step response related data), environmental adaptability detection data (temperature, humidity, vibration and impact detection data), and durability and stability detection data (such as aging test data). The first performance detection response information is part of the multi-dimensional sensor performance detection data filtered by the target recognition unit. For example, the target recognition unit may extract the measurement error value at a certain specific pressure from the accuracy detection data, or select the amplitude-frequency characteristic value at a certain frequency from the dynamic performance detection data, etc. These values are the basis for subsequent analysis and processing.

[0034] The second performance detection response information is determined based on the multi-dimensional sensing performance detection data and is processed by at least one correlation identification unit. These correlation identification units are in the same detection attribute identification branch as the target identification unit. Similar to the first performance detection response information, the second performance detection response information is also part of the multi-dimensional sensing performance detection data, but is processed by the correlation identification unit. For example, the correlation identification unit may focus on the step response part of the dynamic performance detection data, such as the rise time, overshoot, and settling time data as the second performance detection response information, or the humidity influence detection part in the environmental adaptability detection data, such as the change in sensor output when the humidity changes, as the second performance detection response information.

[0035] The third performance detection response information is determined based on the first performance detection response information and the confidence relationship list paired therewith. First, the confidence relationship list paired with the first performance detection response information is determined according to the target mapping relationship network paired with the target identification unit, and then the third performance detection response information is obtained by comprehensively considering the first performance detection response information and the confidence relationship list. The third performance detection response information is the result after considering the confidence relationship on the basis of the first performance detection response information. For example, if a certain index in the first performance detection response information shows the performance value of the sensor in a certain aspect, a more comprehensive and more reflective result of the sensor performance is obtained through the confidence of different performance levels (such as good, medium, and poor) corresponding to the value in the confidence relationship list, which contains more performance evaluation tendency information.

[0036] The fourth performance detection response information is obtained by processing the respective second performance detection response information by at least one correlation identification unit. The fourth performance detection response information is the result of processing the second performance detection response information by the correlation identification unit. For example, the correlation identification unit analyzes and processes the amplitude-frequency characteristics and phase-frequency characteristics of the sensor at different frequencies for the frequency response detection part in the dynamic performance detection data, and obtains information about the overall performance evaluation of the sensor in the frequency response aspect, which constitutes the fourth performance detection response information.

[0037] In some possible examples, the target multi-element decision tree is a decision structure for analyzing and deciding the performance detection data of the pressure sensor. It contains multiple branches and nodes, processes the input data through different paths and judgment rules to obtain the evaluation result of the performance of the pressure sensor. In the embodiments of the present application, it provides a framework for the analysis of the performance detection data of the pressure sensor. For example, different branches can correspond to different types of performance detection data (static performance, dynamic performance, etc.), and nodes can represent different judgment conditions or data processing steps, so that the performance evaluation result can be gradually obtained according to the multi-dimensional sensing performance detection data of the pressure sensor.

[0038] The detection attribute identification branch is a component of the target multi-decision tree, which includes a detection attribute identification unit and a correlation identification unit. This branch is mainly responsible for identifying and processing specific attributes in the performance detection data of the pressure sensor. It classifies and processes the performance detection data of the pressure sensor according to different attributes. For example, a detection attribute identification branch may be dedicated to processing dynamic performance detection data, in which the detection attribute identification unit and the correlation identification unit identify, analyze and process different parts of the dynamic performance detection data, respectively, to help more detailed and targeted evaluation of the dynamic performance of the pressure sensor.

[0039] The detection attribute identification unit is a unit in the detection attribute identification branch, and in the embodiment of the present application, the target identification unit can be a detection attribute identification unit. Its main function is to obtain specific performance detection response information from multi-dimensional sensor performance detection data. The detection attribute identification unit is responsible for preliminary screening and extraction of multi-dimensional sensor performance detection data. For example, in processing static performance detection data, the detection attribute identification unit can accurately extract meaningful parts from data in multiple aspects such as accuracy detection and repeatability detection, such as extracting measurement error values from accuracy detection data, to provide basic data for subsequent analysis and processing.

[0040] The target mapping relationship network is a data relationship structure, and its relationship network size matches the detection response vector size of the first performance detection response information. It stores the mapping relationship between the first performance detection response information and the performance evaluation, which is constructed based on a large amount of experimental data, empirical data or pre-set rules. The target mapping relationship network is used to determine the confidence relationship list paired with the first performance detection response information. For example, when the first performance detection response information contains the relative error value in the accuracy detection of a pressure sensor, the target mapping relationship network can determine the confidence relationship list of the sensor performance at different levels (such as good, medium, poor) under this relative error value according to the pre-stored relationship between the relative error value and the performance evaluation.

[0041] The confidence relationship list is paired with the first performance detection response information, which lists the confidence values of the sensor performance at different levels (such as good, medium, poor, etc.) based on the first performance detection response information. These confidence values are determined based on the target mapping relationship network. The confidence relationship list provides a basis for determining the third performance detection response information. Through the confidence values in the confidence relationship list, combined with the actual data in the first performance detection response information, the performance of the pressure sensor can be more comprehensively and accurately evaluated, and the third performance detection response information containing the performance evaluation tendency can be obtained.

[0042] On the basis of the foregoing, exemplary descriptions of different performance detection response information and performance index evaluation data are as follows.

[0043] I. Feature vector form of first performance detection response information

[0044] Vector constituent elements

[0045] 1) Static performance related elements;

[0046] For example, taking a pressure sensor as an example, in terms of accuracy, if the measurement error is 3%, this value can be an element in the feature vector. In terms of repeatability, for example, the standard deviation of the output value of 10 measurements of a certain pressure value is 0.5 units (this unit is determined according to the specific physical quantity of the sensor output, such as millivolts of voltage value, etc.), which can also be an element. In addition, if the resolution is 0.05 units (such as pressure units in pascal), it can also be included in the vector.

[0047] 2) Dynamic performance related elements;

[0048] For frequency response, the amplitude-frequency characteristic value at a certain frequency, such as the amplitude-frequency characteristic value at 100 Hz is 0.8 (for example, it is a normalized value relative to the low-frequency amplitude value), and the phase-frequency characteristic value (for example, the phase shift at 100 Hz is 10 degrees, which can be converted into a suitable numerical form and included in the vector). For step response, the rise time is 2 ms, the overshoot is 15%, the settling time is 10 ms, etc. These values can be elements in the feature vector.

[0049] 3) Environmental adaptability related elements;

[0050] In terms of temperature influence, for example, the zero-point temperature drift is ±0.3 mV / °C, and the sensitivity temperature drift is ±0.05% / °C, which can become vector elements. For humidity influence, if the humidity changes from 30% RH to 80% RH, the humidity influence coefficient is 0.03 (the ratio of the change in sensor output to the full-scale output), which can be included in the vector. In terms of vibration and impact influence, for example, under a sinusoidal vibration of 10 Hz-1000 Hz and an amplitude of 1 g, the ratio of the root mean square value of the output fluctuation to the full-scale output is 0.5%, which can be a vector element.

[0051] 4) Durability stability related elements;

[0052] For example, after 1500 hours of aging test, the sensor sensitivity changes from 1.2 mV / Pa to 1.18 mV / Pa, and the aging rate is 1.67%, which can be an element in the feature vector.

[0053] For example, the first performance detection response information feature vector of a pressure sensor is [3%, 0.5, 0.05, 0.8, 10, 2, 15%, 10, ±0.3, ±0.05%, 0.03, 0.5%, 1.67%], and the element order in the embodiment corresponds to the above-mentioned accuracy measurement error, repeatability standard deviation, resolution, 100Hz amplitude-frequency characteristic, 100Hz phase-frequency characteristic, rise time, overshoot, settling time, zero point temperature drift, sensitivity temperature drift, humidity influence coefficient, vibration output fluctuation ratio, and aging rate.

[0054] II. Feature vector form of second performance detection response information

[0055] Vector constituent elements: different correlation identification unit processing result elements;

[0056] For example, if a correlation identification unit is specially used to process the frequency response part of dynamic performance, the amplitude-frequency characteristic and phase-frequency characteristic values of multiple frequency points can be used as elements. For example, the amplitude-frequency characteristic values at 200Hz, 500Hz, and 1000Hz are 0.7, 0.6, and 0.5, and the phase-frequency characteristic values are 15 degrees, 20 degrees, and 30 degrees (converted into a suitable numerical form), which can constitute vector elements.

[0057] Another correlation identification unit processes the humidity influence part in environmental adaptability, and the sensor output change when the humidity changes in different intervals can be used as an element. For example, when the humidity changes from 40%RH-50%RH, 50%RH-60%RH, and 60%RH-70%RH, the ratio of the sensor output change to the full-scale output is 0.02, 0.03, and 0.04, respectively, which can become vector elements.

[0058] For example, for the case where one correlation identification unit processes frequency response and another correlation identification unit processes humidity influence, the second performance detection response information feature vector can be [0.7, 15, 0.6, 20, 0.5, 30, 0.02, 0.03, 0.04], and the element order in the embodiment corresponds to the 200Hz amplitude-frequency characteristic, 200Hz phase-frequency characteristic, 500Hz amplitude-frequency characteristic, 500Hz phase-frequency characteristic, 1000Hz amplitude-frequency characteristic, 1000Hz phase-frequency characteristic, 40-50%RH humidity influence coefficient, 50-60%RH humidity influence coefficient, and 60-70%RH humidity influence coefficient.

[0059] III. Feature vector form of third performance detection response information

[0060] Vector constituent elements: comprehensive performance elements based on confidence relationship;

[0061] The third performance detection response information is obtained based on the first performance detection response information and the confidence relationship list. For example, after comprehensively considering the values of the accuracy, dynamic performance, environmental adaptability, and durability stability in the first performance detection response information and the corresponding confidence relationships, quantitative values about the overall performance at different levels (such as good, medium, and poor) are obtained. For example, the confidence of the overall performance being good is 0.6, the confidence of the overall performance being medium is 0.3, and the confidence of the overall performance being poor is 0.1. These values can be used as vector elements.

[0062] Meanwhile, the weighted comprehensive result of the key performance indicators in the first performance detection response information can also be included. For example, a comprehensive value calculated according to certain weights (for example, the weight of accuracy is 0.3, the weight of dynamic performance is 0.3, the weight of environmental adaptability is 0.2, and the weight of durability stability is 0.2) of the accuracy, dynamic performance, environmental adaptability, and durability stability, such as 0.7 (this value is a value reflecting the comprehensive performance tendency obtained according to specific weighted calculation rules), can also be used as a vector element.

[0063] For example, the third performance detection response information feature vector is [0.6, 0.3, 0.1, 0.7], and the element order in the embodiment of the application corresponds to the confidence values of the overall performance being good, medium, and poor, and the comprehensive performance tendency value, respectively.

[0064] Four, the feature vector form of the fourth performance detection response information

[0065] Vector constituent elements: comprehensive elements of the results processed by the correlation recognition unit;

[0066] For example, a quantitative value obtained by processing the step response part in the dynamic performance by the correlation recognition unit, after comprehensively considering the rise time, overshoot, and settling time. For example, according to a pre-set rule, a value of 0.8 is calculated by comprehensively calculating the rise time of 2 ms, the overshoot of 15%, and the settling time of 10 ms (this value represents the comprehensive performance in the step response aspect, and the calculation rule can be determined according to experience or experiment), which can be used as a vector element.

[0067] For the vibration influence part in the environmental adaptability processed by the correlation recognition unit, a value obtained by comprehensively considering the sensor output stability under different vibration frequencies and amplitudes. For example, in the range of 10 Hz-100 Hz and the amplitude of 0.5 g-1 g, a value of 0.6 is calculated according to the sensor output fluctuation (which represents the comprehensive performance under this vibration condition), which can be used as a vector element.

[0068] For example, the fourth performance detection response information feature vector is [0.8, 0.6], and the element order in the embodiment of the application corresponds to the step response comprehensive performance value and the vibration influence comprehensive performance value, respectively.

[0069] V. Performance index evaluation data in the form of a feature vector

[0070] Vector constituent elements: comprehensive evaluation result elements

[0071] The performance index evaluation data is obtained based on the third performance detection response information and the fourth performance detection response information. For example, the values of the final performance evaluation of the pressure sensor are obtained by comprehensively considering the overall performance confidence value and the comprehensive performance tendency value in the third performance detection response information and the comprehensive performance values processed by each associated recognition unit in the fourth performance detection response information. For example, the probabilities of the final performance evaluation being excellent, good, medium, poor, and bad are 0.2, 0.3, 0.3, 0.1, and 0.1, respectively, and these values can be used as vector elements.

[0072] At the same time, some suggestions or quantitative values of indicators given according to the performance evaluation result can also be included. For example, if the evaluation result is medium, a quantitative improvement suggestion value is given for the aspect that needs to be improved. For example, in the aspect of dynamic performance improvement, the improvement amount is 0.2 (this value represents the degree of improvement needed in the dynamic performance, which is obtained according to the specific evaluation rules), and this value can also be used as a vector element.

[0073] For example, the performance index evaluation data feature vector is [0.2, 0.3, 0.3, 0.1, 0.1, 0.2], and the order of the elements in the embodiment of the application corresponds to the probability values of excellent, good, medium, poor, and bad, respectively, and the dynamic performance improvement suggestion value.

[0074] In a preferred embodiment, the confidence relationship list paired with the first performance detection response information is determined based on at least one target mapping relationship network paired with the target identification unit, and the third performance detection response information is determined based on the first performance detection response information and the confidence relationship list, comprising: determining a detection element mapping relationship network included in at least one of the target mapping relationship networks, evaluating a mapping relationship network and a performance correction mapping relationship network, wherein the detection element mapping relationship network, the evaluation mapping relationship network and the performance correction mapping relationship network have a relationship network size matching a detection response vector size of the first performance detection response information; determining a response element mapping knowledge set based on the first performance detection response information and the detection element mapping relationship network, and determining a response evaluation mapping knowledge set based on the first performance detection response information and the evaluation mapping relationship network; determining the confidence relationship list paired with the first performance detection response information according to the response element mapping knowledge set and the response evaluation mapping knowledge set; determining a response correction mapping knowledge set based on the first performance detection response information and the performance correction mapping relationship network, and determining the third performance detection response information according to the confidence relationship list and the response correction mapping knowledge set.

[0075] In the embodiment, for determining the confidence relationship list paired with the first performance detection response information based on at least one target mapping relationship network paired with the target identification unit, and then determining the third performance detection response information, first, a detection element mapping relationship network, an evaluation mapping relationship network and a performance correction mapping relationship network included in at least one of the target mapping relationship networks are determined. The relationship network size of these relationship networks matches the detection response vector size of the first performance detection response information. Taking the first performance detection response information feature vector mentioned above as an example, for example, the first performance detection response information feature vector is [3%, 0.5, 0.05, 0.8, 10, 2, 15%, 10, ±0.3, ±0.05%, 0.03, 0.5%, 1.67%], the dimension of the vector determines that the structure dimension of the detection element mapping relationship network, the evaluation mapping relationship network and the performance correction mapping relationship network matches the dimension.

[0076] Based on the first performance detection response information and the detection element mapping relationship net, a response element mapping knowledge set is determined. The detection element mapping relationship net stores the mapping relationship between each element (such as the accuracy, repeatability, resolution, and other elements at different numerical values) in the first performance detection response information and other related knowledge (such as the mapping relationship between these elements and the internal structure, working principle, and other knowledge of the sensor). For each element in the first performance detection response information feature vector, such as the measurement error of 3%, the element may correspond to some knowledge about the error source and the influence range in the sensor measurement principle in the detection element mapping relationship net, and these knowledge combinations form the response element mapping knowledge set.

[0077] At the same time, based on the first performance detection response information and the evaluation mapping relationship net, a response evaluation mapping knowledge set is determined. The evaluation mapping relationship net contains the mapping relationship between different numerical values of the first performance detection response information elements and the evaluation standard. For example, for the element of the rise time of 2 ms in the first performance detection response information feature vector, it may correspond to the evaluation level (such as fast, normal, slow, etc.) of the rise time in the same type of sensor and other related evaluation knowledge corresponding to these levels in the evaluation mapping relationship net, and these related knowledge collectively constitutes the response evaluation mapping knowledge set.

[0078] According to the response element mapping knowledge set and the response evaluation mapping knowledge set, a confidence relationship list paired with the first performance detection response information is determined. Since the response element mapping knowledge set contains the principle knowledge behind each element of the first performance detection response information, and the response evaluation mapping knowledge set contains the evaluation related knowledge, the combination of the two can accurately determine the confidence relationship list. For example, according to the knowledge about the source and influence range of the measurement error in the response element mapping knowledge set, and the evaluation level knowledge of the measurement error in the same type of sensor in the response evaluation mapping knowledge set, the confidence values of the sensor performance at different levels (such as good, medium, and poor) when the measurement error is 3% can be determined, such as the confidence of good is 0.3, the confidence of medium is 0.5, and the confidence of poor is 0.2, which constitute the confidence relationship list.

[0079] Based on the first performance detection response information and the performance correction mapping relationship net, a response correction mapping knowledge set is determined. The performance correction mapping relationship net stores the mapping relationship between the numerical values of each element of the first performance detection response information and the possible correction measures or compensation knowledge. For example, for the element of the temperature drift of the sensitivity of ±0.05% / °C in the first performance detection response information feature vector, it may correspond to some technical means for compensating the temperature drift or knowledge on how to correct the influence in subsequent calculations in the performance correction mapping relationship net, and these knowledge combinations form the response correction mapping knowledge set.

[0080] Finally, the third performance detection response information is determined according to the confidence relationship list and the response correction mapping knowledge set. The third performance detection response information is not only based on the preliminary judgment of the sensor performance level in the confidence relationship list, but also combines the knowledge about how to correct or compensate the performance in the response correction mapping knowledge set. For example, based on the confidence relationship list showing that the sensor performance is at a medium level (e.g., good confidence is 0.3, medium confidence is 0.5, and poor confidence is 0.2), combined with the knowledge about correcting measurement errors, temperature drift, and other influences in the response correction mapping knowledge set, a more comprehensive third performance detection response information is obtained, which may include the performance adjustment tendency after considering the correction measures, such as the degree of performance approaching the good level after correction.

[0081] In this way, by dividing the target mapping relationship network into the detection element mapping relationship network, the evaluation mapping relationship network, and the performance correction mapping relationship network in detail, the first performance detection response information can be analyzed and processed more carefully and comprehensively. The confidence relationship list is accurately constructed by determining the response element mapping knowledge set and the response evaluation mapping knowledge set, so that the judgment of the sensor performance at different levels is more scientific and reasonable. The third performance detection response information is obtained by further determining the response correction mapping knowledge set, which not only accurately evaluates the current performance of the sensor, but also considers the performance correction factors, providing more instructive information for subsequent improvement of the sensor performance or optimization of the sensor use. This technical solution helps to improve the accuracy and comprehensiveness of the performance detection and evaluation of the pressure sensor, and provides reliable technical support for the performance optimization, quality control, and reasonable use of the pressure sensor in different application scenarios.

[0082] In an alternative embodiment 1, before obtaining at least one fourth performance detection response information processed by each corresponding second performance detection response information of at least one of the correlation identification units, it further includes: obtaining the u-th second performance detection response information in the multi-dimensional sensor performance detection data by the u-th correlation identification unit, wherein u is an integer not less than 1 and not greater than X, X is the number of correlation identification units included in the detection attribute identification branch; based on at least one correlation mapping relationship network paired with the u-th correlation identification unit, determine the confidence relationship list paired with the u-th second performance detection response information, and determine the u-th fourth performance detection response information based on the u-th second performance detection response information and the confidence relationship list.

[0083] In this alternative embodiment 1, before obtaining at least one fourth performance detection response information processed by at least one corresponding second performance detection response information of each corresponding associated recognition unit, the above-mentioned multi-dimensional sensing performance detection data is taken as an example, which contains a variety of data, and these data constitute the source of the second performance detection response information. First, the u-th associated recognition unit obtains the u-th second performance detection response information in the multi-dimensional sensing performance detection data. In the embodiment of the application, u is an integer, which is not less than 1 and not greater than X, X is the number of associated recognition units included in the detection attribute recognition branch. For example, if X = 3, when u = 1, the first associated recognition unit obtains its corresponding second performance detection response information from the multi-dimensional sensing performance detection data. The content of this information is similar to the first performance detection response information described above, which is also composed of a variety of performance detection related data. For example, if the associated recognition unit focuses on the step response part of the dynamic performance detection data, the second performance detection response information it obtains may contain values such as rise time, overshoot and settling time, for example, rise time is 3ms, overshoot is 10%, settling time is 8ms, etc. These values constitute part of the second performance detection response information.

[0084] Then, based on at least one associated mapping relationship network paired with the u-th associated recognition unit, a confidence relationship list paired with the u-th second performance detection response information is determined. The role of the associated mapping relationship network is similar to the target mapping relationship network mentioned earlier, and its structure is related to the u-th second performance detection response information. For example, for the second performance detection response information containing values such as rise time, overshoot and settling time, the associated mapping relationship network stores the relationship between these values and performance evaluation. Taking the rise time as an example, in the associated mapping relationship network, different ranges of rise time values correspond to different performance level confidences. For example, a rise time of 0-2ms may correspond to a confidence of 0.8 for good performance, 2-5ms may correspond to a confidence of 0.6 for medium performance, etc. There are similar corresponding relationships for overshoot and settling time, such as an overshoot of 0-5% may correspond to a confidence of 0.9 for good performance, a settling time of 0-5ms may correspond to a confidence of 0.7 for good performance, etc. By integrating these information, the confidence relationship list paired with the u-th second performance detection response information can be determined.

[0085] Then, the u-th fourth performance detection response information is determined based on the u-th second performance detection response information and the confidence relationship list. For example, for the second performance detection response information mentioned above with a rise time of 3 ms, an overshoot of 10%, and a settling time of 8 ms, the values in the confidence relationship list are combined. Since the rise time is 3 ms, according to the confidence relationship in the association mapping relationship network, the confidence of the performance at the medium level is high, the overshoot of 10% is also in the range of high confidence of the medium performance level, and the settling time of 8 ms is the same. In view of these circumstances, the u-th fourth performance detection response information is determined. This fourth performance detection response information is a comprehensive evaluation result of the second performance detection response information after considering the confidence relationship. It not only contains the original performance values, but also contains the evaluation tendency of these performance values according to the confidence relationship.

[0086] In this way, by separately obtaining the second performance detection response information for each association recognition unit and determining the confidence relationship list based on the associated mapping relationship network matched therewith, and then determining the fourth performance detection response information, the processing of the second performance detection response information is more detailed and accurate. This approach fully considers the particularity of the performance detection part concerned by each association recognition unit, improving the accuracy of performance evaluation of each part. For example, when processing the step response part in dynamic performance detection data, it can be accurately evaluated according to its own characteristics. At the same time, through the use of association mapping relationship network and confidence relationship list, the determination of the fourth performance detection response information is more scientific and reasonable, not only containing the information of the original performance values, but also incorporating the judgment tendency of the performance level, providing a more reliable basis for the final comprehensive and accurate evaluation of the performance of the pressure sensor.

[0087] In another alternative embodiment 2, the determination of the performance index evaluation data paired with the multi-dimensional sensor performance detection data based on the third performance detection response information and at least one fourth performance detection response information comprises: combining the third performance detection response vector set used to represent the third performance detection response information, and at least one fourth performance detection response vector set used to represent at least one fourth performance detection response information, to obtain a linkage performance detection response vector set; using the linkage performance detection response vector set and the attribute mapping relationship network paired with the detection attribute recognition branch to determine an index evaluation viewpoint label set used to represent the performance index evaluation data.

[0088] In this alternative embodiment 2, for determining the performance index evaluation data paired with the multi-dimensional sensing performance detection data based on the third performance detection response information and the at least one fourth performance detection response information, the process first involves combining the third performance detection response vector set for representing the third performance detection response information, and the at least one fourth performance detection response vector set for representing the at least one fourth performance detection response information, to obtain a linkage performance detection response vector set. Taking the aforementioned feature vector values as an example, the values in the third performance detection response vector set, such as [0.6, 0.3, 0.1, 0.7], correspond to the confidence values of the overall performance at different levels (e.g. good, medium, poor) and the comprehensive performance tendency value respectively. As for the fourth performance detection response vector set, one of the fourth performance detection response vectors is [0.8, 0.6], and the element orders in the embodiment correspond to the step response comprehensive performance value and the vibration influence comprehensive performance value respectively. Combining such third performance detection response vector set and at least one fourth performance detection response vector set forms the linkage performance detection response vector set. For example, if there is only one fourth performance detection response vector set, the linkage performance detection response vector set can be [0.6, 0.3, 0.1, 0.7, 0.8, 0.6], which integrates the relevant values of the third performance detection response information and the fourth performance detection response information, and contains the results of performance evaluation of the pressure sensor from different angles.

[0089] Then, the linkage performance detection response vector set, and the attribute mapping relationship network paired with the detection attribute identification branch, are used to determine the index evaluation viewpoint label set for representing the performance index evaluation data. The attribute mapping relationship network stores the mapping relationship between each value in the linkage performance detection response vector set and the performance index evaluation. For example, the value 0.6 in the linkage performance detection response vector set (e.g. the confidence value of the overall performance in the third performance detection response vector set) can correspond to an evaluation viewpoint label about the comprehensive performance of the sensor in terms of accuracy, stability, etc. in the attribute mapping relationship network, such as “overall performance close to good level and has advantages in some aspects”. For the value 0.8 in the vector set (e.g. the step response comprehensive performance value in the fourth performance detection response vector set), the evaluation viewpoint label in the attribute mapping relationship network can be “step response performance is good but still has room for improvement”. Through the mapping analysis of each value in the linkage performance detection response vector set, the index evaluation viewpoint label set is finally determined. This index evaluation viewpoint label set is a comprehensive and qualitative evaluation result of the performance of the pressure sensor, which describes the performance of the pressure sensor as a whole and takes into account various factors in the third performance detection response information and the fourth performance detection response information.

[0090] Thus, by combining the third performance detection response vector set and the fourth performance detection response vector set into the linkage performance detection response vector set, the integration of performance detection information of different sources is realized. This integration method makes all the numerical information related to performance evaluation concentrated together, avoiding the dispersion of information. Then the attribute mapping relationship network is used to determine the index evaluation viewpoint label set, and the integrated numerical information is converted into qualitative evaluation viewpoint labels, providing an intuitive and comprehensive performance evaluation result for the user. For example, in actual application, the technical personnel can directly understand the performance status of the pressure sensor in various aspects, such as the overall performance level, the specific performance (such as the step response), etc., from the index evaluation viewpoint label set, which helps to quickly judge whether the pressure sensor meets the requirements, whether it needs to be further improved, etc., improving the comprehensiveness and practicality of the performance evaluation of the pressure sensor.

[0091] In still another alternative embodiment 3, the determining of the detection element mapping relationship network, the evaluation mapping relationship network and the performance correction mapping relationship network included in the at least one target mapping relationship network comprises at least one of the following: obtaining a first local detection element relationship network and a second local detection element relationship network, determining the detection element mapping relationship network based on a first relationship feature operation result between the first local detection element relationship network and the second local detection element relationship network; obtaining a first local evaluation relationship network and a second local evaluation relationship network, determining the evaluation mapping relationship network based on a second relationship feature operation result between the first local evaluation relationship network and the second local evaluation relationship network; obtaining a first local performance correction relationship network and a second local performance correction relationship network, determining the performance correction mapping relationship network based on a third relationship feature operation result between the first local performance correction relationship network and the second local performance correction relationship network.

[0092] Further, after determining the performance index evaluation data paired with the multi-dimensional sensing performance detection data based on the third performance detection response information and at least one fourth performance detection response information, the method further comprises: based on the detection attribute recognition branch being a bidirectional long short-term memory branch in the target multi-element decision tree, processing the performance index evaluation data based on a linear transformation branch connected to the bidirectional long short-term memory branch to obtain performance detection linear quantization data; determining a target fusion feature set based on a first local fusion feature set and a second local fusion feature set paired with a detection feature fusion branch; processing the performance detection linear quantization data based on the target fusion feature set to obtain target performance detection data; based on the bidirectional long short-term memory branch being the last bidirectional long short-term memory branch in the target multi-element decision tree, determining a detection attribute recognition result based on the target performance detection data; based on the bidirectional long short-term memory branch not being the last bidirectional long short-term memory branch in the target multi-element decision tree, inputting the target performance detection data into a next bidirectional long short-term memory branch.

[0093] In another alternative embodiment 3, the technical solution revolves around the determination of each relationship network (detection element mapping relationship network, evaluation mapping relationship network, and performance correction mapping relationship network) in the target mapping relationship network and the subsequent processing based on the performance index evaluation data.

[0094] For determining the detection element mapping relationship network, it involves obtaining a first local detection element relationship network and a second local detection element relationship network, and determining based on the first relationship feature operation result therebetween. For example, the first local detection element relationship network can contain the relationship between part of the elements in the static performance detection data of the pressure sensor and other related factors, such as the relationship between the measurement error in the accuracy detection and certain circuit parameters inside the sensor; the second local detection element relationship network can contain the relationship between part of the elements in the dynamic performance detection data and other related factors, such as the relationship between the amplitude-frequency characteristic in the frequency response and the structural characteristics of the sensor. The first relationship feature operation result is an operation integration between the two local relationship networks, and the operation can be based on a certain logical relationship or data association rule. Through this operation result, the detection element mapping relationship network can be determined, which comprehensively maps and associates different performance detection elements of the pressure sensor with other related knowledge.

[0095] In determining the evaluation mapping relationship network, the first partial evaluation relationship network and the second partial evaluation relationship network are obtained, and then the second relationship feature operation result is determined based on the two. The first partial evaluation relationship network may contain the relationship between the evaluation standard of some performance indicators (such as the step response indicator in dynamic performance) under certain conditions and other factors (such as the average level of the same type of sensor); the second partial evaluation relationship network may contain the relationship between the evaluation standard of some indicators (such as the zero point temperature drift under the influence of temperature) in environmental adaptability detection data and other related factors (such as the requirement of application scene on temperature stability). The second relationship feature operation result integrates the two partial evaluation relationship networks, thereby determining the evaluation mapping relationship network, which comprehensively defines the mapping relationship between different performance detection data and evaluation standards.

[0096] In determining the performance correction mapping relationship network, the first partial performance correction relationship network and the second partial performance correction relationship network are obtained, and the third relationship feature operation result is determined based on the two. The first partial performance correction relationship network may contain the relationship between the preliminary correction strategy of some performance indicators (such as the aging rate in durability stability) and other characteristics of the sensor (such as material characteristics); the second partial performance correction relationship network may contain the relationship between the correction strategy of some parameters (such as the phase frequency characteristic in frequency response) in dynamic performance and other related factors (such as the signal processing method). The third relationship feature operation result integrates the two partial performance correction relationship networks, thereby determining the performance correction mapping relationship network, which clearly defines the mapping relationship between different performance detection data and performance correction strategies.

[0097] Further, after determining the performance indicator evaluation data paired with the multi-dimensional sensor performance detection data based on the third performance detection response information and at least one fourth performance detection response information, there are a series of processing steps. When the detection attribute recognition branch is a bidirectional long short-term memory branch in the target multi-element decision tree, the performance indicator evaluation data is processed based on the linear transformation branch connected with the bidirectional long short-term memory branch to obtain performance detection linear quantization data. For example, the performance indicator evaluation data may exist in the form of a set of qualitative evaluation labels, such as {“overall performance is good, step response needs to be optimized”}, and the linear transformation branch may convert these qualitative data into quantifiable data, such as corresponding the overall performance is good to the numerical value 0.8 (representing the quantification of the degree of performance excellence), corresponding the step response needs to be optimized to the numerical value 0.3 (representing the quantification of the degree of improvement needed), and so on, thereby obtaining the performance detection linear quantization data.

[0098] Then, the target fusion feature set is determined based on the first local fusion feature set and the second local fusion feature set paired with the detection feature fusion branch. The first local fusion feature set can include features extracted from part of the performance detection data, such as a feature vector [0.8, 10] extracted from the frequency response part of the dynamic performance detection data (where 0.8 is the amplitude-frequency characteristic at a certain frequency, and 10 is the phase-frequency characteristic); the second local fusion feature set can include features extracted from the environmental adaptability detection data, such as [±0.3, 0.03] (where ±0.3 is the zero-point temperature drift, and 0.03 is the humidity influence coefficient). The target fusion feature set is determined by specific fusion rules (such as weighted average, feature splicing, etc.).

[0099] Then, the target performance detection data is obtained by processing the performance detection linear quantization data based on the target fusion feature set. For example, the target fusion feature set can adjust or supplement each value in the performance detection linear quantization data, and correct the overall performance quantization value 0.8 in the performance detection linear quantization data according to the related features in the target fusion feature set, to obtain more accurate target performance detection data.

[0100] When the bidirectional long short-term memory branch is the last bidirectional long short-term memory branch in the target multi-element decision tree, the detection attribute recognition result is determined based on the target performance detection data. This detection attribute recognition result is the final recognition conclusion for the performance of the pressure sensor, for example, it can be "the pressure sensor achieves excellent level in all detection performances and can be applied to high-precision measurement scenarios". When the bidirectional long short-term memory branch is not the last bidirectional long short-term memory branch in the target multi-element decision tree, the target performance detection data is input into the next bidirectional long short-term memory branch for further analysis and processing.

[0101] As can be seen, by using the combination operation of the local relationship network to determine each relationship network in the target mapping relationship network, the relationship network can be more flexible and more detailed, and the accuracy and comprehensiveness of the relationship network are improved, thereby improving the processing effect of the first performance detection response information. In subsequent processing based on performance index evaluation data, through the synergistic effect of the bidirectional long short-term memory branch, the linear transformation branch, and the detection feature fusion branch, the performance index evaluation data can be processed in multiple dimensions. From qualitative to quantitative conversion, feature fusion, and transfer processing between multiple bidirectional long short-term memory branches, the evaluation of the performance of the pressure sensor is more comprehensive, in-depth, and accurate, which helps to more accurately identify the performance state of the pressure sensor, and provides a more reliable basis for quality control, performance optimization, and reasonable application of the pressure sensor.

[0102] In a further technical solution, the determining of the detection element mapping relationship network based on the first relationship feature operation result between the first local detection element relationship network and the second local detection element relationship network comprises: obtaining a cross-detection element relationship network; determining the detection element mapping relationship network based on a fourth relationship feature operation result between the local correlation mapping relationship network corresponding to the first relationship feature operation result and the cross-detection element relationship network; and wherein the cross-detection element relationship network is used to determine the detection element mapping relationship network corresponding to each of the plurality of detection attribute recognition units.

[0103] The determining of the evaluation mapping relationship network based on the second relationship feature operation result between the first local evaluation relationship network and the second local evaluation relationship network comprises: obtaining a cross-evaluation relationship network; determining the evaluation mapping relationship network based on a fifth relationship feature operation result between the local correlation mapping relationship network corresponding to the second relationship feature operation result and the cross-evaluation relationship network; and wherein the cross-evaluation relationship network is used to determine the evaluation mapping relationship network corresponding to each of the plurality of detection attribute recognition units.

[0104] The determining of the performance correction mapping relationship network based on the third relationship feature operation result between the first local performance correction relationship network and the second local performance correction relationship network comprises: obtaining a cross-performance correction relationship network; determining the performance correction mapping relationship network based on a sixth relationship feature operation result between the local correlation mapping relationship network corresponding to the third relationship feature operation result and the cross-performance correction relationship network; and wherein the cross-evaluation relationship network is used to determine the performance correction mapping relationship network corresponding to each of the plurality of detection attribute recognition units.

[0105] In this further technical solution, for the process of determining the detection element mapping relationship network, first, the detection element mapping relationship network is determined based on the first relationship feature operation result between the first local detection element relationship network and the second local detection element relationship network. In this process, a cross-detection element relationship network is obtained. For example, the first local detection element relationship network can contain the relationship between some elements in the accuracy detection of the pressure sensor and other characteristics of the sensor, such as the relationship between the measurement error and the sensor chip precision; the second local detection element relationship network can involve the relationship between some elements and other related factors in dynamic performance detection, such as the relationship between the frequency response and the internal structure of the sensor. The first relationship feature operation result is the result of a specific operation between the two local relationship networks, and this operation is based on the internal logical association between them.

[0106] Then, the fourth relationship feature operation result between the local correlation mapping relationship network corresponding to the first relationship feature operation result and the cross detection element relationship network is determined, and finally this result is determined as the detection element mapping relationship network. The cross detection element relationship network has a special role, which is used to determine the detection element mapping relationship network corresponding to each detection attribute identification unit. For example, there are three detection attribute identification units, and the cross detection element relationship network will construct the detection element mapping relationship network from the above operation result in a way suitable for each detection attribute identification unit according to the characteristics and needs of different detection attribute identification units. For example, for the first detection attribute identification unit, if it mainly focuses on the repeatability aspect in static performance detection, the cross detection element relationship network will filter and adjust the relationship content related to repeatability from the overall operation result to construct the detection element mapping relationship network part suitable for the detection attribute identification unit.

[0107] For determining the evaluation mapping relationship network, the second relationship feature operation result between the first local evaluation relationship network and the second local evaluation relationship network is operated. Similarly, the cross evaluation relationship network is obtained first. The first local evaluation relationship network may contain the relationship between the evaluation standard of the step response index in dynamic performance detection in different application scenarios and other related factors (such as industry standards, cost limitations, etc.); the second local evaluation relationship network may contain the relationship between the evaluation standard of the temperature influence index in environmental adaptability detection under specific conditions and other related factors (such as the temperature range of the device operating environment, etc.). The second relationship feature operation result is the integrated operation result of the two local evaluation relationship networks.

[0108] Then, the fifth relationship feature operation result between the local correlation mapping relationship network corresponding to the second relationship feature operation result and the cross evaluation relationship network is determined, and is determined as the evaluation mapping relationship network. The cross evaluation relationship network is used to determine the evaluation mapping relationship network corresponding to each detection attribute identification unit. For example, there are two detection attribute identification units, and for the first detection attribute identification unit, if it focuses on the frequency response evaluation in dynamic performance detection, the cross evaluation relationship network will construct the evaluation mapping relationship network part related to the frequency response evaluation according to the overall operation result to meet the needs of the detection attribute identification unit for the evaluation mapping relationship network.

[0109] In determining the performance correction mapping relationship network, a third relationship characteristic operation result is operated based on the first local performance correction relationship network and the second local performance correction relationship network. A cross performance correction relationship network is obtained first. The first local performance correction relationship network can contain correction strategies for the aging rate index in the durability stability detection and the relationship between the sensor material characteristics and the manufacturing process; the second local performance correction relationship network can contain the phase frequency characteristic correction strategies in the dynamic performance detection and the relationship between the signal processing algorithm and the circuit design. The third relationship characteristic operation result is the operation result of the two local performance correction relationship networks.

[0110] Then, the sixth relationship characteristic operation result is determined between the local association mapping relationship network corresponding to the third relationship characteristic operation result and the cross performance correction relationship network, and is determined as the performance correction mapping relationship network. The cross performance correction relationship network in the embodiment of the application is used to determine the performance correction mapping relationship network corresponding to each of the plurality of detection attribute recognition units. For example, if there are four detection attribute recognition units, for the third detection attribute recognition unit, if it mainly involves humidity influence correction in environmental adaptability detection, the cross performance correction relationship network will filter and construct the performance correction mapping relationship network part related to the humidity influence correction from the overall operation result to adapt to the requirements of this detection attribute recognition unit.

[0111] Based on the above technical solution, by introducing the cross detection element relationship network, the cross evaluation relationship network and the cross performance correction relationship network, the different requirements of the plurality of detection attribute recognition units can be more accurately met when constructing the detection element mapping relationship network, the evaluation mapping relationship network and the performance correction mapping relationship network. The collaborative operation mode of the multiple relationship networks makes the construction of each mapping relationship network more detailed, comprehensive and targeted. For example, when constructing the detection element mapping relationship network for different detection attribute recognition units, the relevant relationship content can be accurately extracted and adjusted from the overall operation result according to the focus of each unit, thereby improving the accuracy and effectiveness of the pressure sensor performance detection data processing, and laying a solid foundation for more accurately evaluating the performance of the pressure sensor.

[0112] In an optional technical idea, before the target identification unit obtains the first performance detection response information in the multi-dimensional sensing performance detection data, it also includes: obtaining a first decision tree model, wherein the first decision tree model includes a plurality of cascaded sensing signal embedding mining components and a plurality of cascaded sensing signal embedding analysis components, the sensing signal embedding mining component and the sensing signal embedding analysis component include cascaded bidirectional long short-term memory branches, linear transformation branches and detection feature fusion branches; based on at least one signal attribute mapping relationship network in the first decision tree model paired with the bidirectional long short-term memory branch, at least one association attribute mapping relationship network paired with a plurality of detection attribute identification units in the bidirectional long short-term memory branch is determined; adjust the association attribute mapping relationship network to obtain a plurality of attribute mapping feature subsets paired with the association attribute mapping relationship network, wherein the relationship network size of each of the plurality of attribute mapping feature subsets is smaller than the relationship network size of the association attribute mapping relationship network; update at least one signal attribute mapping relationship network paired with the bidirectional long short-term memory branch based on a plurality of attribute mapping feature subsets to obtain a second decision tree model; debug the second decision tree model to obtain the target multi-element decision tree, wherein the target multi-element decision tree is used to process the pressure sensor detection data stream.

[0113] In some preferred embodiments, before the second decision tree model is debugged to obtain the target multi-element decision tree, it also includes: adjusting the sample fusion feature set in the second decision tree model paired with the detection feature fusion branch to obtain a plurality of local fusion feature sets matched with the sample fusion feature set, wherein the relationship network size of each of the plurality of local fusion feature sets is smaller than the relationship network size of the sample fusion feature set; update the sample fusion feature set paired with the detection feature fusion branch using a plurality of local fusion feature sets to obtain the second decision tree model to be debugged.

[0114] Further, the second decision tree model is debugged to obtain the target multi-element decision tree, wherein the target multi-element decision tree is used for processing pressure sensor detection data streams, and the method comprises: obtaining a historical detection data set, wherein the historical detection data set comprises a plurality of pressure sensor detection data streams and training annotations respectively paired with the plurality of pressure sensor detection data streams; debugging the second decision tree model using the historical detection data set; on the basis that the second decision tree model does not meet the debugging requirements, optimizing model weights associated with the second decision tree model, wherein the model weights are model variables in a model variable set indicated by a plurality of attribute mapping feature subsets and a plurality of local fusion feature sets; and on the basis that the second decision tree model meets the debugging requirements, determining the second decision tree model as the target multi-element decision tree.

[0115] Under this optional technical idea, before obtaining the first performance detection response information in the multi-dimensional sensor performance detection data through the target recognition unit, the operation of constructing and training the target multi-element decision tree is further included.

[0116] First, a first decision tree model is obtained, which includes a plurality of cascaded sensor signal embedding mining components and a plurality of cascaded sensor signal embedding analysis components, and there are cascaded bidirectional long short-term memory branches, linear transformation branches and detection feature fusion branches in these components. For example, in the sensor signal embedding mining component, the bidirectional long short-term memory branch may be responsible for processing the time sequence related information in the pressure sensor detection data, the linear transformation branch performs linear conversion operation on the data, and the detection feature fusion branch fuses features from different sources.

[0117] Based on at least one signal attribute mapping relationship network paired with the bidirectional long short-term memory branch in the first decision tree model, at least one associated attribute mapping relationship network paired with a plurality of detection attribute recognition units in the bidirectional long short-term memory branch is determined. For example, in the first decision tree model, the signal attribute mapping relationship network stores the mapping relationship between the detection data processed by the bidirectional long short-term memory branch and other related attributes, and through a specific algorithm or rule, the associated attribute mapping relationship network related to each detection attribute recognition unit is determined. For example, for a detection attribute recognition unit mainly concerned about the frequency response data in the dynamic performance of the pressure sensor, the associated attribute mapping relationship network will filter and adjust the mapping relationship part related to the frequency response from the signal attribute mapping relationship network.

[0118] Then, the association attribute mapping relationship network is adjusted to obtain a plurality of attribute mapping feature subsets paired with the association attribute mapping relationship network. The relationship network size of each of the attribute mapping feature subsets is smaller than the relationship network size of the association attribute mapping relationship network. For example, the association attribute mapping relationship network can contain complete mapping relationship of frequency response from 10 Hz to 1000 Hz, and the attribute mapping feature subset obtained after adjustment can only focus on the mapping relationship of frequency response from 100 Hz to 300 Hz, so that the range of the relationship network is reduced and more targeted.

[0119] The second decision tree model is obtained by updating at least one signal attribute mapping relationship network paired with the bidirectional long short-term memory branch based on the plurality of attribute mapping feature subsets. The updating operation makes the signal attribute mapping relationship network on which the bidirectional long short-term memory branch relies more accurate and efficient when processing data.

[0120] In some preferred embodiments, before the second decision tree model is debugged, there is an additional operation. The sample fusion feature set paired with the detection feature fusion branch in the second decision tree model is adjusted to obtain a plurality of local fusion feature sets matched with the sample fusion feature set, and the relationship network size of each of the local fusion feature sets is smaller than the relationship network size of the sample fusion feature set. For example, the sample fusion feature set contains fusion features of static performance, dynamic performance, environmental adaptability and durability stability of the pressure sensor, and the local fusion feature set obtained after adjustment can only focus on the fusion features of dynamic performance and environmental adaptability, so that the range of the feature set is reduced.

[0121] The sample fusion feature set paired with the detection feature fusion branch is updated by using the plurality of local fusion feature sets to obtain the second decision tree model to be debugged. The updating operation makes the sample fusion feature set on which the detection feature fusion branch relies more in line with the requirements when processing data.

[0122] Further, the second decision tree model is debugged to obtain a target multi-element decision tree. First, a historical detection data set is obtained, which contains a plurality of pressure sensor detection data streams and training annotations respectively paired with the data streams. For example, the pressure sensor detection data stream can contain detection data such as accuracy, repeatability, frequency response, temperature influence, and the training annotation is a label of the performance of the pressure sensor corresponding to the detection data, such as good, medium or poor.

[0123] The second decision tree model is debugged by using a historical detection data set. During the debugging, if the second decision tree model does not meet the debugging requirements, the model weights associated with the second decision tree model are optimized. The model weights in the embodiment of the present application are model variables in a model variable set indicated by a plurality of attribute mapping feature subsets and a plurality of local fusion feature sets. For example, if the model has a large deviation in the accuracy of predicting the pressure sensor, the model variable weights involved in the attribute mapping feature subsets and the local fusion feature sets related to the accuracy are adjusted.

[0124] If the second decision tree model meets the debugging requirements, the second decision tree model is determined as a target multi-element decision tree, which is used to process the pressure sensor detection data stream.

[0125] In this way, through the step-by-step adjustment and optimization of the first decision tree model, including the operations of constructing the associated attribute mapping relationship network, the attribute mapping feature subset, and the local fusion feature set, the decision tree model is more accurate and efficient. When constructing the associated attribute mapping relationship network and the attribute mapping feature subset, different detection attribute recognition units can be optimized, improving the accuracy of the bidirectional long short-term memory branch in processing data. When constructing the local fusion feature set, the data processing capability of the detection feature fusion branch is optimized. Through the debugging and weight optimization by using the historical detection data set, the accuracy and reliability of the model are further improved, and the target multi-element decision tree obtained finally can more effectively process the pressure sensor detection data stream, thereby providing more accurate and reliable evaluation basis for the performance detection of the pressure sensor.

[0126] In an independent embodiment, before determining at least one associated attribute mapping relationship network respectively paired with a plurality of detection attribute recognition units in the bidirectional long short-term memory branch based on at least one signal attribute mapping relationship network paired with the bidirectional long short-term memory branch in the first decision tree model, the following steps are cyclically implemented until a target down-sampling coding feature is determined from a plurality of down-sampling coding features, wherein the down-sampling coding feature includes a plurality of adjustable model variables, each adjustable model variable respectively indicates a round of model variable down-sampling processing implemented on the first decision tree model, the model variable down-sampling processing includes at least one of feature sampling and feature updating, and the target down-sampling coding feature is used to improve the first decision tree model into the second decision tree model.

[0127] (1) Obtain a plurality of down-sampling coding features obtained in the previous round of processing;

[0128] (2) Process the first decision tree model respectively according to a plurality of down-sampling coding features to obtain associated decision tree models respectively corresponding to a plurality of down-sampling coding features;

[0129] (3) Based on the target historical detection dataset, obtain the decision analysis error of each of the multiple associated decision tree models;

[0130] (4) Based on the decision analysis error of each of the multiple associated decision tree models, determine the noise judgment weights that are paired with each of the multiple downsampled coding features;

[0131] (5) Based on the fact that the noise determination weight does not include the first downsampled coding feature among the multiple downsampled coding features, at least one second downsampled coding feature is determined from the multiple downsampled coding features based on the noise determination weight; the at least one second downsampled coding feature is sequentially optimized to obtain at least one optimized second downsampled coding feature; the at least one optimized second downsampled coding feature is added to the multiple downsampled coding features;

[0132] (6) Based on the first downsampling coding feature among the multiple downsampling coding features, which includes the first downsampling coding feature whose noise determination weight meets the target threshold, the first downsampling coding feature is determined as the target downsampling coding feature.

[0133] In this standalone embodiment, before determining at least one associated attribute mapping network that is paired with multiple detection attribute recognition units in the bidirectional long short-term memory branch based on at least one signal attribute mapping network paired with the bidirectional long short-term memory branch in the first decision tree model, there is a loop operation to determine a target downsampling coding feature, which is used to improve the first decision tree model into a second decision tree model.

[0134] First, the downsampled encoded features contain multiple adjustable model variables, each indicating a round of model variable downsampling processing performed on the first decision tree model, which includes at least one of feature sampling and feature updating.

[0135] At the start of the loop, multiple downsampled encoded features obtained from the previous round of processing are acquired. For example, in the first round, there might be an initial set of downsampled encoded features, which are obtained by initially setting the relevant variables in the first decision tree model based on certain rules.

[0136] Then, the first decision tree model is processed according to the plurality of down-sampling encoding features respectively, to obtain a plurality of associated decision tree models respectively corresponding to the plurality of down-sampling encoding features. Taking the first decision tree model containing a plurality of cascaded sensor signal embedding mining components and a plurality of cascaded sensor signal embedding analysis components as an example, after adjusting the relevant variables in these components by each down-sampling encoding feature, different associated decision tree models are obtained. These associated decision tree models may differ in structure and data processing method due to the differences in down-sampling encoding features.

[0137] Then, based on a target historical detection data set, the decision analysis errors of the plurality of associated decision tree models are obtained. The target historical detection data set contains a plurality of pressure sensor detection data streams and training annotations respectively paired with the data streams. For example, when a certain associated decision tree model processes the accuracy detection data in the historical detection data set, it may deviate from the accurate results in the training annotations, and this deviation can be a decision analysis error. For a historical detection data set containing frequency response, temperature influence and other aspects of data, the associated decision tree model will produce corresponding decision analysis errors when processing these data.

[0138] According to the decision analysis errors of the plurality of associated decision tree models, noise judgment weights respectively paired with the plurality of down-sampling encoding features are determined. The larger the decision analysis error, the higher the corresponding noise judgment weight, indicating that the down-sampling encoding feature may introduce more "noise" or uncertainty in model processing. For example, when an associated decision tree model processes dynamic performance detection data of a pressure sensor, the decision analysis error is large, and the noise judgment weight of the corresponding down-sampling encoding feature may be high.

[0139] Based on the plurality of down-sampling encoding features not including a first down-sampling encoding feature whose noise judgment weight meets a target threshold, at least one second down-sampling encoding feature is determined from the plurality of down-sampling encoding features based on the noise judgment weights. The target threshold is a predetermined judgment standard. If no down-sampling encoding feature has a noise judgment weight that meets this standard, some down-sampling encoding features are selected as second down-sampling encoding features according to the noise judgment weights. Then, the at least one second down-sampling encoding feature is sequentially subjected to feature optimization to obtain at least one optimized second down-sampling encoding feature. For example, some model variables related to dynamic performance detection data in the down-sampling encoding features are adjusted and optimized, such as adjusting the variable weight related to frequency response. The at least one optimized second down-sampling encoding feature is added to the plurality of down-sampling encoding features for the next round of circulation.

[0140] And in multiple down-sampling encoded features, including the first down-sampling encoded feature whose noise judgment weight value meets the target threshold, the first down-sampling encoded feature is determined as the target down-sampling encoded feature. Once the noise judgment weight value of a certain down-sampling encoded feature reaches the target threshold, it is considered that this down-sampling encoded feature meets the requirements and can be used as the target down-sampling encoded feature for improving the first decision tree model into the second decision tree model.

[0141] It can be seen that through the cyclic processing and screening of the down-sampling encoded features, the target down-sampling encoded feature most suitable for improving the first decision tree model into the second decision tree model can be effectively determined from numerous possible down-sampling encoded features. In this process, the noise judgment weight value is determined based on the decision analysis error, which can quantify the influence of each down-sampling encoded feature on the model accuracy. Through the feature optimization and screening process, the quality of the down-sampling encoded features is continuously improved, and the target down-sampling encoded feature obtained finally can make the improved second decision tree model more accurate and effective when processing pressure sensor detection data, laying a good foundation for subsequent operations such as constructing the associated attribute mapping relationship network, and improving the accuracy and reliability of the entire pressure sensor performance detection system.

[0142] In summary, the embodiment of the present application first acquires the first performance detection response information from the multi-dimensional sensing performance detection data by using the target recognition unit. This structure based on the target multi-element decision tree detection attribute recognition branch can classify and recognize various performance detection data, ensuring the accuracy and pertinence of data acquisition. Then, the confidence degree relationship list is determined by the target mapping relationship network paired with the target recognition unit to obtain the third performance detection response information, which can fully consider the complex relationship between various detection information and improve the accuracy of evaluation. The associated recognition unit processes the second performance detection response information to obtain the fourth performance detection response information, and the multi-unit cooperation realizes comprehensive analysis of different performance data. Finally, the performance index evaluation data is determined based on the third and fourth performance detection response information, which can provide comprehensive, accurate and scientific evaluation results for the performance of the pressure sensor, and help to accurately control the performance of the pressure sensor in multiple links such as production, research and development, and use.

[0143] Further, Figure 2 The structural block diagram of the pressure sensor performance detection system 300 is shown, which includes a memory 310 for storing program instructions and data, and a processor 320 coupled with the memory 310 for executing the instructions in the memory 310 to implement the above-mentioned method.

[0144] Further, a computer storage medium is also provided, which contains instructions that, when executed on a processor, implement the above-mentioned method.

[0145] In several embodiments provided by the embodiments of the present application, it should be understood that the disclosed apparatus and method can also be implemented by other manners. The apparatus and method embodiments described above are only illustrative, for example, the flowcharts and block diagrams in the drawings show the possible implementation architecture, function and operation of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment or a part of code, which includes one or more executable instructions for implementing the specified logic function. It should also be noted that in some alternative implementation manners, the functions noted in the blocks can also occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can also be executed in reverse order depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for executing the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0146] In addition, each functional module in the embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0147] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a network device, or the like) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes. It should be noted that in this document, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the processes, methods, articles or devices that include a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or device that includes the element.

[0148] The above descriptions are only the preferred embodiments of the present application, not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A method of detecting performance of a pressure sensor, characterized by, The method is applied to a pressure sensor performance detection system, and the method comprises: A target recognition unit is used to obtain first performance detection response information in multi-dimensional sensor performance detection data, wherein the target recognition unit is one of detection attribute recognition branches of a target multi-element decision tree, the detection attribute recognition branches comprise the target recognition unit and at least one associated recognition unit, and the at least one associated recognition unit is used to process at least one second performance detection response information respectively determined based on the multi-dimensional sensor performance detection data; At least one target mapping relationship network matched with the target recognition unit is used to determine a confidence relationship list matched with the first performance detection response information, and third performance detection response information is determined based on the first performance detection response information and the confidence relationship list, wherein a relationship network size of the target mapping relationship network matches a detection response vector size of the first performance detection response information; At least one fourth performance detection response information processed by the at least one associated recognition unit from the respective second performance detection response information is obtained; Performance index evaluation data matched with the multi-dimensional sensor performance detection data is determined based on the third performance detection response information and the at least one fourth performance detection response information.

2. The method of claim 1, wherein, The at least one target mapping relationship network matched with the target recognition unit is used to determine the confidence relationship list matched with the first performance detection response information, and the third performance detection response information is determined based on the first performance detection response information and the confidence relationship list, and the method comprises: A detection element mapping relationship network included in the at least one target mapping relationship network is determined, and an evaluation mapping relationship network and a performance correction mapping relationship network are evaluated, wherein a relationship network size of the detection element mapping relationship network, the evaluation mapping relationship network and the performance correction mapping relationship network matches a detection response vector size of the first performance detection response information; A response element mapping knowledge set is determined based on the first performance detection response information and the detection element mapping relationship network, and a response evaluation mapping knowledge set is determined based on the first performance detection response information and the evaluation mapping relationship network; The confidence relationship list matched with the first performance detection response information is determined according to the response element mapping knowledge set and the response evaluation mapping knowledge set; A response correction mapping knowledge set is determined based on the first performance detection response information and the performance correction mapping relationship network, and the third performance detection response information is determined according to the confidence relationship list and the response correction mapping knowledge set.

3. The method of claim 2, wherein, Before the at least one fourth performance detection response information processed by the at least one associated recognition unit from the respective second performance detection response information is obtained, the method further comprises: acquire the u-th second performance detection response information in the multi-dimensional sensing performance detection data through the u-th association identification unit, where u is an integer not less than 1 and not greater than X, X is the number of the association identification units included in the detection attribute identification branch; determine a confidence degree relationship list paired with the u-th second performance detection response information based on at least one association mapping relationship network paired with the u-th association identification unit, and determine the u-th fourth performance detection response information based on the u-th second performance detection response information and the confidence degree relationship list.

4. The method of claim 2, wherein, The performance index evaluation data paired with the multi-dimensional sensing performance detection data is determined based on the third performance detection response information and at least one fourth performance detection response information, including: combining a third performance detection response vector set used for representing the third performance detection response information and at least one fourth performance detection response vector set used for representing at least one fourth performance detection response information respectively to obtain a linkage performance detection response vector set; determining an index evaluation viewpoint label set used for representing the performance index evaluation data by using the linkage performance detection response vector set and an attribute mapping relationship network paired with the detection attribute identification branch.

5. The method of claim 2, wherein, The detection element mapping relationship network, the evaluation mapping relationship network and the performance correction mapping relationship network included in the at least one target mapping relationship network are determined, including at least one of the following: acquire a first local detection element relationship network and a second local detection element relationship network, and determine the detection element mapping relationship network based on a first relationship feature operation result between the first local detection element relationship network and the second local detection element relationship network; acquire a first local evaluation relationship network and a second local evaluation relationship network, and determine the evaluation mapping relationship network based on a second relationship feature operation result between the first local evaluation relationship network and the second local evaluation relationship network; acquire a first local performance correction relationship network and a second local performance correction relationship network, and determine the performance correction mapping relationship network based on a third relationship feature operation result between the first local performance correction relationship network and the second local performance correction relationship network.

6. The method of claim 5, wherein, After determining the performance index evaluation data paired with the multi-dimensional sensing performance detection data based on the third performance detection response information and at least one fourth performance detection response information, including: based on a linear transformation branch connected with the bidirectional long short-term memory branch in the target multi-element decision tree, processing the performance index evaluation data to obtain performance detection linear quantization data; determining a target fusion feature set based on a first local fusion feature set and a second local fusion feature set paired with a detection feature fusion branch; processing the performance detection linear quantization data based on the target fusion feature set to obtain target performance detection data; On the basis that the bidirectional long short-term memory branch is the last bidirectional long short-term memory branch in the target multi-element decision tree, a detection attribute recognition result is determined based on the target performance detection data; On the basis that the bidirectional long short-term memory branch is not the last bidirectional long short-term memory branch in the target multi-element decision tree, the target performance detection data is input into the next bidirectional long short-term memory branch.

7. The method of claim 5, wherein, The determination of the detection element mapping relationship network based on the first relationship feature operation result between the first local detection element relationship network and the second local detection element relationship network comprises: obtaining a cross-detection element relationship network; and determining a fourth relationship feature operation result between a local correlation mapping relationship network corresponding to the first relationship feature operation result and the cross-detection element relationship network as the detection element mapping relationship network; wherein the cross-detection element relationship network is used to determine the detection element mapping relationship network corresponding to each of the plurality of detection attribute recognition units. The determination of the evaluation mapping relationship network based on the second relationship feature operation result between the first local evaluation relationship network and the second local evaluation relationship network comprises: obtaining a cross-evaluation relationship network; and determining a fifth relationship feature operation result between a local correlation mapping relationship network corresponding to the second relationship feature operation result and the cross-evaluation relationship network as the evaluation mapping relationship network; wherein the cross-evaluation relationship network is used to determine the evaluation mapping relationship network corresponding to each of the plurality of detection attribute recognition units. The determination of the performance correction mapping relationship network based on the third relationship feature operation result between the first local performance correction relationship network and the second local performance correction relationship network comprises: obtaining a cross-performance correction relationship network; and determining a sixth relationship feature operation result between a local correlation mapping relationship network corresponding to the third relationship feature operation result and the cross-performance correction relationship network as the performance correction mapping relationship network; wherein the cross-evaluation relationship network is used to determine the performance correction mapping relationship network corresponding to each of the plurality of detection attribute recognition units.

8. The method of claim 1, wherein, Before the target recognition unit obtains the first performance detection response information in the multi-dimensional sensing performance detection data, the method further comprises: obtaining a first decision tree model, wherein the first decision tree model comprises a plurality of cascaded sensing signal embedding mining components and a plurality of cascaded sensing signal embedding analysis components, and the sensing signal embedding mining components and the sensing signal embedding analysis components comprise cascaded bidirectional long short-term memory branches, linear transformation branches and detection feature fusion branches; determining at least one correlation attribute mapping relationship network corresponding to each of the plurality of detection attribute recognition units in the bidirectional long short-term memory branch based on at least one signal attribute mapping relationship network in the first decision tree model paired with the bidirectional long short-term memory branch; determining at least one correlation attribute mapping relationship network corresponding to each of the plurality of detection attribute recognition units in the bidirectional long short-term memory branch based on at least one signal attribute mapping relationship network in the first decision tree model paired with the bidirectional long short-term memory branch; The association attribute mapping relationship network is adjusted to obtain a plurality of attribute mapping feature subsets paired with the association attribute mapping relationship network, wherein the relationship network size of each of the plurality of attribute mapping feature subsets is smaller than the relationship network size of the association attribute mapping relationship network; At least one of the signal attribute mapping relationship networks paired with the bidirectional long short-term memory branch is updated based on the plurality of attribute mapping feature subsets to obtain a second decision tree model; The second decision tree model is debugged to obtain the target multi-element decision tree, wherein the target multi-element decision tree is used for processing pressure sensor detection data streams; Before the second decision tree model is debugged to obtain the target multi-element decision tree, the method further includes: The sample fusion feature set paired with the detection feature fusion branch in the second decision tree model is adjusted to obtain a plurality of local fusion feature sets matched with the sample fusion feature set, wherein the relationship network size of each of the plurality of local fusion feature sets is smaller than the relationship network size of the sample fusion feature set; The sample fusion feature set paired with the detection feature fusion branch is updated using the plurality of local fusion feature sets to obtain the second decision tree model to be debugged; The second decision tree model is debugged to obtain the target multi-element decision tree, wherein the target multi-element decision tree is used for processing pressure sensor detection data streams, including: A historical detection data set is obtained, wherein the historical detection data set includes a plurality of pressure sensor detection data streams and training annotations paired with the plurality of pressure sensor detection data streams, respectively; The second decision tree model is debugged using the historical detection data set; On the basis that the second decision tree model does not meet the debugging requirements, a model weight associated with the second decision tree model is optimized, wherein the model weight is a model variable in a model variable set indicated by the plurality of attribute mapping feature subsets and the plurality of local fusion feature sets; On the basis that the second decision tree model meets the debugging requirements, the second decision tree model is determined as the target multi-element decision tree.

9. A pressure sensor performance detection system, characterized by, including: a memory for storing program instructions and data; a processor coupled with the memory, for executing the instructions in the memory to implement the method of any one of claims 1-8.

10. A computer storage medium, characterized in that, instructions that, when executed on a processor, implement the method of any one of claims 1-8.

Images