WO2020107445A1 - 一种传感器的自动校准方法和系统 - Google Patents

一种传感器的自动校准方法和系统 Download PDF

Info

Publication number
WO2020107445A1
WO2020107445A1 PCT/CN2018/118686 CN2018118686W WO2020107445A1 WO 2020107445 A1 WO2020107445 A1 WO 2020107445A1 CN 2018118686 W CN2018118686 W CN 2018118686W WO 2020107445 A1 WO2020107445 A1 WO 2020107445A1
Authority
WO
WIPO (PCT)
Prior art keywords
sensor
data
calibration
module
sensors
Prior art date
Application number
PCT/CN2018/118686
Other languages
English (en)
French (fr)
Inventor
张巧丽
孙宝石
Original Assignee
苏州数言信息技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 苏州数言信息技术有限公司 filed Critical 苏州数言信息技术有限公司
Priority to PCT/CN2018/118686 priority Critical patent/WO2020107445A1/zh
Publication of WO2020107445A1 publication Critical patent/WO2020107445A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01DMEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D18/00Testing or calibrating apparatus or arrangements provided for in groups G01D1/00 - G01D15/00

Definitions

  • the invention relates to the field of automatic calibration, in particular to an automatic calibration method and system for sensors.
  • the patent "Sensor Calibration Apparatus and Method" with application number 201310722409.X includes: an I2C interface that is connected to the sensor via an I2C signal line; and a calibration module that communicates with the sensor via the I2C interface.
  • the calibration module includes : Calibration core, which can provide the calibration value to the sensor to calibrate the output of the sensor; Start value register, which stores the initial calibration value, the calibration kernel writes the initial calibration value as the calibration value to the sensor via the I2C interface; Target value register, which Store the expected output value of the sensor; and the tolerance register, which stores the tolerance value, where the calibration kernel reads the output value of the sensor after calibration based on the calibration value, and judges whether the difference between it and the expected output value is within the tolerance value If it is within the tolerance value, the calibration of the sensor is successfully completed, otherwise the calibration value is modified and the new calibration value is written to the sensor until the difference between the sensor's output value and the expected output value is within the tolerance value.
  • This type of method is a simple value adjustment calibration method, and its limitations are:
  • the calibration value needs to be adjusted continuously, which is better for simple linear sensors, but for non-linear sensors, sensors that do not meet the linear law, there is no way to do calibration
  • the other is the calibration using the generated coefficient table.
  • the patent "Sensor Calibration Method" with application number 201710258680.0 numbering M sensors; calculating the calibration coefficient of each sensor, and generating a sensor deviation calibration coefficient table including the number of each sensor and the calibration coefficient of each sensor; calculation The calibration coefficient of the joint module when the joint module is used with each sensor, and generates a calibration coefficient table of the joint module deviation including the number of each sensor and the calibration coefficient of the joint module when the joint module is used with each sensor; The sensor deviation calibration coefficient table and the joint module deviation calibration coefficient table generate the comprehensive deviation calibration coefficient table of M sensors and the joint module; and the client that installs the sensor, the joint module and the comprehensive deviation calibration coefficient table from the sensor number In the deviation calibration coefficient table, the corresponding comprehensive calibration coefficient is called to calibrate the sensor.
  • the calibration formula varies with the environment. It is difficult to accurately calibrate the sensor by adjusting the coefficient.
  • the disadvantages of the sensor calibration method and equipment in the prior art include the following aspects:
  • the method is one-sided: only supports the sensor calibration that meets the rules of a certain formula. It is impossible to calibrate the sensor with complex and changeable calibration formulas as the external conditions change.
  • the senor is not adaptive, and it needs to be re-calibrated after a period of use or when the external conditions change.
  • the purpose of the present invention is to provide an automatic sensor calibration method and system for solving the technical problems of traditional sensor calibration with high one-sidedness, poor adaptability, low degree of automation, and inability to record data.
  • the present invention proposes the following technical solutions:
  • An automatic sensor calibration system includes a sensor terminal, a gateway, a cloud platform, a standard measurement module, a calibration module, and a big data platform;
  • the sensor terminal includes a plurality of sensors, the sensors are target calibration objects, and the sensors are used to collect sensor data and upload to the gateway;
  • the gateway is used to connect the sensor to the cloud platform and transfer the sensor data uploaded by the sensor to the cloud platform;
  • the standard measurement module, calibration module, and big data platform are all connected to the cloud platform;
  • the standard measurement module is used to obtain the true value data of the sensor data corresponding to each sensor and transmit it to the cloud platform;
  • the cloud platform is used to transfer the obtained sensor data and real value data to the big data platform;
  • the big data platform is used to store and transfer the acquired sensor data, real value data, and other statistical data that automatically statistics and affects the accuracy of the sensor to the calibration module;
  • the calibration module is used to analyze the obtained sensor data, real value data, and other statistical data that automatically statistics and affects the accuracy of the sensor, to obtain the corresponding calibration model of the target calibrated sensor;
  • the calibration model is used to calibrate data during sensor calibration and output the calibrated results.
  • a part of the sensors in the sensor terminal constitute a basic environment sensor
  • the basic environment sensor is a sensor corresponding to a pre-selected physical parameter that affects the accuracy of other sensors.
  • the calibration module includes a data acquisition module, a data filtering module, and a data modeling module;
  • the data obtaining module is used to obtain data from a big data platform
  • the data filtering module is used to filter and filter the data in the acquired data module to remove abnormal data and duplicate data;
  • the data modeling module is used to generate a correlation condition list, a calibration ranking list, and a calibration model for the target calibrated sensor;
  • the correlation condition list is composed of sensors or other condition data corresponding to physical parameters that affect the accuracy of the target calibrated sensor selected from the basic environmental sensors and other condition data through variable correlation calculation methods;
  • the calibration ranking list refers to the degree of influence of each sensor on the accuracy of sensors in the entire system sorted out with reference to the correlation list corresponding to the sensors calibrated by each target, and the magnitude of the influence on the accuracy of sensors in the entire system Sort in order;
  • the calibration model is used to calibrate the target calibrated sensor in the order of the calibration ranking list and with reference to the real value data.
  • the system further includes a machine learning module, the machine learning module is connected to the big data platform and the calibration module;
  • the machine learning module is used to enable and generate a new calibration model to replace the calibration model generated by the calibration module when the accuracy of the calibration model generated by the calibration module is low.
  • the machine learning module is used to obtain the correlation condition list of the target calibrated sensor from the calibration module, and obtain the sensor data in the correlation list from the big data platform, and characterize the sensor data; meanwhile The machine learning module is used to obtain the sensor data, real value data of the target calibrated sensor and other condition data in the correlation condition list from the big data platform;
  • the machine learning module performs machine learning based on the obtained data to obtain a calibration model of the target-calibrated sensor.
  • the machine learning module learns according to the following method:
  • Step 1 Refer to the correlation condition list to simulate different environmental conditions for the target calibrated sensor
  • Step 2 Record the sensor data of the target calibrated sensor under various environmental conditions over a period of time, and obtain its characteristic value, and at the same time obtain the true value under the corresponding environmental conditions;
  • Step 3 Perform machine learning with reference to real values and feature values to obtain a calibration model.
  • the accuracy index of the calibration model generated by the calibration module itself includes the sum variance, the sum of square errors, the mean square error, the root mean square, the standard deviation, and the determination coefficient One or more.
  • the calibration model is stored on the sensor terminal and/or cloud platform.
  • the invention also discloses an automatic calibration method of the sensor, which is consistent with the basic idea of the above system, and is introduced as follows.
  • the sensor collects sensor data and uploads it to the gateway;
  • the gateway connects the sensor to the cloud platform and transfers the sensor data uploaded by the sensor to the cloud platform;
  • the standard measurement module obtains the true value data of the sensor data corresponding to each sensor and transmits it to the cloud platform;
  • the cloud platform transfers the obtained sensor data and real value data to the big data platform
  • the big data platform stores the acquired sensor data, real value data, and other statistical data that automatically statistics and affects the accuracy of the sensor and passes it to the calibration module;
  • the calibration module analyzes the obtained sensor data, real value data and other statistical data that automatically statistics and affects the accuracy of the sensor to obtain the corresponding calibration model of the target calibration sensor;
  • the calibration model calibrates data during sensor calibration and outputs the calibrated results.
  • the calibration module includes a data acquisition module, a data filtering module, and a data modeling module;
  • the data obtaining module obtains data from a big data platform
  • the data filtering module filters and filters the data in the acquired data module to remove abnormal data and duplicate data
  • the data modeling module generates a correlation condition list, a calibration ranking list, and a calibration model for the target calibrated sensor
  • the correlation condition list is composed of sensors or other condition data corresponding to physical parameters that affect the accuracy of the target calibrated sensor selected from the basic environmental sensors and other condition data through variable correlation calculation methods;
  • the basic environmental sensor is a sensor corresponding to a physical parameter pre-selected from the sensor terminal and affecting the accuracy of other sensors;
  • the calibration ranking list refers to the degree of influence of each sensor on the accuracy of sensors in the entire system sorted out with reference to the correlation list corresponding to the sensors calibrated by each target, and the magnitude of the influence on the accuracy of sensors in the entire system Sort in order;
  • the calibration model calibrates the target calibrated sensor in the order of the calibration ranking list and with reference to the real value data.
  • the machine learning module when the accuracy of the calibration model generated by the calibration module is low, the machine learning module is enabled and a new calibration model is generated to replace the calibration model generated by the calibration module itself;
  • the machine learning module obtains the correlation condition list of the target calibrated sensor from the calibration module, and obtains the sensor data in the correlation list from the big data platform, and characterizes the sensor data; at the same time, the machine learning The module obtains the sensor data, real value data of the target calibrated sensor and other condition data in the correlation condition list from the big data platform;
  • the machine learning module performs machine learning based on the obtained data to obtain a calibration model of the target-calibrated sensor.
  • the technical solution of the present invention provides a set of sensor detection methods and systems that support the calibration of multiple sensors.
  • the technical solution of the present invention has the following advantages over traditional technical solutions:
  • the method and system support batch calibration of sensors, which can calibrate a large number of sensors at the same time, saving time and labor.
  • the method and system can generate calibration models for various types of sensors, various linear and non-linear sensors, various directly integrated sensor components or integrated sensor products can be supported, and can be automatically adjusted according to changes in the environment Calibration model of the sensor.
  • the operation is simple.
  • the system automatically collects data, calculates data, and generates a calibration model.
  • the model can be directly stored in the sensor terminal or can be stored in the cloud. It supports offline and online use.
  • FIG. 2 is a schematic diagram of the functional structure of the calibration module of the present invention.
  • FIG. 3 is a schematic diagram of the functional structure of the machine learning module of the present invention.
  • an automatic sensor calibration system includes a sensor terminal, a gateway, a cloud platform, a standard measurement module, a calibration module, a big data platform, and a management control terminal.
  • the sensor terminal includes multiple sensors, and the types of the sensors may also be multiple types, such as temperature and humidity, brightness, color, formaldehyde, CO2, CO, PM2.5, PM10, TVOC, current, and so on.
  • the sensor terminal can be an integrated sensor component or a finished integrated sensor. The difference between them is that integrated components require the terminal's own MCU to collect the signal, and then convert the signal value into a measured value, and the integrated product can directly obtain the measured value.
  • the sensor is a target calibration object, and the sensor is used to collect sensor data and upload to the gateway.
  • the gateway is used for network connection between the sensor and the cloud platform, and transmits the sensor data uploaded by the sensor to the cloud platform.
  • the sensor terminal can bring its own gateway function according to demand, then the sensor data can be directly transmitted to the cloud platform.
  • the standard measurement module, calibration module, big data platform and management control terminal are all connected to the cloud platform.
  • the standard metering module needs to integrate a peripheral standard meter to obtain the true value data of the sensor data corresponding to each sensor and pass it to the cloud platform.
  • the standard metering module is only used during calibration, and this module is not required when the sensor is actually used.
  • the cloud platform is used to transfer the acquired sensor data and real value data to the big data platform.
  • the big data platform is used to transfer the acquired sensor data, real value data and other condition data (including but not limited to sensor usage time, running time, etc.) that automatically statistics and affect the accuracy of the sensor to the calibration module, and these data It is also stored on the big data platform.
  • real value data and other condition data including but not limited to sensor usage time, running time, etc.
  • the calibration module is the core part of the specific embodiment of the present invention, and is used to analyze the obtained sensor data, real value data, and other statistical data that automatically statistics and affects the accuracy of the sensor to obtain the corresponding calibration of the target calibrated sensor model.
  • the calibration model is used to calibrate data during sensor calibration and output the calibrated results.
  • the management control terminal is used to provide the user with a view of each sensor and to facilitate the calibration of each sensor.
  • the applicant tried to find the weight of different sensors on the system based on the influencing factors of the sensors, thereby adjusting the sensors in the order of weight to ensure the efficient calibration of the entire system.
  • the specific embodiment of the present invention proposes the concept of "basic environment sensor", that is, a part of the sensors in the sensor terminal constitute a basic environment sensor, and the basic environment sensor is a pre-selected one that affects the accuracy of other sensors The sensor corresponding to the physical parameter.
  • a temperature sensor that is related to detecting temperature can be classified as a basic environmental sensor.
  • the above specific basic environmental sensors can be flexibly selected and combined according to their own needs. Taking a step back, if the selection of basic environmental sensors cannot be determined, all sensors with possible correlations can be included. Except for the increase in the number of data samples that need to be collected and processed, the algorithm and process remain unchanged.
  • the combination of the above basic environmental sensors is to select the corresponding correlation condition list for each specific sensor.
  • the calibration module is described in detail below.
  • the calibration module includes a data acquisition module, a data filtering module and a data modeling module.
  • the data acquisition module is used to acquire data from a big data platform, that is, sensor data, real value data, and other condition data that are automatically calculated and affect the accuracy of the sensor.
  • the data filtering module is used for filtering and filtering the data in the acquired data module to remove abnormal data and duplicate data to obtain an effective data set.
  • the data modeling module is used to generate a correlation condition list, a calibration ranking list, and a calibration model for the target calibrated sensor.
  • the correlation condition list is obtained by the correlation condition selection module in the data modeling module, which is selected from basic environmental sensors and other condition data through variable correlation calculation methods (including but not limited to Pearson correlation algorithm) It consists of sensor or other condition data (including but not limited to sensor usage time, running time, etc.) corresponding to the physical parameters that affect the accuracy of the target calibrated sensor.
  • the calibration sorting list needs to sort and calibrate different types of sensors in the system, that is, the sorting according to the influence of the sensors on the system described above. That is, each sensor sorted with reference to the correlation list corresponding to the sensors calibrated by each target affects the degree of accuracy of the sensors in the entire system, and is sorted in order of affecting the accuracy of the sensors in the entire system.
  • the weights of the weighted sensors can be accumulated to obtain the degree of influence of various sensors on the accuracy of the sensors in the entire system, and sorted in the order of influence.
  • the above weight accumulation method includes, but is not limited to, summing the weights of the corresponding sensors distributed in the respective correlation lists to obtain the weights of the sensors for the entire system.
  • the various sensors are sorted in the order of the above weights relative to the entire system, and the calibration sequence is performed in this order.
  • the accuracy index of the calibration model generated by the calibration module itself includes one or more of sum variance, square sum error, mean square deviation, root mean square, standard deviation, and determination coefficient between the calibration result and the true value.
  • the machine learning module is connected to the big data platform and the calibration module.
  • the machine learning module is used to enable and generate a new calibration model to replace the calibration model generated by the calibration module when the accuracy of the calibration model generated by the calibration module is low.
  • the machine learning module is used to obtain the correlation condition list of the target calibrated sensor from the calibration module, and obtain the sensor data in the correlation list from the big data platform, and characterize the sensor data; meanwhile The machine learning module is used to obtain the sensor data, real value data and other condition data in the correlation condition list of the target calibrated sensor from the big data platform.
  • the machine learning module performs machine learning based on the obtained data to obtain a calibration model of the target-calibrated sensor.
  • the machine learning module learns as follows:
  • Step 1 Refer to the correlation condition list to simulate different environmental conditions for the target calibrated sensor.
  • Step 2 Record the sensor data of the target calibrated sensor under various environmental conditions over a period of time, and obtain its characteristic value, and at the same time obtain the true value under the corresponding environmental conditions.
  • Step 3 Perform machine learning with reference to real values and feature values to obtain a calibration model.
  • the calibration model is stored on the sensor terminal and/or the cloud platform.
  • the sensor data collected by the sensor the sensor data that affects the sensor, and other influencing factor data need to be passed into the calibration model to obtain data that is closer to the true value after calibration.
  • the sensor collects sensor data during the calibration phase and actual use phase and uploads it to the gateway.
  • the gateway connects the sensor to the cloud platform through the network, and transmits the sensor data uploaded by the sensor to the cloud platform.
  • the standard measurement module acquires the true value data of the sensor data corresponding to each sensor during the calibration stage and transmits it to the cloud platform.
  • the cloud platform transfers the obtained sensor data and real value data to the big data platform.
  • the big data platform transfers the acquired sensor data, true value data, and other statistical data that automatically statistics and affects the accuracy of the sensor to the calibration module.
  • the calibration module analyzes the obtained sensor data, real value data, and other statistical data that automatically statistics and affects the accuracy of the sensor, and obtains the corresponding calibration model of the target calibrated sensor.
  • the calibration model calibrates data during sensor calibration and outputs the calibrated results.
  • the calibration module includes a data acquisition module, a data filtering module, and a data modeling module.
  • the data obtaining module obtains data from a big data platform.
  • the data filtering module filters and filters the data in the acquired data module to remove abnormal data and duplicate data.
  • the data modeling module generates a correlation condition list, a calibration ranking list, and a calibration model for the target calibrated sensor.
  • the correlation condition list is composed of sensors or other condition data corresponding to physical parameters that affect the accuracy of the target calibrated sensor selected from the basic environmental sensors and other condition data through variable correlation calculation methods;
  • the basic environmental sensor is a sensor corresponding to a physical parameter pre-selected from the sensor terminal and affecting the accuracy of other sensors.
  • the calibration ranking list refers to the degree of influence of each sensor on the accuracy of sensors in the entire system sorted out with reference to the correlation list corresponding to the sensors calibrated by each target, and the magnitude of the influence on the accuracy of sensors in the entire system Sort in order.
  • the calibration model calibrates the target calibrated sensor in the order of the calibration ranking list and with reference to the real value data.
  • the machine learning module is activated and a new calibration model is generated to replace the calibration model generated by the calibration module itself.
  • the machine learning module obtains the correlation condition list of the target calibrated sensor from the calibration module, and obtains the sensor data in the correlation list from the big data platform, and characterizes the sensor data; at the same time, the machine learning The module obtains the sensor data, real value data of the target calibrated sensor and other condition data in the correlation condition list from the big data platform.
  • the machine learning module performs machine learning based on the obtained data to obtain a calibration model of the target-calibrated sensor.
  • the user starts the calibration mode, the instruction is sent to the sensor terminal through the cloud platform, the sensor receives the instruction, enters the calibration mode, starts to quickly collect data, and sends the data to the cloud platform, after the cloud platform receives the data, processes the data and join
  • the current condition data (temperature, humidity, device duration) is stored in the big data platform.
  • the standard meter needs to be turned on to obtain the true value of the sensor at the same time.
  • manual intervention is required, such as the true value of formaldehyde.
  • the system automatically tags the data according to the time when the sensor collects data and real data, and starts model training.
  • a calibration model is obtained, which corresponds to the sensor and stored in the cloud platform.
  • the sensor will have a unique ID in the system According to this unique ID, the calibration model of this sensor can be found, so it can be calibrated online in the cloud.
  • the calibration model can also be delivered to the sensor terminal through the cloud platform, and the sensor itself records the calibration.
  • the system issues instructions to make the sensor enter the working mode.
  • the sensor uses the calibration model to calibrate the data and output the calibrated result.
  • the above system and method are applied to, for example, a school classroom indoor environment system, and the system is calibrated.
  • the system contains air quality detection sensors, temperature and humidity sensors, and brightness sensors. Users can see the current environmental status through the management control terminal. Since the sensors are purchased directly from existing products, there is a certain error in the accuracy of the sensor, and the sensor in the entire environment needs to be calibrated. Integrate the sensor into the automatic calibration system. The user uses the interface to start calibration and start collecting data. At the same time, a standard meter is required to input real data. Since the air quality sensor is sensitive to temperature, temperature changes will affect the accuracy of the sensor, so it needs to be collected in each Sensor data under different temperatures and different air quality environments. After a large amount of data collection, the system automatically creates a good model for the sensor after training and completes automatic calibration. In this way, the sensor can get more accurate results at different temperatures. The effect of temperature on the sensor is eliminated.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

一种传感器的自动校准系统和方法。传感器用于采集传感器数据并上传至网关;网关将传感器所上传的传感器数据传递给云平台;标准计量模块用于获取各个传感器对应的传感器数据的真实值数据,并传递给云平台;云平台用于将获得的传感器数据、真实值数据传递给大数据平台;大数据平台用于将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据存储并传递给校准模块;校准模块用于对获得的数据进行分析,获得目标校准传感器相应的校准模型或依靠机器学习获得校准模型;校准模型用于在传感器校准时校准数据并输出校准后的结果。

Description

一种传感器的自动校准方法和系统 技术领域
本发明涉及自动校准领域,特别是对传感器的自动校准方法和系统。
背景技术
随着物联网领域的快速发展,传感器的应用也越发的广泛,深入到各行各业,大量的传感器应用在各种不同的环境中,而不同的环境会对传感器的电子器件产生一定程度的影响,温度、湿度、电磁场、器件的使用时长等等都会对传感器产生影响,造成传感器数据的不准确,甚至越来越偏离实际值。因此传感器的校准也越发的重要。但是,即使是同样的传感器器件本身也是缺乏一致性的,对一致性差的传感器,需要对每一批次甚至是每一个传感器成品进行校验,而目前,大多数的传感器校准方法都需要现场人工通过一定的装置系统来进行校验,操作麻烦且耗费人力。而校准后的传感器在实际使用中由于环境的变化,各个外界条件的改变导致传感器又需要重新校准,十分不方便。
下面通过几个具体的方案介绍现有技术中常见的传感器校准方法和装置。
一种是进行简单的值校准。例如申请号为201310722409.X的专利“传感器校准装置和方法”,该装置包括:I2C接口,其经由I2C信号线连接至传感器;以及校准模块,其经由该I2C接口与传感器通信,该校准模块包括:校准内核,其能向传感器提供校准值以校准传感器的输出;起始值寄存器,其存储初始化校准值,校准内核经由I2C接口将该初始化校准值作为校准值写入传感器;目标值寄存器,其存储传感器的期望输出值;以及容差寄存器,其存储容差值,其中校准内核读取传感器基于校准值进行校准之后的输出值,并判断其与期望输出值之差是否在容差值之内,如果在容差值之内则成功完成对传感器的校准,否则修改校准值并将新的校准值写入传感器,直至传感器的输出值与期望输出值之差在容差值之内。
这类方法属于简单的值调整校准方法,局限性在于:
1、需要不停的调整校准值,对简单的线性传感器效果较好,但对于非线性传感器,不满足线性规律的传感器,没办法做校准
2、适应性差,换个环境需要重新校准。
另一种是利用生成系数表校准。例如申请号201710258680.0的专利“传感器校准方法”:对M个传感器进行编号;计算每个传感器的校准系数,并生成包括每个传感器的编号及每个传感器的校准系数的传感器偏差校准系数表;计算联用模块与每个传感器联用时联用模块的校准系数,并生成包括每个传感器的编号及联用模块与每个传感器联用时联用模块的校准系数的联用模块偏差校准系数表;根据传感器偏差校准系数表及联用模块偏差校准系数表生成M个传感器与联用模块的综合偏差校准系数表;以及安装传感器、联用模块及综合偏差校准系数表的客户端根据传感器的编号从综合偏差校准系数表中调用相应的综合校准系数对传感器进行校准。
这类方法需要为每个传感器生成一个校准系数,局限性在于:
1、对于复杂的非线性传感器,校准公式随环境复杂多变,仅仅通过系数的调整很难对传感器进行准确地校准。
2、没有自动化系统的概念,一一为每个传感器编号校准效率低下。
综上所述,现有技术中对传感器的校准方法和设备的缺点包括如下几个方面:
一、方法片面性:仅支持满足某种特定公式规律的传感器校准,对随着外界条件变化,校准公式复杂多变的传感器无法校准。
二、传感器没有自适应性,使用一段时间或者外界条件改变后,需要重新校准。
三、缺乏自动化,对人工依赖大,且校准之后无法适应多变的环境,一旦环境变化,需要重新校准。
四、没有系统化,采样数据、校准数据、校准历史等具有价值的数据没有记录,没法追溯,且没有进行大量有意义的分析。
发明内容
本发明目的在于提供一种传感器的自动校准方法和系统,用于解决传统的传感器校准片面性高、适应性差、自动化程度低且无法进行数据记录的技术问题。
为达成上述目的,本发明提出如下技术方案:
一种传感器的自动校准系统,包括传感器终端、网关、云平台、标准计量模块、校准模块、大数据平台;
所述传感器终端包括多个传感器,所述传感器为目标校准对象,且所述传感器用于采集传感器数据并上传至网关;
所述网关用于将传感器和云平台进行网络连接,并将传感器所上传的传感器数据传递给云平台;
所述标准计量模块、校准模块、大数据平台均连接于云平台上;
所述标准计量模块用于获取各个传感器对应的传感器数据的真实值数据,并传递给云平台;
所述云平台用于将获得的传感器数据、真实值数据传递给大数据平台;
所述大数据平台用于将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据存储并传递给校准模块;
所述校准模块用于对获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行分析,获得目标校准的传感器相应的校准模型;
所述校准模型用于在传感器校准时校准数据并输出校准后的结果。
进一步的,在本发明中,所述传感器终端中的一部分传感器组成基础环境传感器,所述基础环境传感器为预选出的对其他传感器的准确度造成影响的物理参数所对应的传感器。
进一步的,在本发明中,所述校准模块包括获取数据模块、数据过滤模块以及数据建模模块;
所述获取数据模块用于从大数据平台上获取数据;
所述数据过滤模块用于将获取数据模块中的数据进行过滤筛选,以去除异常数据以及重复数据;
所述数据建模模块用于对目标校准的传感器生成相关性条件列表、校准排序列表以及校准模型;
所述相关性条件列表为由通过变量相关性计算方法从基础环境传感器、其他条件数据中筛选出的对目标校准的传感器的准确度造成影响的物理参数所对应的传感器或其他条件数据组成;
所述校准排序列表为参照各个目标校准的传感器对应的相关性列表排序出的每个传感器对整个系统中的传感器的准确度的影响程度,并以对整个系统中的传感器的准确度影响大小的顺序进行排序;
所述校准模型用于按照校准排序列表的顺序并参照真实值数据对目标校准的传感器进行校准。
进一步的,在本发明中,所述系统还包括机器学习模块,所述机器学习模块连接于大数据平台和校准模块上;
所述机器学习模块用于校准模块自身生成的校准模型准确度较低时启用并生成新的校准模型以替代校准模块自身生成的校准模型;
所述机器学习模块用于从校准模块处获得目标校准的传感器的相关性条件列表,并从大数据平台中获得上述相关性列表中的传感器数据,将这些传感器数据进行特征化处理;同时所述机器学习模块用于从大数据平台上获得目标校准的传感器的传感器数据、真实值数据以及相关性条件列表中的其他条件数据;
所述机器学习模块基于上述获得的数据进行机器学习,获得目标校准的传感器的校准模型。
进一步的,在本发明中,所述机器学习模块按照如下方法进行学习:
步骤1、参照相关性条件列表,为目标校准的传感器模拟不同的环境条件;
步骤2、记录一段时间内各个环境条件下该目标校准的传感器的传感器数据,并求取其特征值,同时获得对应环境条件下的真实值;
步骤3、参照真实值和特征值进行机器学习,获得校准模型。
进一步的,在本发明中,所述校准模块自身生成的校准模型的准确度指标包括校准结果与真实值之间的和方差、误差平方和、均方差、均方根、标准差、确定系数中的一种或几种。
进一步的,在本发明中,所述校准模型存储于传感器终端和/或云平台上。
本发明同时还公开了一种传感器的自动校准方法,该自动方法与上述系统基本思想一致,介绍如下。
传感器采集传感器数据并上传至网关;
网关将传感器和云平台进行网络连接,并将传感器所上传的传感器数据传递给云平台;
标准计量模块获取各个传感器对应的传感器数据的真实值数据,并传递给云平台;
云平台将获得的传感器数据、真实值数据传递给大数据平台;
大数据平台将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行存储并传递给校准模块;
校准模块对获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行分析,获得目标校准的传感器相应的校准模型;
校准模型在传感器校准时校准数据并输出校准后的结果。
进一步的,在上述方法中,校准模块包括获取数据模块、数据过滤模块以及数据建模模块;
所述获取数据模块从大数据平台上获取数据;
所述数据过滤模块将获取数据模块中的数据进行过滤筛选,以去除异常数据以及重复数据;
所述数据建模模块对目标校准的传感器生成相关性条件列表、校准排序列表以及校准模型;
所述相关性条件列表为由通过变量相关性计算方法从基础环境传感器、其他条件数据中筛选出的对目标校准的传感器的准确度造成影响的物理参数所对应的传感器或其他条件数据组成;所述基础环境传感器为从传感器终端中预选出的对其他传感器的准确度造成影响的物理参数所对应的传感器;
所述校准排序列表为参照各个目标校准的传感器对应的相关性列表排序出的每个传感器对整个系统中的传感器的准确度的影响程度,并以对整个系统中的传感器的准确度影响大小的顺序进行排序;
所述校准模型按照校准排序列表的顺序并参照真实值数据对目标校准的传感器进行校准。
进一步的,在上述方法中,所述校准模块自身生成的校准模型准确度较低时启用机器学习模块并生成新的校准模型替代校准模块自身生成的校准模型;
所述机器学习模块从校准模块处获得目标校准的传感器的相关性条件列表,并从大数据平台中获得上述相关性列表中的传感器数据,将这些传感器数据进行特征化处理;同时所述机器学习模块从大数据平台上获得目标校准的传感器的传感器数据、真实值数据以及相关性条件列表中的其他条件数据;
所述机器学习模块基于上述获得的数据进行机器学习,获得目标校准的传感器的校准模型。
有益效果:
由以上技术方案可知,本发明的技术方案提供了一套传感器检测的方法和系统,支持多种传感器的校准,本发明的技术方案相比于传统的技术方案具有以下优势:
1、能够支持各种复杂多变的传感器的自动校准,具有通用性。
2、该方法和系统支持传感器的批量校准,可以同时对大批量的传感器进行校准,节省时间和人力。
3、该方法和系统可以为多种类型的传感器生成校准模型,各种线性、非线性规律的传感器,各种直接集成传感器元器件或者集成传感器成品均能支持,且能根据环境的变化自动调整传感器的校准模型。
4、操作简单,系统自动化采集数据、运算数据、生成校准模型,可将模型直接存储到传感器终端,也可以存储在云端,同时支持离线和在线使用。
应当理解,前述构思以及在下面更加详细地描述的额外构思的所有组合只要在这样的构思不相互矛盾的情况下都可以被视为本公开的发明主题的一部分。
结合附图从下面的描述中可以更加全面地理解本发明教导的前述和其他方面、实施例和特征。本发明的其他附加方面例如示例性实施方式的特征和/或有益效果将在下面的描述中显见,或通过根据本发明教导的具体实施方式的实践中得知。
附图说明
附图不意在按比例绘制。在附图中,在各个图中示出的每个相同或近似相同的组成部分可以用相同的标号表示。为了清晰起见,在每个图中,并非每个组成部分均被标记。现在,将通过例子并参考附图来描述本发明的各个方面的实施例,其中:
图1为本发明的系统的结构示意图;
图2为本发明的校准模块的功能结构示意图;
图3为本发明的机器学习模块的功能结构示意图;
图4为本发明的校准模块的完整流程示意图;
[根据细则26改正28.01.2019] 
图5为本发明的整个系统校准流程示意图。
具体实施方式
为了更了解本发明的技术内容,特举具体实施例并配合所附图式说明如下。
在本公开中参照附图来描述本发明的各方面,附图中示出了许多说明的实 施例。本公开的实施例不必定意在包括本发明的所有方面。应当理解,上面介绍的多种构思和实施例,以及下面更加详细地描述的那些构思和实施方式可以以很多方式中任意一种来实施,这是因为本发明所公开的构思和实施例并不限于任何实施方式。另外,本发明公开的一些方面可以单独使用,或者与本发明公开的其他方面的任何适当组合来使用。
如图1所示的一种传感器的自动校准系统,包括传感器终端、网关、云平台、标准计量模块、校准模块、大数据平台以及管理控制端。
所述传感器终端包括多个传感器,传感器的类型也可以是多种,如温湿度、光亮度、颜色、甲醛、CO2、CO、PM2.5、PM10、TVOC、电流等等。传感器终端可以为集成传感器元器件,也可以为集成传感器成品。它们的差异在于,集成元器件,需终端自身的MCU采集信号,再由信号值转换为测量值,而集成成品,可以直接拿到测量值。
所述传感器为目标校准对象,且所述传感器用于采集传感器数据并上传至网关。
所述网关用于将传感器和云平台进行网络连接,并将传感器所上传的传感器数据传递给云平台。
当然,传感器终端可根据需求自带网关功能,那么传感器数据则可以直接向云平台传输。
所述标准计量模块、校准模块、大数据平台、管理控制端均连接于云平台上。
所述标准计量模块需集成外围标准计量仪,用于获取各个传感器对应的传感器数据的真实值数据,并传递给云平台。标准计量模块只在校准时使用,传感器实际使用时无需此模块。
所述云平台用于将获得的传感器数据、真实值数据传递给大数据平台。
所述大数据平台用于将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据(包括但不限于传感器使用时长、运行时长 等)传递给校准模块,同时这些数据也存储于大数据平台上。
所述校准模块是本发明的具体实施例的核心部分,用于对获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行分析,获得目标校准的传感器相应的校准模型。
所述校准模型用于在传感器校准时校准数据并输出校准后的结果。
所述管理控制端用于提供用户对各个传感器的查看,以及对各个传感器校准提供便捷化。
由于整个系统中的传感器种类较多,影响传感器的准确度的因素较多,因此,对传感器的校准可谓是牵一发而动全身。故如何能够对整个系统中的传感器进行合理有效的校准,是本发明需要切实解决的问题。
针对上述问题,申请人从传感器的影响因素出发,试图找到不同的传感器对系统的影响权重,由此,按照权重大小的顺序调节传感器,从而确保整个系统高效率的校准。
因此,本发明的具体实施例提出了“基础环境传感器”的概念,即将所述传感器终端中的一部分传感器组成基础环境传感器,所述基础环境传感器为预选出的对其他传感器的准确度造成影响的物理参数所对应的传感器。
例如温度对电子元器件有影响,因此温度必然对很多传感器的准确度有影响,因此与检测温度相关的传感器即温度传感器即可归类为基础环境传感器。
上述具体的基础环境传感器的可以根据自身的需要,灵活的选取进行组合。退一步讲,如果无法确定基础环境传感器的选取,则可以将有相关性可能的传感器都包含进来,除了需要采集和处理的数据样本增加外,算法和过程不变。
上述基础环境传感器的组合是为了接下来为每个具体的传感器选出对应的相关性条件列表。
这需要使用校准模块,下面具体介绍校准模块。
如图2、图3和图4所示,所述校准模块包括获取数据模块、数据过滤模块以及数据建模模块。
所述获取数据模块用于从大数据平台上获取数据,即传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据。
所述数据过滤模块用于将获取数据模块中的数据进行过滤筛选,以去除异常数据以及重复数据,得到一个有效的数据集。
所述数据建模模块用于对目标校准的传感器生成相关性条件列表、校准排序列表以及校准模型。
所述相关性条件列表由数据建模模块中的相关性条件选择模块获得,该模块通过变量相关性计算方法(包括但不限于皮尔逊相关算法)从基础环境传感器、其他条件数据中筛选出的对目标校准的传感器的准确度造成影响的物理参数所对应的传感器或其他条件数据(包括但不限于传感器使用时长、运行时长等)组成。
所述校准排序列表需要将系统中不同类型的传感器做校准排序,也即前述所述的按照传感器对系统影响大小排序。即参照各个目标校准的传感器对应的相关性列表排序出的每个传感器对整个系统中的传感器的准确度的影响程度,并以对整个系统中的传感器的准确度影响大小的顺序进行排序。
具体来讲,针对传感器A,在相关列表中存在的N个影响传感器A的准确度的参数所对应的传感器,这些传感器对传感器A的影响具有各自的权重。那么对于整个系统而言,可以将有权重的传感器的权重进行累积,获得各种传感器对整个系统中的传感器的准确度的影响程度,并以影响大小的顺序进行排序。
当然,上述权重的累积方法包括但不限于将分布于各个相关列表中对应的传感器的权重进行求和,获得该类传感器对于整个系统的权重。
将各种传感器按照上述相对整个系统的权重大小的顺序进行排序,校准顺序按照该顺序进行。
校准时,参照真实值数据先进行拟合,可以拟合成直线或者多项式曲线,对拟合效果好的传感器,则可以直接得到校准模型,对拟合效果不好的传感器,则需要调用机器学习模块,生成校准模型。
所述校准模块自身生成的校准模型的准确度指标包括校准结果与真实值之间的和方差、误差平方和、均方差、均方根、标准差、确定系数中的一种或几种。
下面针对机器学习模块进行介绍。
所述机器学习模块连接于大数据平台和校准模块上。
所述机器学习模块用于校准模块自身生成的校准模型准确度较低时启用并生成新的校准模型以替代校准模块自身生成的校准模型。
所述机器学习模块用于从校准模块处获得目标校准的传感器的相关性条件列表,并从大数据平台中获得上述相关性列表中的传感器数据,将这些传感器数据进行特征化处理;同时所述机器学习模块用于从大数据平台上获得目标校准的传感器的传感器数据、真实值数据以及相关性条件列表中的其他条件数据。
所述机器学习模块基于上述获得的数据进行机器学习,获得目标校准的传感器的校准模型。
所述机器学习模块按照如下方法进行学习:
步骤1、参照相关性条件列表,为目标校准的传感器模拟不同的环境条件。
步骤2、记录一段时间内各个环境条件下该目标校准的传感器的传感器数据,并求取其特征值,同时获得对应环境条件下的真实值。
步骤3、参照真实值和特征值进行机器学习,获得校准模型。
无论是校准模块还是机器学习获得的校准模型,所述校准模型存储于传感器终端和/或云平台上。
传感器工作中,只需传入传感器采集的传感器数据以及对此传感器有影响的传感器数据以及其他影响因素数据,代入校准模型,便可得到校准后更接近真实值的数据。
本发明的具体实施例的系统对应的一种传感器的自动校准方法如下:
传感器在校准阶段和实际使用阶段采集传感器数据并上传至网关。
网关将传感器和云平台进行网络连接,并将传感器所上传的传感器数据传 递给云平台。
标准计量模块在校准阶段获取各个传感器对应的传感器数据的真实值数据,并传递给云平台。
云平台将获得的传感器数据、真实值数据传递给大数据平台。
大数据平台将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据传递给校准模块。
校准模块对获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行分析,获得目标校准的传感器相应的校准模型。
校准模型在传感器校准时校准数据并输出校准后的结果。
进一步,在上述方法中,校准模块包括获取数据模块、数据过滤模块以及数据建模模块。
所述获取数据模块从大数据平台上获取数据。
所述数据过滤模块将获取数据模块中的数据进行过滤筛选,以去除异常数据以及重复数据。
所述数据建模模块对目标校准的传感器生成相关性条件列表、校准排序列表以及校准模型。
所述相关性条件列表为由通过变量相关性计算方法从基础环境传感器、其他条件数据中筛选出的对目标校准的传感器的准确度造成影响的物理参数所对应的传感器或其他条件数据组成;所述基础环境传感器为从传感器终端中预选出的对其他传感器的准确度造成影响的物理参数所对应的传感器。
所述校准排序列表为参照各个目标校准的传感器对应的相关性列表排序出的每个传感器对整个系统中的传感器的准确度的影响程度,并以对整个系统中的传感器的准确度影响大小的顺序进行排序。
所述校准模型按照校准排序列表的顺序并参照真实值数据对目标校准的传感器进行校准。
作为优选的,所述校准模块自身生成的校准模型准确度较低时启用机器学 习模块并生成新的校准模型替代校准模块自身生成的校准模型。
所述机器学习模块从校准模块处获得目标校准的传感器的相关性条件列表,并从大数据平台中获得上述相关性列表中的传感器数据,将这些传感器数据进行特征化处理;同时所述机器学习模块从大数据平台上获得目标校准的传感器的传感器数据、真实值数据以及相关性条件列表中的其他条件数据。
所述机器学习模块基于上述获得的数据进行机器学习,获得目标校准的传感器的校准模型。
当然,实际应用中,还需要对上述方法进行适应性调整,参照图5。
用户开启校准模式,指令通过云平台下发至传感器终端,传感器收到指令,进入校准模式,开始快速采集数据,并将数据送至云平台,云平台收到数据后,对数据进行加工,加入当前条件数据(温度、湿度、器件使用时长),存入大数据平台。同时,需要开启标准计量仪,获取传感器同一时刻的真实值。对于复杂无法集成到我们的系统中的计量仪,需要人工干预,如甲醛的真实值。之后,系统根据传感器采集数据和真实数据的时间,自动为数据打上标签,并开始模型训练,训练结束后得到校准模型,将此模型对应到传感器存入云平台,传感器在系统中会有唯一ID,根据此唯一ID既能查到此传感器的校准模型,因此可以在云端在线校准。当然也可以将校准模型通过云平台下发至传感器终端,由传感器自身记录校准。至此,校准过程结束,系统下发指令让传感器进入工作模式,传感器采集到数据后,使用校准模型校准数据,输出校准后的结果。
将上述系统和方法应用到例如学校教室室内环境系统,对该系统进行校准。
该系统中包含空气质量检测传感器,温湿度传感器,光亮度传感器,用户通过管理控制端可看到当前环境状态。由于传感器都是直接购买现有产品,传感器精度有一定的误差,需要对整个环境中的传感器做校准。集成传感器到自动校准系统中,用户使用界面开启校准,开始采集数据,同时需要标准计量仪输入真实数据,由于空气质量传感器对温度比较敏感,温度的变化会影响传感器的精度,因此需要采集在各个不同温度,不同空气质量环境下的传感器数据, 在进行的大量的数据采集后,训练后系统自动为传感器创建好模型,完成自动校准。这样,在不同的温度下,传感器均能得到更精确的结果。消除了温度对传感器的影响。
又如,现在有大量好几个批次同种类型传感器,需要做出厂校准,如果这些传感器一致性很好,则只需取一个传感器,集成到自动校准系统中,模拟不同环境,采集数据,进行校准。如果这些传感器各个批次间的一致性不好,但是每个批次间的一致性较好,则从每个批次中取一个传感器,集成到自动校准系统,模拟不同环境,采集数据,进行校准。如果这些传感器每个批次间的一致性不好,则取所有的传感器,集成到系统,一一校准。
虽然本发明已以较佳实施例揭露如上,然其并非用以限定本发明。本发明所属技术领域中具有通常知识者,在不脱离本发明的精神和范围内,当可作各种的更动与润饰。因此,本发明的保护范围当视权利要求书所界定者为准。

Claims (10)

  1. 一种传感器的自动校准系统,其特征在于:包括传感器终端、网关、云平台、标准计量模块、校准模块、大数据平台;
    所述传感器终端包括多个传感器,所述传感器为目标校准对象,且所述传感器用于采集传感器数据并上传至网关;
    所述网关用于将传感器和云平台进行网络连接,并将传感器所上传的传感器数据传递给云平台;
    所述标准计量模块、校准模块、大数据平台均连接于云平台上;
    所述标准计量模块用于获取各个传感器对应的传感器数据的真实值数据,并传递给云平台;
    所述云平台用于将获得的传感器数据、真实值数据传递给大数据平台;
    所述大数据平台用于将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据存储并传递给校准模块;
    所述校准模块用于对获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行分析,获得目标校准的传感器相应的校准模型;
    所述校准模型用于在传感器校准时校准数据并输出校准后的结果。
  2. 根据权利要求1所述的传感器的自动校准系统,其特征在于:所述传感器终端中的一部分传感器组成基础环境传感器,所述基础环境传感器为预选出的对其他传感器的准确度造成影响的物理参数所对应的传感器。
  3. 根据权利要求2所述的传感器的自动校准系统,其特征在于:所述校准模块包括获取数据模块、数据过滤模块以及数据建模模块;
    所述获取数据模块用于从大数据平台上获取数据;
    所述数据过滤模块用于将获取数据模块中的数据进行过滤筛选,以去除异常数据以及重复数据;
    所述数据建模模块用于对目标校准的传感器生成相关性条件列表、校准排 序列表以及校准模型;
    所述相关性条件列表为由通过变量相关性计算方法从基础环境传感器、其他条件数据中筛选出的对目标校准的传感器的准确度造成影响的物理参数所对应的传感器或其他条件数据组成;
    所述校准排序列表为参照各个目标校准的传感器对应的相关性列表排序出的每个传感器对整个系统中的传感器的准确度的影响程度,并以对整个系统中的传感器的准确度影响大小的顺序进行排序;
    所述校准模型用于按照校准排序列表的顺序并参照真实值数据对目标校准的传感器进行校准。
  4. 根据权利要求3所述的传感器的自动校准系统,其特征在于:所述系统还包括机器学习模块,所述机器学习模块连接于大数据平台和校准模块上;
    所述机器学习模块用于校准模块自身生成的校准模型准确度较低时启用并生成新的校准模型以替代校准模块自身生成的校准模型;
    所述机器学习模块用于从校准模块处获得目标校准的传感器的相关性条件列表,并从大数据平台中获得上述相关性列表中的传感器数据,将这些传感器数据进行特征化处理;同时所述机器学习模块用于从大数据平台上获得目标校准的传感器的传感器数据、真实值数据以及相关性条件列表中的其他条件数据;
    所述机器学习模块基于上述获得的数据进行机器学习,获得目标校准的传感器的校准模型。
  5. 根据权利要求4所述的传感器的自动校准系统,其特征在于:所述机器学习模块按照如下方法进行学习:
    步骤1、参照相关性条件列表,为目标校准的传感器模拟不同的环境条件;
    步骤2、记录一段时间内各个环境条件下该目标校准的传感器的传感器数据,并求取其特征值,同时获得对应环境条件下的真实值;
    步骤3、参照真实值和特征值进行机器学习,获得校准模型。
  6. 根据权利要求4所述的传感器的自动校准系统,其特征在于:所述校准 模块自身生成的校准模型的准确度指标包括校准结果与真实值之间的和方差、误差平方和、均方差、均方根、标准差、确定系数中的一种或几种。
  7. 根据权利要求1-6中任意一条所述的传感器的自动校准系统,其特征在于:所述校准模型存储于传感器终端和/或云平台上。
  8. 一种传感器的自动校准方法,其特征在于:所述传感器终端中的传感器采集传感器数据并上传至网关;
    网关将传感器和云平台进行网络连接,并将传感器所上传的传感器数据传递给云平台;
    标准计量模块获取各个传感器对应的传感器数据的真实值数据,并传递给云平台;
    云平台将获得的传感器数据、真实值数据传递给大数据平台;
    大数据平台将获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行存储并传递给校准模块;
    校准模块对获得的传感器数据、真实值数据以及自动统计的且影响传感器准确度的其他条件数据进行分析,获得目标校准的传感器相应的校准模型;
    校准模型在传感器校准时校准数据并输出校准后的结果。
  9. 根据权利要求8所述的传感器的自动校准方法,其特征在于:校准模块包括获取数据模块、数据过滤模块以及数据建模模块;
    所述获取数据模块从大数据平台上获取数据;
    所述数据过滤模块将获取数据模块中的数据进行过滤筛选,以去除异常数据以及重复数据;
    所述数据建模模块对目标校准的传感器生成相关性条件列表、校准排序列表以及校准模型;
    所述相关性条件列表为由通过变量相关性计算方法从基础环境传感器、其他条件数据中筛选出的对目标校准的传感器的准确度造成影响的物理参数所对应的传感器或其他条件数据组成;所述基础环境传感器为从传感器终端中预选 出的对其他传感器的准确度造成影响的物理参数所对应的传感器;
    所述校准排序列表为参照各个目标校准的传感器对应的相关性列表排序出的每个传感器对整个系统中的传感器的准确度的影响程度,并以对整个系统中的传感器的准确度影响大小的顺序进行排序;
    所述校准模型按照校准排序列表的顺序并参照真实值数据对目标校准的传感器进行校准。
  10. 根据权利要求9所述的传感器的自动校准方法,其特征在于:所述校准模块自身生成的校准模型准确度较低时启用机器学习模块并生成新的校准模型替代校准模块自身生成的校准模型;
    所述机器学习模块从校准模块处获得目标校准的传感器的相关性条件列表,并从大数据平台中获得上述相关性列表中的传感器数据,将这些传感器数据进行特征化处理;同时所述机器学习模块从大数据平台上获得目标校准的传感器的传感器数据、真实值数据以及相关性条件列表中的其他条件数据;
    所述机器学习模块基于上述获得的数据进行机器学习,获得目标校准的传感器的校准模型。
PCT/CN2018/118686 2018-11-30 2018-11-30 一种传感器的自动校准方法和系统 WO2020107445A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/118686 WO2020107445A1 (zh) 2018-11-30 2018-11-30 一种传感器的自动校准方法和系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/118686 WO2020107445A1 (zh) 2018-11-30 2018-11-30 一种传感器的自动校准方法和系统

Publications (1)

Publication Number Publication Date
WO2020107445A1 true WO2020107445A1 (zh) 2020-06-04

Family

ID=70852641

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/118686 WO2020107445A1 (zh) 2018-11-30 2018-11-30 一种传感器的自动校准方法和系统

Country Status (1)

Country Link
WO (1) WO2020107445A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4033243A1 (en) * 2021-01-22 2022-07-27 Infineon Technologies AG Gas sensing device and method for determining a calibrated measurement value for a concentration of a target gas

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104236613A (zh) * 2014-07-14 2014-12-24 北京理工大学 公路网传感设备便携式监测诊断及现场检定系统
US20150127289A1 (en) * 2013-11-01 2015-05-07 Honeywell International Inc. Systems and methods for off-line and on-line sensor calibration
CN204461646U (zh) * 2014-12-31 2015-07-08 四川金网通电子科技有限公司 基于云端的传感器数值标定系统
CN106405007A (zh) * 2016-08-30 2017-02-15 河北先河环保科技股份有限公司 气体传感器、颗粒物传感器的新校准方法
CN106441402A (zh) * 2016-08-31 2017-02-22 北京众清科技有限公司 一种传感器组件校准方法、装置及系统
JP2017167999A (ja) * 2016-03-18 2017-09-21 善郎 水野 センサ管理システム
CN108469273A (zh) * 2018-02-27 2018-08-31 济宁中科云天环保科技有限公司 基于机器学习算法的云端数据联调校准方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150127289A1 (en) * 2013-11-01 2015-05-07 Honeywell International Inc. Systems and methods for off-line and on-line sensor calibration
CN104236613A (zh) * 2014-07-14 2014-12-24 北京理工大学 公路网传感设备便携式监测诊断及现场检定系统
CN204461646U (zh) * 2014-12-31 2015-07-08 四川金网通电子科技有限公司 基于云端的传感器数值标定系统
JP2017167999A (ja) * 2016-03-18 2017-09-21 善郎 水野 センサ管理システム
CN106405007A (zh) * 2016-08-30 2017-02-15 河北先河环保科技股份有限公司 气体传感器、颗粒物传感器的新校准方法
CN106441402A (zh) * 2016-08-31 2017-02-22 北京众清科技有限公司 一种传感器组件校准方法、装置及系统
CN108469273A (zh) * 2018-02-27 2018-08-31 济宁中科云天环保科技有限公司 基于机器学习算法的云端数据联调校准方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4033243A1 (en) * 2021-01-22 2022-07-27 Infineon Technologies AG Gas sensing device and method for determining a calibrated measurement value for a concentration of a target gas
US11953481B2 (en) 2021-01-22 2024-04-09 Infineon Technologies Ag Gas sensing device and method for determining a calibrated measurement value for a concentration of a target gas

Similar Documents

Publication Publication Date Title
CN109631973B (zh) 一种传感器的自动校准方法和系统
CN101033989B (zh) 环境监测装置及方法
CN205373919U (zh) 一种基于远程传输的用于工业现场测量校验仪
CN111830350B (zh) 能耗计量方法、装置及电器
CN114047472B (zh) 一种智能电表的计量误差监测系统及其监测方法、装置
CN104111436A (zh) 数字多用表自动检定系统及方法
CN108319664A (zh) 一种大坝及工程安全监测数据粗差识别方法及系统
CN105628218A (zh) 一种光模块芯片温度校准方法
CN114994261A (zh) 环境空气监测的传感器远程校准方法
WO2020107445A1 (zh) 一种传感器的自动校准方法和系统
WO2020056812A1 (zh) 用于评价室内环境质量的环境参数权重确定方法及系统
CN105424074A (zh) 一种传感器校准系统及其控制方法、装置
CN102494743A (zh) 水表小流量自动检测装置及检测方法
CN110702202A (zh) 一种基于云计算平台的电子天平在线计量方法及系统
CN112765787B (zh) 动态环境下考虑性能指标聚类的退化建模与寿命预测方法
CN103090899A (zh) 传感器检测装置及检测方法
CN116879223A (zh) 近红外水分仪标定方法、装置、设备及可读存储介质
WO2020021344A1 (zh) 环境传感器协同校准方法
CN104991222B (zh) 计量自动化终端质量评价系统
CN202511851U (zh) 一种水表大流量自动检测装置
CN206146591U (zh) 基于压力变送器性能测试系统
CN110705186B (zh) 通过rbf粒子群优化算法即时在线仪表校验和诊断方法
CN114993501A (zh) 一种基于边缘计算的ntc温度传感器校准方法
CN110750756B (zh) 通过最优支持向量机算法即时在线仪表校验和诊断方法
CN107748342A (zh) 库仑计校准方法、移动终端及计算机存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18941402

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18941402

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 03.02.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 18941402

Country of ref document: EP

Kind code of ref document: A1