Eight-station parallel resistance calibration method and device based on high-precision resistance network

Abstract

The application discloses an eight-station parallel resistance calibration method and device based on a high-precision resistance network, and relates to the technical field of intelligent sensors.The application comprises the following steps: acquiring resistance range configuration information of each handheld digital multimeter, determining standard resistance values corresponding to each range, and placing the multimeters in eight calibration stations; synchronously outputting the standard resistance values corresponding to each station to each calibration station through a high-precision resistance network; reading resistance measurement values displayed by each multimeter through an image sensor; determining error correction coefficients corresponding to each station through data expansion and comprehensive evaluation according to historical calibration data, correcting the resistance measurement values of each station, comparing the corrected measurement values with the corresponding standard resistance values, and determining resistance calibration results of each multimeter according to the comparison results.The application solves the technical problems that the existing resistance calibration method for handheld digital multimeters relies on manual operation, which leads to low efficiency, and each station cannot be synchronously and parallelly calibrated, which leads to poor accuracy of the calibration results.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensor technology, and specifically to an eight-position parallel resistance calibration method and apparatus based on a high-precision resistance network. Background Technology

[0002] Handheld digital multimeters are widely used portable measuring instruments in the field of electronic measurement. The accuracy of their resistance measurement function directly affects the reliability of electrical equipment testing, circuit repair, and product quality inspection. Therefore, handheld digital multimeters need to be calibrated regularly in resistance mode to ensure that their measurement accuracy meets technical specifications.

[0003] Currently, resistance calibration of handheld digital multimeters mainly relies on a single-unit, successive calibration method. Operators must connect each meter to a standard resistance source one by one, completing calibration for each range before moving on to the next. This process is time-consuming, inefficient, and unsuitable for large-scale calibration needs. Furthermore, reading meter measurements, calculating errors, and determining pass / fail all require manual intervention, which is cumbersome and prone to errors. The differences in range configurations between different meter models necessitate manual consultation of technical manuals, further increasing the complexity of calibration preparation. Therefore, existing calibration methods are no longer adequate for current requirements on calibration efficiency and automation. Summary of the Invention

[0004] This invention provides an eight-station parallel resistance calibration method and apparatus based on a high-precision resistance network. This invention solves the technical problems of existing handheld digital multimeter resistance calibration methods, which rely on manual operation, resulting in low efficiency and poor accuracy of calibration results due to the inability to perform synchronous parallel calibration at each station.

[0005] In view of the above problems, the present invention provides an eight-station parallel resistance calibration method based on a high-precision resistance network, the method comprising: Obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range according to the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration stations respectively. The standard resistance value corresponding to each of the eight calibration stations is synchronously output to the eight calibration stations through a high-precision resistor network. The resistance measurement values ​​displayed on each handheld digital multimeter are read by the image sensor corresponding to each calibration station; Based on the historical calibration data of each calibration station, the error correction coefficient corresponding to each station is determined through data expansion and comprehensive evaluation. The resistance measurement value of each station is then corrected according to the error correction coefficient to obtain the corrected measurement value of each calibration station. The calibrated measured values ​​at each calibration station are compared with the corresponding standard resistance values, and the resistance calibration results of each handheld digital multimeter are determined based on the comparison results.

[0006] This invention also provides an eight-station parallel resistance calibration device based on a high-precision resistance network, comprising: The range configuration and workstation deployment module is used to obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range according to the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration workstations respectively. A standard resistor synchronous output module is used to synchronously output the standard resistor value corresponding to each of the eight calibration stations through a high-precision resistor network. The measurement value image recognition and reading module is used to read the resistance measurement values ​​displayed by each handheld digital multimeter through the image sensor corresponding to each calibration station; The error correction coefficient determination and correction module is used to determine the error correction coefficient corresponding to each calibration station based on the historical calibration data of each calibration station through data expansion and comprehensive evaluation, and to correct the resistance measurement value of each station according to the error correction coefficient to obtain the corrected measurement value of each calibration station. The calibration result determination and report generation module is used to compare the corrected measured values ​​of each calibration station with the corresponding standard resistance values, and determine the resistance calibration results of each handheld digital multimeter based on the comparison results.

[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention enables parallel calibration of eight instruments by combining eight calibration stations with eight independent output channels of a high-precision resistor network, significantly improving batch calibration efficiency. A pre-stored model-range configuration mapping table automates the acquisition of range configuration information and standardizes the determination of standard resistance values, avoiding subjective biases from manual review and point selection. A field-programmable gate array (FPGA) generates a synchronous trigger signal, ensuring all eight channels output standard resistance values ​​simultaneously, eliminating inconsistencies in measurement start times at each station caused by time-sharing output. Automatic image acquisition via an image sensor, combined with tilt correction and optical character recognition, replaces manual reading, improving reading efficiency and accuracy. Data augmentation using a conditional generative adversarial network (GAN) combined with comprehensive evaluation and iterative supplementation strategies, along with a station embedding layer mapping discrete station numbers to dense embedding vectors to fuse multi-source features for error prediction, achieves precise determination of error correction coefficients. Automatic acquisition of tolerance thresholds for comparison and judgment, combined with a retesting mechanism to avoid single-time misjudgments, and automatic generation of calibration reports form a complete automated closed loop from data acquisition to report output, effectively improving the reliability of calibration results and work efficiency.

[0008] In summary, this invention solves the technical problems of low efficiency and poor accuracy of calibration results caused by the reliance on manual operation in existing handheld digital multimeter resistance calibration methods, as well as the inability of each workstation to perform synchronous parallel calibration. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating the eight-position parallel resistance calibration method based on a high-precision resistance network provided in an embodiment of the present invention. Figure 2 , Figure 3 This is a schematic diagram illustrating the structure of the eight stations in the eight-station parallel resistance calibration method based on a high-precision resistance network provided in this embodiment of the invention. Figure 4 This is a schematic diagram of the structure of an eight-station parallel resistance calibration device based on a high-precision resistance network provided in an embodiment of the present invention. The components represented by each number in the attached diagram are explained below: The module includes: a range configuration and workstation deployment module 11, a standard resistance synchronous output module 12, a measured value image recognition and reading module 13, an error correction coefficient determination and correction module 14, and a calibration result judgment and report generation module 15. Detailed Implementation

[0011] This invention provides an eight-station parallel resistance calibration method and apparatus based on a high-precision resistance network, which specifically solves the technical problems of low efficiency caused by manual operation in existing handheld digital multimeter resistance calibration methods and poor accuracy of calibration results due to the inability of each station to be calibrated synchronously and in parallel.

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0013] Example 1, as Figure 1 As shown, this invention provides an eight-station parallel resistor calibration method based on a high-precision resistor network, the method comprising: S100: Obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range according to the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration stations respectively.

[0014] The resistance ranges and measurement ranges of different models of handheld digital multimeters vary significantly. Current calibration methods require operators to manually consult the technical manuals of each model to obtain range configuration information, then manually select calibration points within each range and determine the standard resistance value, finally placing the instrument in the corresponding station and recording the correspondence. This manual operation mode is not only extremely cumbersome and time-consuming in preparation, severely limiting the overall efficiency of eight-station parallel calibration, but also prone to errors such as range identification errors, standard resistance value selection deviations, or chaotic station correspondence records during manual consultation and point selection, leading to deviations in calibration results or even damage to the instrument. Especially in the eight-station parallel calibration scenario, the eight stations may simultaneously correspond to multiple different models of instruments, with differentiated range configuration information for each station, further drastically increasing the complexity and error probability of manual processing. Therefore, how to achieve automated acquisition of resistance range configuration information, standardized and adaptive determination of standard resistance values ​​for each range, and efficient and accurate allocation between the eight stations and the instrument to be calibrated has become a pressing technical problem to be solved.

[0015] Step S100 in the method provided in this embodiment of the invention includes: Obtain the model information of each handheld digital multimeter to be calibrated, and query the pre-stored model-range configuration mapping table according to the model information to obtain the resistance range configuration information of each handheld digital multimeter. The resistance range configuration information includes the resistance range supported by each handheld digital multimeter and the range of each resistance range. Based on the resistance range configuration information, a calibration point is selected within the range of each resistance range to determine the standard resistance value corresponding to each range. Place the eight handheld digital multimeters to be calibrated into the eight calibration stations respectively.

[0016] In this embodiment of the invention, the model information of each handheld digital multimeter to be calibrated is first obtained. Specifically, the panel image of the handheld digital multimeter can be acquired by the image sensor corresponding to each calibration station, and the model information can be extracted after optical character recognition. The operator manually inputs the model information of each handheld digital multimeter through the human-machine interface of the host computer. The model identification code affixed to the surface of the handheld digital multimeter is read by a barcode or QR code scanning device. In this embodiment of the invention, the method of acquiring panel images by image sensors and extracting model information through optical character recognition is used to achieve automated acquisition of model information.

[0017] Furthermore, after obtaining the model information of each handheld digital multimeter, the pre-stored model-range configuration mapping table is consulted to obtain the resistance range configuration information of each handheld digital multimeter. The model-range configuration mapping table is pre-stored in the calibration system's storage module. This mapping table records the resistance range configuration information of various commonly used handheld digital multimeter models. The resistance range configuration information includes the resistance ranges supported by each handheld digital multimeter and the range of each resistance range. When a new model is not recorded in the model-range configuration mapping table, a prompt message is issued via the host computer, allowing the operator to manually enter the range configuration information for that model, and the model-range configuration mapping table is automatically updated.

[0018] Furthermore, based on the resistance range configuration information, a calibration point is selected within the range of each resistance range to determine the standard resistance value corresponding to each range.

[0019] Specifically, for each resistance range, calibration points are selected within its range according to preset rules. These preset rules include, but are not limited to, the following: selecting the midpoint of the range as the calibration point, for example, selecting 100Ω as the standard resistance value for a range of 0Ω to 200Ω; selecting the upper limit of the range as the calibration point, for example, selecting 200Ω as the standard resistance value for a range of 0Ω to 200Ω; or selecting one or more points within the range that are distributed exponentially or logarithmically as calibration points. In this embodiment of the invention, the upper limit of the range for each range is selected as the standard resistance value because the upper limit of the range can generally more comprehensively reflect the measurement accuracy of the handheld digital multimeter at that range.

[0020] For example, suppose a handheld digital multimeter to be calibrated is model A, and its resistance range configuration information includes 200Ω range, 2kΩ range, 20kΩ range, 200kΩ range, 2MΩ range, and 20MΩ range. The 200Ω range corresponds to a range of 0Ω to 200Ω, the 2kΩ range corresponds to a range of 0Ω to 2kΩ, the 20kΩ range corresponds to a range of 0Ω to 20kΩ, the 200kΩ range corresponds to a range of 0Ω to 200kΩ, the 2MΩ range corresponds to a range of 0Ω to 2MΩ, and the 20MΩ range corresponds to a range of 0Ω to 20MΩ. According to the rule of selecting the upper limit of the range as the calibration point, the standard resistance values ​​corresponding to each range are as follows: 200Ω range corresponds to 200Ω, 2kΩ range corresponds to 2kΩ, 20kΩ range corresponds to 20kΩ, 200kΩ range corresponds to 200kΩ, 2MΩ range corresponds to 2MΩ, and 20MΩ range corresponds to 20MΩ.

[0021] It should be noted that the above method of selecting the upper limit of the measurement range as the calibration point is only an illustrative example. In practical applications, one or more calibration points can be selected within the measurement range of each measurement range according to the requirements of specific calibration specifications. This embodiment does not limit this.

[0022] Furthermore, after determining the standard resistance values ​​corresponding to each range of each handheld digital multimeter, the eight handheld digital multimeters to be calibrated were placed into the eight calibration stations respectively.

[0023] like Figure 2 and Figure 3As shown, specifically, eight calibration stations are set up on the calibration bench. Each calibration station includes a fixture for securing the handheld digital multimeter, test probes or terminals for connecting to the standard resistor output channel of the high-precision resistor network, and an image sensor for acquiring images from the display panel. The operator places the eight handheld digital multimeters to be calibrated into the eight calibration stations sequentially and secures them using the fixtures. During placement, it is essential to ensure a reliable electrical connection between the test input terminals of each handheld digital multimeter and the test probes or terminals of the corresponding calibration station, while simultaneously ensuring that the display panel of each handheld digital multimeter faces the corresponding image sensor so that the image sensor can clearly acquire the display panel image.

[0024] After placing eight handheld digital multimeters into the eight calibration stations, the control module of the calibration system records the correspondence between each station number and the model of the handheld digital multimeter placed in that station, and stores this correspondence in the calibration system's storage module for later retrieval during the calibration process. Simultaneously, based on the model and range configuration information of the handheld digital multimeters corresponding to each station, the control module determines the standard resistance value corresponding to each range at each station and generates a calibration task sequence for each station.

[0025] For example, assuming that a handheld digital multimeter of model A is installed at station 1 and a handheld digital multimeter of model B is installed at station 2, the control module records the correspondence between station 1 and model A, and station 2 and model B. When performing calibration at the 200Ω range, the control module determines that the standard resistance value for station 1 is 200Ω, and station 2 has no calibration task at the 200Ω range, thus achieving differentiated configuration of calibration tasks for each station.

[0026] This invention utilizes a pre-stored model-range configuration mapping table to automatically retrieve complete resistance range configuration information based on model information. This completely eliminates the tedious manual process of consulting technical manuals one by one, significantly shortening calibration preparation time and preventing range identification errors caused by manual review and data entry. Simultaneously, by automatically selecting calibration points within the range of each obtained range according to preset rules to determine the standard resistance value, the standardization and consistency of standard resistance value selection are ensured, avoiding uncontrollable variables introduced by subjective human experience.

[0027] Furthermore, by placing eight instruments into eight calibration stations and automatically recording the correspondence between station numbers and instrument models, a precise station configuration basis is provided for the high-precision resistor network to synchronously output differentiated standard resistance values ​​matching each instrument model to each station. This ensures that each station does not interfere with others and receives the correct standard resistance excitation during parallel calibration. In summary, this solution comprehensively improves the intelligence and automation level of the calibration method, enabling it to flexibly adapt to the mixed-line calibration needs of various instrument models and providing reliable support for large-scale, high-efficiency resistance calibration work.

[0028] S200: The standard resistance value corresponding to each of the eight calibration stations is synchronously output to the eight calibration stations through a high-precision resistor network.

[0029] In existing handheld digital multimeter resistance calibration methods, high-precision resistance networks typically output standard resistance values ​​to each calibration station using a time-sharing or sequential output method. This means each channel outputs its own standard resistance value at different times. Because the response time of resistance measurements varies among handheld digital multimeters, and instruments at different stations may be in different stable states, the time-sharing output method leads to inconsistent timing of the standard resistance excitation received at each station, resulting in asynchronous measurement start times for the instruments at each station. This time asynchrony introduces additional measurement uncertainty in an eight-station parallel calibration scenario, as instruments at each station may perform measurements under different conditions such as fluctuating ambient temperature and changes in circuit thermal equilibrium, leading to decreased comparability of calibration results.

[0030] Meanwhile, if the output resistance values ​​of each channel do not match the standard resistance values ​​actually required by each station, the output resistance values ​​may exceed the instrument's range, posing a risk of damaging the instrument to be calibrated. Furthermore, in traditional methods, the network topology is unclear, and the channel configuration lacks a unified synchronous control mechanism, making it difficult to ensure that all eight channels complete resistance value switching and output simultaneously. Therefore, how to achieve independent resistance value configuration and synchronous output control for the eight output channels of a high-precision resistance network, enabling each station to obtain its matching standard resistance value at the same time, has become a pressing technical problem that needs to be solved.

[0031] Step S200 in the method provided in this embodiment of the invention includes: Obtain the network topology of the high-precision resistor network, which includes eight independent standard resistor output channels. Configure the output resistance value of each standard resistor output channel according to the standard resistor value corresponding to each workstation; A synchronous trigger signal is generated by a field-programmable gate array and simultaneously sent to the eight standard resistor output channels of the high-precision resistor network, so that each standard resistor output channel outputs the configured standard resistor value to the corresponding calibration station at the same time.

[0032] In this embodiment of the invention, the network topology of the high-precision resistor network is first obtained. The high-precision resistor network is an eight-channel programmable standard resistor output module. Its network topology includes eight independent standard resistor output channels, each of which can independently programmatically adjust its output resistance value. The output channels are electrically isolated from each other, and the resistance configuration and output status of any channel do not affect the normal operation of other channels, thereby ensuring that each of the eight calibration stations receives independent and accurate standard resistor excitation.

[0033] Specifically, the high-precision resistor network uses a combination of high-precision resistors and relay arrays to achieve programmable resistance adjustment. Each output channel contains multiple sets of precision resistors and corresponding control relays. By controlling the on / off states of different relay combinations, multiple sets of precision resistors can be connected in series, parallel, or series-parallel combinations to output the target resistance value. The resistance adjustment resolution, accuracy, and range of each output channel are designed according to actual calibration requirements, typically needing to meet the output accuracy requirements of the standard resistance values ​​for each range. The network topology information of the high-precision resistor network is pre-stored in the calibration system's storage module. The calibration system's control module can read this stored information to obtain the topology information of each output channel, including its number, address, control method, and electrical parameters.

[0034] Furthermore, after obtaining the network topology of the high-precision resistor network, the output resistance value of each standard resistor output channel is configured according to the standard resistance value corresponding to each station.

[0035] Specifically, the control module of the calibration system determines the standard resistance values ​​corresponding to each calibration station at each range and generates resistance configuration instructions for each standard resistance output channel. These instructions include the target channel number and the target output resistance value. The control module sends these configuration instructions to the high-precision resistor network via a communication bus. Upon receiving the instructions, the high-precision resistor network's internal channel control unit parses them and calculates the corresponding relay combination control logic based on the target output resistance value. This logic then drives the corresponding relays to switch the output resistance value of the corresponding channel to the target value.

[0036] For example, suppose the control module of the calibration system determines the standard resistance values ​​for each station as follows: station 1 has a standard resistance value of 200Ω, station 2 has a standard resistance value of 400Ω, station 3 has a standard resistance value of 2kΩ, station 4 has a standard resistance value of 4kΩ, station 5 has a standard resistance value of 20kΩ, station 6 has a standard resistance value of 40kΩ, station 7 has a standard resistance value of 200kΩ, and station 8 has a standard resistance value of 400kΩ. The control module then generates eight resistance configuration instructions: channel 1 is configured as 200Ω, channel 2 as 400Ω, channel 3 as 2kΩ, channel 4 as 4kΩ, channel 5 as 20kΩ, channel 6 as 40kΩ, channel 7 as 200kΩ, and channel 8 as 400kΩ, and sends these instructions to the high-precision resistor network via the communication bus. Upon receiving the instructions, the high-precision resistor network drives the relay array of each channel to switch the output resistance value of each channel to the corresponding target value.

[0037] Furthermore, after the output resistance value of each standard resistor output channel is configured, a synchronous trigger signal is generated by a field-programmable gate array and simultaneously sent to the eight standard resistor output channels of the high-precision resistor network, so that each standard resistor output channel outputs the configured standard resistance value to the corresponding calibration station at the same time.

[0038] Specifically, the control module of the calibration system includes a field-programmable gate array (FPGA) chip, which serves as the source of the synchronization trigger signal. The FPGA chip internally contains synchronization trigger signal generation logic, capable of simultaneously generating synchronization trigger pulse signals from multiple output pins with extremely high timing accuracy upon receiving a synchronization output command from the control module. These synchronization trigger pulse signals are connected via signal transmission cables to the trigger input terminals of eight standard resistor output channels of a high-precision resistor network.

[0039] When the field-programmable gate array (FPGA) generates a synchronous trigger signal, this signal simultaneously reaches the trigger input terminals of all eight standard resistor output channels. Upon receiving the trigger signal, each standard resistor output channel locks its pre-configured relay array state and turns on to its output terminal, thereby simultaneously outputting the corresponding standard resistance value to its respective connected calibration station. Since the time deviation between the output channels of the synchronous trigger signal generated by the FPGA is on the nanosecond level, far less than the resistance measurement response time of a handheld digital multimeter, it can be considered that all eight stations receive standard resistor excitation simultaneously.

[0040] It should be noted that the synchronization trigger signal generated by the field-programmable gate array (FPGA) can be either level-triggered or edge-triggered. In this embodiment of the invention, a rising edge triggering method is used, meaning that the FPGA outputs a synchronous rising edge pulse signal, and each standard resistor output channel immediately executes its output action upon detecting the rising edge. The logic for generating the synchronization trigger signal of the FPGA can be programmed using a hardware description language, and its synchronization accuracy is guaranteed by the high-speed clock and parallel output structure within the FPGA.

[0041] Furthermore, to ensure that the output resistance of each channel is stable during synchronous output, the control module reserves a delay sufficient for the relays to operate and the resistance to stabilize after sending the resistance configuration command. The synchronous output command is sent to the field-programmable gate array only after this delay. After synchronous output is complete, the handheld digital multimeters at each calibration station simultaneously begin receiving standard resistance excitation and performing resistance measurements. The displayed resistance values ​​are synchronously acquired and read by the image sensors at each station, thus initiating the subsequent measurement value reading and error correction process.

[0042] By acquiring a high-precision resistor network topology containing eight independent standard resistor output channels, and configuring the output resistance value of each channel according to the standard resistance value corresponding to each station, it is ensured that the standard resistance value obtained at each calibration station is accurately matched with the range of the instrument to be calibrated in that station, thus avoiding the risk of instrument damage due to mismatched standard resistance values ​​from the source.

[0043] Simultaneously, a synchronous trigger signal is generated by a field-programmable gate array (FPGA) and sent to eight standard resistor output channels at the same time. This ensures that each channel outputs its configured standard resistor value to its corresponding calibration station at the same moment, completely eliminating the problem of inconsistent measurement start times at each station caused by time-division multiplexing. This guarantees that all eight stations receive the standard resistor excitation at exactly the same time, effectively eliminating the additional uncertainty introduced by time asynchrony into the calibration results. Leveraging its parallel processing capabilities and programmability, the FPGA can generate synchronous trigger signals with extremely high timing accuracy. The switching and settling times of the output resistance values ​​of each channel are highly consistent, greatly improving the time synchronization and overall calibration accuracy of the eight-station parallel calibration. This provides a solid time reference for the accurate comparison of subsequent measurements from each station and the reliability determination of the calibration results.

[0044] S300: Read the resistance measurement value displayed by each handheld digital multimeter through the image sensor corresponding to each calibration station.

[0045] In existing handheld digital multimeter resistance calibration methods, reading resistance measurements primarily relies on manual visual readings. This is not only inefficient and labor-intensive, but also prone to errors due to prolonged readings, especially in scenarios with eight parallel workstations where simultaneous manual reading of multiple instruments is difficult. While some solutions attempt to use image acquisition followed by optical character recognition, the placement angles of the instruments at each workstation are often inconsistent, resulting in varying degrees of tilt and perspective distortion on the display panel image. This causes the digital characters to be stretched or distorted, severely interfering with the accuracy of optical character recognition and potentially leading to misidentification and incorrect measurement values, thus affecting the reliability of subsequent calibration results. Therefore, how to automate the reading of resistance measurements and effectively eliminate the adverse effects of image tilt and distortion on character recognition has become a pressing technical problem to be solved.

[0046] Step S300 in the method provided in this embodiment of the invention includes: The display panel image of the handheld digital multimeter is acquired by the image sensor corresponding to each calibration station; The display panel image is subjected to tilt correction processing to obtain a corrected display panel image; The corrected display panel image is input into a pre-trained optical character recognition model, and the output is the resistance measurement value displayed by the handheld digital multimeter.

[0047] In this embodiment of the invention, firstly, the display panel image of the handheld digital multimeter is acquired by the image sensor corresponding to each calibration station.

[0048] Specifically, an image sensor is installed above or in front of each calibration station. The image sensor is preferably an industrial-grade digital camera with autofocus and auto-exposure capabilities, enabling it to capture clear images of the display panel under various ambient lighting conditions. The installation position and shooting angle of each image sensor are pre-adjusted to ensure that its optical axis is approximately perpendicular to the display panel plane of the handheld digital multimeter in the corresponding calibration station, thereby reducing distortion caused by the initial shooting angle. In an eight-station parallel calibration scenario, each image sensor corresponds to one of the eight calibration stations and can independently complete the acquisition of images from the instrument display panel at its respective station.

[0049] After the high-precision resistor network completes the synchronous output of standard resistance values, the handheld digital multimeters at each calibration station begin resistance measurement and display the results. Once the measurement results stabilize, the calibration system's control module simultaneously sends image acquisition trigger commands to each image sensor via the image acquisition control bus. Upon receiving the trigger command, each image sensor simultaneously captures images of its corresponding handheld digital multimeter's display panel, resulting in eight display panel images. These images are color or grayscale digital images encompassing the handheld digital multimeter's LCD or digital tube display area. Each image sensor transmits the acquired display panel images to the calibration system's image processing module via a data transmission interface for subsequent processing.

[0050] To ensure the quality of image acquisition, the control module can detect and adjust the ambient brightness through the automatic exposure control function of the image sensor before sending the image acquisition trigger command, ensuring that the brightness and contrast of the display panel image are within a suitable range for optical character recognition.

[0051] Furthermore, after acquiring the display panel images of each workstation, the display panel images are subjected to tilt correction processing to obtain the corrected display panel images.

[0052] Specifically, when handheld digital multimeters are placed in the calibration station, the placement angle of the instrument may vary slightly due to positioning errors of the meter fixture, dimensional tolerances of the instrument housing, or different placement methods by the operator. This causes the display panel plane to tilt at a certain angle relative to the optical axis of the image sensor, resulting in perspective distortion of the display area and digital characters in the acquired display panel image. The purpose of tilt correction processing is to eliminate this perspective distortion and correct the tilted display panel image into a positive, distortion-free image.

[0053] In this embodiment of the invention, a minimum bounding rectangle correction method based on contour extraction is used to correct the tilt of the display panel image. The specific steps are as follows: First, the acquired raw display panel image is preprocessed. The raw image is converted into a grayscale image, and then the grayscale image is binarized to make the display panel area and the background area significantly distinguishable in grayscale, resulting in a binarized image.

[0054] Secondly, contour detection is performed on the binarized image to extract all contours. The area of ​​each contour is calculated, and the contour with the largest area is selected as the target contour for the display panel. Since the display panel usually occupies the largest continuous area in the image, area filtering can effectively eliminate interfering contours in the background.

[0055] Next, obtain the minimum bounding rectangle of the target outline. The minimum bounding rectangle can completely enclose the target outline with the smallest rectangular area, and the rotation angle information of this rectangle is also output. This rotation angle is the tilt angle of the display panel relative to the image coordinate system.

[0056] Finally, using the center point of the smallest bounding rectangle as the rotation center and the negative value of the rotation angle of this rectangle as the rotation angle, an affine transformation is performed on the original display panel image to rotate the tilted display panel to a positive orientation, resulting in the corrected display panel image. This method requires only one rotation angle to complete the correction, resulting in less computation and simpler implementation. Furthermore, it provides sufficient correction accuracy for slight tilts of planar targets such as display panels. In the corrected display panel image, the display area presents a positive, untilted rectangular shape, and the digits within it are restored to their normal orientation, providing good input image quality for subsequent optical character recognition.

[0057] Furthermore, after obtaining the corrected display panel image, the corrected display panel image is input into a pre-trained optical character recognition model, and the output is the resistance measurement value displayed by the handheld digital multimeter.

[0058] In this embodiment of the invention, the optical character recognition model preferably adopts an existing open-source OCR model, eliminating the need for training from scratch and labeling a large number of training samples. Currently, there are various mature open-source OCR solutions. These models have been pre-trained on large-scale public datasets containing multiple fonts, languages, and shooting conditions, possessing good character recognition accuracy and cross-scene generalization capabilities, and can be directly used for practical deployment. In practical applications, a suitable model can be selected for integration from existing open-source OCR models based on the calibration system's hardware configuration, recognition accuracy requirements, and inference speed requirements. Examples include various open-source OCR models based on CNN+RNN+CTC architecture or Transformer architecture.

[0059] The optical character recognition model requires no additional training before use; it can directly load the pre-trained weights provided by the open-source model for inference and recognition. The calibration system's image processing module inputs the calibrated display panel image into the model via the programming interface provided by the open-source OCR model. After forward inference calculations, the model outputs the recognized character sequence, which represents the resistance measurement value displayed on the handheld digital multimeter. For example, if the display panel shows "200.0Ω", the optical character recognition model outputs the character sequence "200.0" along with the unit symbol. The image processing module further extracts the numerical portion "200.0" from the output as the resistance measurement value.

[0060] For different models of handheld digital multimeters, the font, size, decimal point position, and representation of characters on the display panel may vary. Open-source OCR models are typically pre-trained on large-scale datasets containing multiple fonts, languages, and shooting conditions, exhibiting good generalization ability for different display styles and various real-world shooting scenarios, and can adapt to the character recognition needs of various meter models.

[0061] Furthermore, to ensure the reliability of resistance measurement readings, after the optical character recognition model outputs the recognition result, the image processing module can also verify the result. For example, it checks whether the result contains valid numeric characters, decimal points, and unit symbols, and whether the number of digits is within a reasonable range. If the verification passes, the identified resistance measurement value is output and stored in the calibration system's data storage module for subsequent error correction and calibration result determination. If the verification fails, a re-image acquisition and re-recognition process is triggered until the reading is successful or the preset retry limit is reached.

[0062] This invention automatically acquires images of the display panel using image sensors at each workstation, replacing manual readings and significantly improving reading efficiency and automation. By performing tilt correction on the acquired images, perspective distortion caused by differences in instrument placement angles is effectively eliminated, restoring characters to a positive, distortion-free form and significantly reducing the probability of character misrecognition. Inputting the corrected images into a pre-trained optical character recognition model accurately outputs the resistance measurement values ​​for each workstation. These steps ensure the accuracy, consistency, and efficiency of reading measurement values ​​at each workstation in an eight-workstation parallel scenario, providing a reliable data foundation for subsequent error correction and calibration determination.

[0063] S400: Based on the historical calibration data of each calibration station, the error correction coefficient corresponding to each station is determined through data expansion and comprehensive evaluation, and the resistance measurement value of each station is corrected according to the error correction coefficient to obtain the corrected measurement value of each calibration station.

[0064] In the resistance calibration process of handheld digital multimeters, historical calibration data often suffers from uneven sample distribution. The number of samples for certain workstation-range combinations is severely insufficient, or the samples are unevenly distributed within the standard resistance value range. This leads to a significant decrease in the prediction accuracy of error prediction models trained on historical data in sparse sample regions, making it impossible to accurately predict measurement errors within the current calibration cycle, thus affecting the accuracy of error correction coefficients. Furthermore, the workstation number, as a discrete categorical feature, has a complex nonlinear relationship with error that is difficult for traditional regression models to effectively capture. Therefore, how to obtain sufficient and evenly distributed training samples from limited and unevenly distributed historical calibration data through data augmentation, and how to construct an error prediction model that effectively integrates multi-source heterogeneous features such as workstation number, standard resistance value, and ambient temperature to accurately determine the error correction coefficients for each workstation, has become a pressing technical problem to be solved.

[0065] Step S400 in the method provided in this embodiment of the invention includes: Acquire historical calibration data accumulated by each calibration station in previous calibrations. The historical calibration data includes historical standard resistance value, historical resistance measurement value, historical ambient temperature, historical station number, historical range setting, and corresponding historical measurement error. The historical measurement error is the difference between the historical resistance measurement value and the historical standard resistance value. Based on the historical standard resistance values, historical ambient temperatures, historical workstation numbers, and historical measurement ranges in the historical calibration data, expanded calibration data is generated through a pre-trained data generation model. The expanded calibration data includes pseudo standard resistance values, pseudo ambient temperatures, pseudo workstation numbers, pseudo measurement ranges, and pseudo measurement errors. A comprehensive evaluation is performed on the expanded calibration data. Based on the comprehensive evaluation results, the workstation-range combination that needs to be supplemented and the input feature interval corresponding to the workstation-range combination are determined. The input feature interval is then used as a supplementary condition and input into the data generation model to iteratively generate supplementary calibration data until the preset data sufficiency requirements are met, thereby obtaining the target expanded calibration dataset. Using the pseudo-standard resistance value, pseudo-ambient temperature, and pseudo-station number in the target expanded calibration dataset as input features, and the pseudo-measurement error in the target expanded calibration dataset as supervision label, an error prediction model is trained. Input the current ambient temperature, the standard resistance value of each calibration station, and the station number into the error prediction model, and output the predicted error value for the current calibration cycle. The negative of the predicted error value is used as the error correction coefficient for each calibration station, and the corresponding error correction coefficient is added to the resistance measurement value of each calibration station to obtain the corrected measurement value of each calibration station.

[0066] In this embodiment of the invention, firstly, historical calibration data accumulated by each calibration station in previous calibrations is acquired. This historical calibration data is stored in the historical database of the calibration system. After each calibration is completed, the system automatically stores the relevant data for that calibration into the database, forming a continuously accumulating historical calibration dataset.

[0067] Historical calibration data includes historical standard resistance values, historical resistance measurements, historical ambient temperatures, historical workstation numbers, historical range settings, and corresponding historical measurement errors. Specifically, the historical standard resistance value is the standard resistance value output by the high-precision resistance network to each workstation during each calibration; the historical resistance measurement value is the resistance measurement value displayed on the handheld digital multimeter at each workstation, read by the image sensor and optical character recognition model during each calibration; the historical ambient temperature is the ambient temperature value collected and recorded by the temperature sensor installed on the calibration bench during each calibration; the historical workstation number is the unique identifier for each calibration workstation, ranging from 1 to 8; the historical range setting is the resistance range setting of the handheld digital multimeter during each calibration, such as 200Ω or 2kΩ; and the historical measurement error is the difference between the historical resistance measurement value and the historical standard resistance value, i.e., historical measurement error = historical resistance measurement value - historical standard resistance value.

[0068] Furthermore, based on historical standard resistance values, historical ambient temperatures, historical workstation numbers, and historical measurement ranges from the historical calibration data, expanded calibration data is generated through a pre-trained data generation model. The expanded calibration data includes pseudo-standard resistance values, pseudo-ambient temperatures, pseudo-workstation numbers, pseudo-measurement ranges, and pseudo-measurement errors.

[0069] In the actual generation process, historical standard resistance values, historical ambient temperatures, historical workstation numbers, and historical measurement ranges from historical calibration data are used as input data to generate the model. Based on the distribution characteristics of these conditional variables, the model generates corresponding pseudo-standard resistance values, pseudo-ambient temperatures, pseudo-workstation numbers, pseudo-measurement ranges, and pseudo-measurement errors. The pseudo-measurement error is a simulated error value generated by the data generation model under given conditional variables, based on the error distribution patterns learned from historical data. Through this data generation model, a large amount of expanded calibration data can be generated from limited historical calibration data without increasing the number of actual calibrations, effectively expanding the scale of the training sample.

[0070] The pre-training process of the data generation model includes: Multiple sets of historical calibration data samples are obtained. Each set of historical calibration data samples includes historical standard resistance value samples, historical ambient temperature samples, historical workstation number samples, historical range samples, and corresponding historical measurement error samples. Based on conditional generative adversarial networks, an initial data generation model is constructed, which includes a generator and a discriminator. The generator generates a pseudo-expansion sample based on the historical standard resistance value sample, the historical ambient temperature sample, the historical workstation number sample, and the historical range sample. The pseudo-expansion sample includes a pseudo standard resistance value, a pseudo ambient temperature, a pseudo workstation number, a pseudo range, and a pseudo measurement error. The discriminator receives the pseudo-expansion sample and the real expansion sample composed of the historical calibration data sample, and determines the authenticity of the input sample. The generator and the discriminator alternately update the network parameters until the discriminator can no longer distinguish between the pseudo-expanded samples and the real expanded samples, at which point training stops, and the generator at the point where training stops is used as the pre-trained data generation model.

[0071] In this embodiment of the invention, firstly, multiple sets of historical calibration data samples are obtained from historical calibration data. Each set of historical calibration data samples includes historical standard resistance value samples, historical ambient temperature samples, historical workstation number samples, historical range samples, and corresponding historical measurement error samples. These samples originate from various calibration records stored in the historical database of the calibration system.

[0072] Secondly, based on conditional generative adversarial networks (GANs), an initial data generation model is constructed, comprising a generator and a discriminator. The generator generates pseudo-expansion samples based on historical standard resistance value samples, historical ambient temperature samples, historical workstation number samples, and historical measurement range samples. These pseudo-expansion samples contain pseudo-standard resistance values, pseudo-ambient temperatures, pseudo-workstation numbers, pseudo-measurement ranges, and pseudo-measurement errors. Specifically, the generator uses historical standard resistance values, historical ambient temperatures, historical workstation numbers, and historical measurement ranges as conditional inputs. Internally, it transforms these conditions through a multi-layer fully connected network, outputting the corresponding pseudo-measurement error, which, together with the input conditional variables, constitutes a complete pseudo-expansion sample.

[0073] The discriminator receives fake augmented samples and real augmented samples composed of historical calibration data samples, and determines the authenticity of the input samples. Specifically, the discriminator takes the complete sample as input, extracts features through a multi-layer fully connected network, and finally outputs a scalar value between 0 and 1, representing the probability that the input sample is a real sample. The closer the value is to 1, the more real the discriminator considers the sample; the closer the value is to 0, the more fake the discriminator considers the sample.

[0074] The generator and discriminator alternately update the network parameters. In each training round, the generator's parameters are first fixed, and real augmented samples and pseudo-augmented samples are input into the discriminator respectively. The discriminator's loss is calculated, and its parameters are updated to make the discriminator distinguish between real and pseudo samples as accurately as possible. Then, the discriminator's parameters are fixed again, and a conditional variable is input into the generator to generate new pseudo-augmented samples. These pseudo-augmented samples are then input into the discriminator, and the generator's loss is calculated, and its parameters are updated to make the generator generate pseudo samples that can fool the discriminator as much as possible. Through this adversarial training method, the generator and discriminator continuously compete and improve each other.

[0075] Repeat the alternating update process described above until the discriminator can no longer distinguish between pseudo-expanded samples and real expanded samples, that is, when the discriminator's output probability for both pseudo and real samples is close to 0.5, stop training and use the generator at the point of training stop as the pre-trained data generation model. At this point, the generator has learned the true distribution of historical calibration data and can generate pseudo-expanded samples that are highly consistent with the distribution of real historical data based on given condition variables.

[0076] Furthermore, a comprehensive evaluation is performed on the expanded calibration data. Based on the evaluation results, the workstation-range combinations requiring supplementary data and the corresponding input feature intervals are determined. These input feature intervals are then used as supplementary conditions and input into the data generation model to iteratively generate supplementary calibration data until the preset data sufficiency requirements are met, resulting in the target expanded calibration dataset, including: Step 1: Count the number of samples for each station-range combination in the expanded calibration data, and mark the station-range combinations with a sample number lower than a preset threshold as the first combination to be supplemented; Step 2: For each workstation-range combination where the number of samples has reached the preset number threshold, the standard resistance value range corresponding to the workstation-range combination is divided into multiple sub-intervals. The sample distribution density in each sub-interval is calculated. Workstation-range combinations in sub-intervals where the sample distribution density is lower than the preset density threshold are marked as the second combination to be supplemented. Step 3: Take the first and second supplementary combinations as supplementary combinations, and determine the standard resistance value sub-range and ambient temperature range corresponding to the insufficient samples in each supplementary combination, so as to obtain the supplementary standard resistance value sub-range and supplementary ambient temperature range corresponding to each supplementary combination. Step 4: For each combination to be supplemented, take the midpoint value of the sub-interval of the standard resistance value to be supplemented as the representative value of the standard resistance value, take the midpoint value of the ambient temperature range to be supplemented as the representative value of the ambient temperature, and take the representative value of the standard resistance value, the representative value of the ambient temperature, the workstation number corresponding to the combination to be supplemented, and the range level corresponding to the combination to be supplemented as the target generation conditions, input them into the data generation model, and generate the expanded calibration data; Step 5: Repeat steps 1 to 4 until the number of samples for all workstation-range combinations is greater than or equal to the preset number threshold, and the sample distribution density of each workstation-range combination in each sub-interval is greater than or equal to the preset density threshold. At this point, integrate all the extended calibration data into the target extended calibration dataset.

[0077] Specifically, in this embodiment of the invention, the comprehensive evaluation and iterative supplementation process is performed according to the following steps.

[0078] Step 1: Count the number of samples for each station-range combination in the expanded calibration data. A station-range combination refers to a data group determined by the station number and the range setting. For example, "station 1-200Ω range" is one station-range combination, "station 1-2kΩ range" is another, and so on. Station-range combinations with a sample size below a preset threshold are marked as the first combinations to be supplemented. The preset threshold is determined based on the minimum sample size required for training the error prediction model; for example, it can be set to 50 samples. When the sample size for a certain station-range combination is less than 50, it is considered that the sample size for that combination is insufficient to support model training, and supplementary data is needed.

[0079] Step 2: For each workstation-range combination where the number of samples has reached the preset threshold, the standard resistance value range corresponding to the workstation-range combination is divided into multiple sub-intervals. The sample distribution density in each sub-interval is calculated, and the workstation-range combination with a sample distribution density lower than the preset density threshold is marked as the second combination to be supplemented.

[0080] Specifically, for each station-range combination, the standard resistance value range is the range of that range. For example, for the "station 1-200Ω range" combination, the standard resistance value range is 0Ω to 200Ω. This range is divided into multiple sub-ranges of equal width or frequency, for example, 10 sub-ranges, each with a width of 20Ω. The number of existing samples in each sub-range is counted, and the sample distribution density of each sub-range is calculated. The sample distribution density is equal to the number of samples divided by the width of the sub-range. If the sample distribution density of a certain sub-range is lower than a preset density threshold, it indicates that the sample coverage of that station-range combination within the standard resistance value sub-range is insufficient, and supplementary data is needed. The preset density threshold can be set according to actual needs; for example, it can be set to include at least one sample per unit resistance value.

[0081] Step 3: Take the first and second supplementary combinations together as supplementary combinations, and determine the standard resistance value sub-range and ambient temperature range corresponding to the insufficient samples in each supplementary combination, so as to obtain the supplementary standard resistance value sub-range and supplementary ambient temperature range corresponding to each supplementary combination.

[0082] For the first combination to be supplemented, since the number of samples for the entire combination is insufficient, the sub-range of standard resistance values ​​to be supplemented is the entire measurement range, and the sub-range of ambient temperature to be supplemented is the ambient temperature range covered by the combination in the historical calibration data. For the second combination to be supplemented, the sub-range of standard resistance values ​​to be supplemented is a specific sub-range where the sample distribution density is lower than a preset density threshold, and the sub-range of ambient temperature to be supplemented is the ambient temperature range covered by existing samples within that sub-range or the overall ambient temperature range of the combination in the historical calibration data.

[0083] Step 4: For each combination to be supplemented, the midpoint value of the sub-interval of the standard resistance value to be supplemented is taken as the representative value of the standard resistance value, the midpoint value of the ambient temperature range to be supplemented is taken as the representative value of the ambient temperature, and the representative value of the standard resistance value, the representative value of the ambient temperature, the workstation number corresponding to the combination to be supplemented, and the range corresponding to the combination to be supplemented are taken as the target generation conditions, input into the data generation model, and the expanded calibration data is generated.

[0084] For example, assuming a certain combination to be supplemented is "station 3-2kΩ range", the sub-range of the standard resistance value to be supplemented is 1.0kΩ to 1.5kΩ, and the range of the ambient temperature to be supplemented is 22℃ to 26℃, then the representative value of the standard resistance value is taken as 1.25kΩ, and the representative value of the ambient temperature is 24℃. The station number "3", the range "2kΩ range", the standard resistance value "1.25kΩ" and the ambient temperature "24℃" are used as the target generation conditions input data to generate the model. The model generates the corresponding pseudo measurement error and outputs one or more supplementary calibration data.

[0085] Step 5: Repeat steps 1 to 4. After each generation of supplementary calibration data, recalculate the sample size and density for each station-range combination, re-mark the combinations to be supplemented, and generate supplementary data again, until the sample size for all station-range combinations is greater than or equal to a preset threshold, and the sample density for each station-range combination within each sub-interval is greater than or equal to a preset density threshold. At this point, integrate all the extended calibration data into the target extended calibration dataset.

[0086] Through the above iterative supplementation process, the target expanded calibration dataset has achieved sufficient sample quantity and balanced distribution under each workstation-range combination, providing a high-quality data foundation for the training of subsequent error prediction models.

[0087] Further, using the pseudo-standard resistance values, pseudo-ambient temperatures, and pseudo-station numbers from the target expanded calibration dataset as input features, and the pseudo-measurement errors from the target expanded calibration dataset as supervision labels, an error prediction model is trained, including: An initial error prediction model is constructed based on a deep neural network. The initial error prediction model includes a workstation embedding layer, a feature concatenation layer, multiple fully connected hidden layers, and an output layer. The workstation embedding layer is used to map the workstation number of each calibration workstation into a trainable embedding vector. During training, the pseudo workstation number of each training sample in the target augmented calibration dataset is input into the workstation embedding layer to obtain the corresponding embedding vector. The embedding vector, the pseudo standard resistance value of the training sample, and the pseudo ambient temperature of the training sample are input into the feature splicing layer for splicing to obtain the spliced ​​feature. The spliced ​​feature is then sequentially input into the multiple fully connected hidden layers and the output layer to obtain the prediction error value of the training sample. With the training objective of minimizing the difference between the predicted error value and the pseudo-measurement error corresponding to the training sample, the network parameters of the initial error prediction model are updated until the verification convergence is obtained, thus obtaining the error prediction model.

[0088] Specifically, an initial error prediction model is constructed based on a deep neural network. The initial error prediction model includes a workstation embedding layer, a feature concatenation layer, multiple fully connected hidden layers, and an output layer.

[0089] The workstation embedding layer maps the workstation numbers of each calibration workstation to trainable embedding vectors. Since the workstation numbers are discrete categorical features, ranging from 1 to 8, directly using one-hot encoding would result in high feature dimensionality and an inability to capture potential similarities between workstations. The workstation embedding layer maps each workstation number to a low-dimensional, dense embedding vector. The dimension of the embedding vector can be set according to actual needs, such as 8 or 16 dimensions. The embedding vectors corresponding to each workstation number are continuously updated during model training through backpropagation, thereby learning the unique error bias patterns of each workstation.

[0090] During training, the pseudo-workstation number of each training sample in the target augmented calibration dataset is input into the workstation embedding layer to obtain the corresponding embedding vector. The embedding vector, the pseudo-standard resistance value of the training sample, and the pseudo-ambient temperature of the training sample are then input into the feature concatenation layer for concatenation to obtain the concatenated feature. The concatenated feature simultaneously contains the category information of the workstation number, the numerical information of the standard resistance value, and the numerical information of the ambient temperature, achieving a unified representation of multi-source heterogeneous features.

[0091] The concatenated features are sequentially input into multiple fully connected hidden layers and an output layer to obtain the prediction error values ​​for the training samples. The fully connected hidden layers perform non-linear transformations on the concatenated features, extracting high-order interaction information between features. Each fully connected hidden layer contains several neurons, and the activation function can be the ReLU function. Dropout regularization can be introduced between hidden layers to prevent overfitting. The output layer consists of a single neuron that outputs the prediction error value.

[0092] The training objective is to minimize the difference between the predicted error value and the pseudo-measurement error corresponding to the training samples. The network parameters of the initial error prediction model are updated until the validation convergence, yielding the error prediction model. Specifically, the mean squared error loss function can be used as the optimization objective, and the adaptive moment estimation optimization algorithm is used to update the network parameters. The target is expanded, and the calibration dataset is proportionally divided into training and validation sets. Parameters are updated on the training set, and loss changes are monitored on the validation set. Training stops when the validation set loss no longer decreases for several consecutive training epochs, and the model at which training stops is taken as the error prediction model.

[0093] Furthermore, after the error prediction model is trained, the current ambient temperature and the standard resistance value and station number of each calibration station are input into the error prediction model, and the predicted error value of the current calibration cycle is output.

[0094] Specifically, within the current calibration cycle, the control module of the calibration system collects the current ambient temperature through a temperature sensor and, in conjunction with the standard resistance value of each station at the current range and the station number, constructs an input feature vector for each station. The input feature vectors of each station are then input into a trained error prediction model. After forward inference calculations, the model outputs the predicted error value for each station under the current calibration conditions.

[0095] The negative of the predicted error value is used as the error correction coefficient for each calibration station, i.e., error correction coefficient = -predicted error value. The principle is that the predicted error value represents the measurement deviation that a handheld digital multimeter may produce under given conditions. To compensate for this deviation, a correction amount of opposite direction and equal magnitude needs to be applied to the measured value. For example, if the error prediction model predicts that the measurement error of a certain station under current conditions is +0.5Ω, that is, the measured value is 0.5Ω larger than the standard value, then the error correction coefficient is -0.5Ω.

[0096] Finally, the resistance measurements at each calibration station are added to the corresponding error correction coefficient to obtain the corrected measurement value for each calibration station. That is, corrected measurement value = resistance measurement value + error correction coefficient = resistance measurement value - predicted error value. Through this correction, systematic measurement deviations caused by instrument characteristics, station differences, and environmental factors are effectively eliminated, making the corrected measurement value closer to the true standard resistance value.

[0097] For example, suppose that during the current calibration cycle, station 1 is calibrated at the 200Ω setting, the standard resistance value is 200Ω, and the current ambient temperature is 25℃. Inputting station number "1", standard resistance value "200Ω", and ambient temperature "25℃" into the trained error prediction model, the model outputs a predicted error of +0.3Ω. This indicates that under the conditions of station 1, standard resistance value 200Ω, and ambient temperature 25℃, the historical statistical measurement error of the handheld digital multimeter is on average 0.3Ω too high. Therefore, the error correction coefficient is -0.3Ω.

[0098] Subsequently, the resistance measurement value displayed on the handheld digital multimeter at workstation 1 was read as 200.4Ω using an image sensor. Adding an error correction factor of -0.3Ω to the resistance measurement value of 200.4Ω yielded a corrected measurement value of 200.1Ω. The deviation between the corrected measurement value of 200.1Ω and the standard resistance value of 200Ω was only 0.1Ω, a significant reduction compared to the uncorrected deviation of 0.4Ω, thus verifying the effectiveness of the error correction.

[0099] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0100] This invention utilizes a conditional generative adversarial network to generate pseudo-expansion samples that closely match the distribution of real historical data. Combined with comprehensive evaluation and iterative supplementation strategies, it ensures sufficient and balanced sample distribution for each workstation-range combination within each standard resistance value sub-interval, guaranteeing model training quality from a data perspective. A workstation embedding layer maps discrete workstation numbers to trainable dense embedding vectors, effectively capturing the unique error deviation patterns of each workstation. These patterns are then fused with continuous features such as standard resistance values ​​and ambient temperature, and nonlinearly fitted using a deep neural network, achieving accurate prediction of error values. This technical solution significantly improves the accuracy of error correction coefficient determination and the reliability of calibration results under limited historical data conditions.

[0101] S500: Compare the calibrated measured values ​​of each calibration station with the corresponding standard resistance values, and determine the resistance calibration results of each handheld digital multimeter based on the comparison results.

[0102] In existing handheld digital multimeter resistance calibration methods, error comparison and pass / fail determination mainly rely on manual review of the technical specifications of each model and range, manual calculation of tolerance thresholds, and comparison with measurement errors. This is not only inefficient and prone to errors, but also significantly increases the complexity of manual processing in eight-station parallel calibration scenarios involving multiple models and ranges. Furthermore, traditional methods typically perform only one measurement and one determination. When the measurement error slightly exceeds the tolerance threshold, a single measurement result is insufficient to reliably reflect the instrument's true metrological performance. Directly determining it as unqualified may lead to misclassifying marginally qualified instruments as unqualified, resulting in unnecessary repair or scrapping costs.

[0103] Furthermore, after calibration, manual data processing and calibration report compilation are required, further extending the calibration cycle. Therefore, how to achieve automated comparison of measured values ​​after calibration at each workstation with standard resistance values, automated acquisition and comparison of tolerance thresholds for each model and range, automated retesting and secondary judgment when measured values ​​exceed tolerances, and automated generation of calibration result reports have become urgent technical problems to be solved.

[0104] Step S500 in the method provided in this embodiment of the invention includes: The difference between the corrected measured value and the corresponding standard resistance value at each calibration station is calculated as the final measurement error at each calibration station. Based on the model information of each handheld digital multimeter, the tolerance threshold corresponding to each range is obtained from the pre-stored model-range configuration mapping table, and the absolute value of the final measurement error of each calibration station is compared with the corresponding tolerance threshold. When the absolute value of the final measurement error is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is determined to be calibrated successfully in the current range. When the absolute value of the final measurement error is greater than the tolerance threshold, the standard resistance value output by the standard resistance output channel remains unchanged. The resistance measurement value displayed by the handheld digital multimeter is read again by the image sensor. The final measurement error is recalculated based on the reread resistance measurement value. The absolute value of the recalculated final measurement error is compared with the tolerance threshold again. If the comparison result is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated successfully in the current range. If the comparison result is greater than the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated unsuccessfully in the current range. Based on the calibration pass / fail results of each calibration station at each range, and the final measurement error of each calibration station, the resistance calibration results of each handheld digital multimeter are generated.

[0105] In this embodiment of the invention, firstly, the difference between the corrected measured value of each calibration station and the corresponding standard resistance value is calculated as the final measurement error of each calibration station.

[0106] Specifically, for each calibration station, the calibrated measurement value at the current range is obtained, while the standard resistance value currently output by that station is a known quantity. The control module of the calibration system reads the calibrated measurement value and the corresponding standard resistance value of each station, and calculates the final measurement error as: calibrated measurement value - standard resistance value.

[0107] The final measurement error can be positive, negative, or zero. The calculated final measurement error is stored in the calibration system's data storage module for subsequent pass / fail determination and report generation.

[0108] For example, assuming station 1 is calibrated at the 200Ω range, with a standard resistance value of 200Ω and a measured value of 200.1Ω after calibration, the final measurement error is 200.1Ω - 200Ω = +0.1Ω. Assuming station 2 is calibrated at the 2kΩ range, with a standard resistance value of 2kΩ and a measured value of 1.998kΩ after calibration, the final measurement error is 1.998kΩ - 2kΩ = -0.002kΩ = -2Ω.

[0109] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0110] Furthermore, after calculating the final measurement error of each calibration station, based on the model information of each handheld digital multimeter, the tolerance threshold corresponding to each range is obtained from the pre-stored model-range configuration mapping table, and the absolute value of the final measurement error of each calibration station is compared with the corresponding tolerance threshold.

[0111] Specifically, the model-range configuration mapping table stores not only the resistance range and range of each model of handheld digital multimeter, but also the tolerance threshold corresponding to each range. The tolerance threshold refers to the maximum absolute value of the measurement error allowed by the handheld digital multimeter at that range, and is usually determined by the accuracy class of the instrument. The control module of the calibration system queries the model-range configuration mapping table based on the model information and current range information of the handheld digital multimeter corresponding to each station to obtain the tolerance threshold for each station at the current range.

[0112] For example, suppose the handheld digital multimeter at station 1 is model A, with a tolerance threshold of ±0.5Ω in the 200Ω range. The handheld digital multimeter at station 2 is model B, with a tolerance threshold of ±4Ω in the 2kΩ range. The final measurement error at station 1 is +0.1Ω, and the final measurement error at station 2 is -2Ω. After the control module obtains the above tolerance thresholds from the model-range configuration mapping table, it compares the absolute value of the final measurement error at station 1, |+0.1Ω|=0.1Ω, with the tolerance threshold of 0.5Ω, and compares the absolute value of the final measurement error at station 2, |-2Ω|=2Ω, with the tolerance threshold of 4Ω.

[0113] Furthermore, when the absolute value of the final measurement error is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed calibrated successfully in the current range. When the absolute value of the final measurement error is greater than the tolerance threshold, it is not directly deemed unqualified, but a retest process is initiated. Specifically, the standard resistance value output by the standard resistance output channel remains unchanged, and the resistance measurement value displayed on the handheld digital multimeter is read again by the image sensor. Based on the reread resistance measurement value, the final measurement error is recalculated, and the absolute value of the recalculated final measurement error is compared again with the tolerance threshold.

[0114] Specifically, the control module of the calibration system sends a hold command to the high-precision resistor network to ensure that the current standard resistance value output by each standard resistor output channel remains unchanged. At the same time, the control module sends an image acquisition trigger command to the image sensor at the corresponding calibration station again. The image sensor re-acquires the display panel image of the handheld digital multimeter at that station, and obtains a new resistance measurement value after tilt correction and optical character recognition. The control module corrects the new resistance measurement value according to the same error correction method as in the previous steps to obtain a new corrected measurement value. Then, the difference between the new corrected measurement value and the standard resistance value is calculated as the new final measurement error.

[0115] If the result of the second comparison is that the absolute value of the recalculated final measurement error is less than or equal to the tolerance threshold, then the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated successfully in the current range. If the result of the second comparison is that the absolute value of the recalculated final measurement error is still greater than the tolerance threshold, then the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated unsuccessfully in the current range.

[0116] The retesting mechanism effectively avoids misjudgments caused by single random fluctuations in the measurement values ​​of handheld digital multimeters, thus improving the reliability of calibration result determination.

[0117] It should be noted that the retesting process can be performed once or multiple times. In this embodiment of the invention, it is preferred to perform one retest, that is, after the first deviation, a second measurement and judgment are performed, and if the second deviation is still exceeded, the result is deemed unqualified. In an alternative embodiment, an upper limit on the number of retests can also be set; if multiple consecutive retests exceed the tolerance, the result is deemed unqualified, and if any retest is qualified, the result is deemed qualified. This embodiment does not limit this.

[0118] Furthermore, after completing the pass / fail determination of all calibration stations at each range, a resistance range calibration report for each handheld digital multimeter is generated based on the pass / fail determination results of each calibration station at each range and the final measurement error of each calibration station.

[0119] Specifically, the calibration report is automatically generated by the report generation module of the calibration system. The report content includes, but is not limited to: (i) Basic calibration information: including calibration date, calibration ambient temperature, calibration system number, calibration report number, etc.; (ii) Information of the calibrated instrument: including the model information, serial number, and station number of the handheld digital multimeter corresponding to each station; (iii) Calibration condition information: including the standard resistance value and tolerance threshold of each station at each range; (iv) Calibration measurement data: including the resistance measurement value, calibrated measurement value, and final measurement error of each station at each range; (v) Calibration result judgment: including the calibration pass or fail judgment result of each station at each range; (vi) Calibration conclusion: summarizing the calibration results of each handheld digital multimeter at each range and giving an overall calibration conclusion.

[0120] Calibration reports can be stored in the calibration system's storage module using a structured data format, or output as a visual calibration report file according to a preset template format. The report generation module has multiple built-in report templates, allowing users to select the appropriate template for different handheld digital multimeter models.

[0121] In summary, the embodiments of the present invention have at least the following technical effects: This invention enables parallel calibration of eight instruments by combining eight calibration stations with eight independent output channels of a high-precision resistor network, significantly improving batch calibration efficiency. A pre-stored model-range configuration mapping table automates the acquisition of range configuration information and standardizes the determination of standard resistance values, avoiding subjective biases from manual review and point selection. A field-programmable gate array (FPGA) generates a synchronous trigger signal, ensuring all eight channels output standard resistance values ​​simultaneously, eliminating inconsistencies in measurement start times at each station caused by time-sharing output. Automatic image acquisition via an image sensor, combined with tilt correction and optical character recognition, replaces manual reading, improving reading efficiency and accuracy. Data augmentation using a conditional generative adversarial network (GAN) combined with comprehensive evaluation and iterative supplementation strategies, along with a station embedding layer mapping discrete station numbers to dense embedding vectors to fuse multi-source features for error prediction, achieves precise determination of error correction coefficients. Automatic acquisition of tolerance thresholds for comparison and judgment, combined with a retesting mechanism to avoid single-time misjudgments, and automatic generation of calibration reports form a complete automated closed loop from data acquisition to report output, effectively improving the reliability of calibration results and work efficiency.

[0122] In summary, this invention solves the technical problems of low efficiency and poor accuracy of calibration results caused by the reliance on manual operation in existing handheld digital multimeter resistance calibration methods, as well as the inability of each workstation to perform synchronous parallel calibration.

[0123] Example 2, as Figure 4As shown, based on the same inventive concept as the eight-position parallel resistance calibration method based on a high-precision resistance network provided in Embodiment 1, this embodiment of the invention also provides an eight-position parallel resistance calibration device based on a high-precision resistance network, comprising: The range configuration and station deployment module 11 is used to obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range according to the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration stations respectively. The standard resistor synchronous output module 12 is used to synchronously output the standard resistor value corresponding to each of the eight calibration stations through a high-precision resistor network. The measurement value image recognition and reading module 13 is used to read the resistance measurement values ​​displayed by each handheld digital multimeter through the image sensor corresponding to each calibration station; The error correction coefficient determination and correction module 14 is used to determine the error correction coefficient corresponding to each calibration station based on the historical calibration data of each calibration station through data expansion and comprehensive evaluation, and to correct the resistance measurement value of each station according to the error correction coefficient to obtain the corrected measurement value of each calibration station. The calibration result determination and report generation module 15 is used to compare the corrected measured value of each calibration station with the corresponding standard resistance value, and determine the resistance calibration result of each handheld digital multimeter based on the comparison result.

[0124] In one embodiment, the range configuration and workstation deployment module 11 is specifically used for: Obtain the model information of each handheld digital multimeter to be calibrated, and query the pre-stored model-range configuration mapping table according to the model information to obtain the resistance range configuration information of each handheld digital multimeter. The resistance range configuration information includes the resistance range supported by each handheld digital multimeter and the range of each resistance range. Based on the resistance range configuration information, a calibration point is selected within the range of each resistance range to determine the standard resistance value corresponding to each range. Place the eight handheld digital multimeters to be calibrated into the eight calibration stations respectively.

[0125] In one embodiment, the standard resistor synchronous output module 12 is specifically used for: Obtain the network topology of the high-precision resistor network, which includes eight independent standard resistor output channels. Configure the output resistance value of each standard resistor output channel according to the standard resistor value corresponding to each workstation; A synchronous trigger signal is generated by a field-programmable gate array and simultaneously sent to the eight standard resistor output channels of the high-precision resistor network, so that each standard resistor output channel outputs the configured standard resistor value to the corresponding calibration station at the same time.

[0126] In one embodiment, the measurement value image recognition and reading module 13 is specifically used for: The display panel image of the handheld digital multimeter is acquired by the image sensor corresponding to each calibration station; The display panel image is subjected to tilt correction processing to obtain a corrected display panel image; The corrected display panel image is input into a pre-trained optical character recognition model, and the output is the resistance measurement value displayed by the handheld digital multimeter.

[0127] In one embodiment, the error correction coefficient determination and correction module 14 is specifically used for: Acquire historical calibration data accumulated by each calibration station in previous calibrations. The historical calibration data includes historical standard resistance value, historical resistance measurement value, historical ambient temperature, historical station number, historical range setting, and corresponding historical measurement error. The historical measurement error is the difference between the historical resistance measurement value and the historical standard resistance value. Based on the historical standard resistance values, historical ambient temperatures, historical workstation numbers, and historical measurement ranges in the historical calibration data, expanded calibration data is generated through a pre-trained data generation model. The expanded calibration data includes pseudo standard resistance values, pseudo ambient temperatures, pseudo workstation numbers, pseudo measurement ranges, and pseudo measurement errors. A comprehensive evaluation is performed on the expanded calibration data. Based on the comprehensive evaluation results, the workstation-range combination that needs to be supplemented and the input feature interval corresponding to the workstation-range combination are determined. The input feature interval is then used as a supplementary condition and input into the data generation model to iteratively generate supplementary calibration data until the preset data sufficiency requirements are met, thereby obtaining the target expanded calibration dataset. Using the pseudo-standard resistance value, pseudo-ambient temperature, and pseudo-station number in the target expanded calibration dataset as input features, and the pseudo-measurement error in the target expanded calibration dataset as supervision label, an error prediction model is trained. Input the current ambient temperature, the standard resistance value of each calibration station, and the station number into the error prediction model, and output the predicted error value for the current calibration cycle. The negative of the predicted error value is used as the error correction coefficient for each calibration station, and the corresponding error correction coefficient is added to the resistance measurement value of each calibration station to obtain the corrected measurement value of each calibration station.

[0128] The pre-training process of the data generation model includes: Multiple sets of historical calibration data samples are obtained. Each set of historical calibration data samples includes historical standard resistance value samples, historical ambient temperature samples, historical workstation number samples, historical range samples, and corresponding historical measurement error samples. Based on conditional generative adversarial networks, an initial data generation model is constructed, which includes a generator and a discriminator. The generator generates a pseudo-expansion sample based on the historical standard resistance value sample, the historical ambient temperature sample, the historical workstation number sample, and the historical range sample. The pseudo-expansion sample includes a pseudo standard resistance value, a pseudo ambient temperature, a pseudo workstation number, a pseudo range, and a pseudo measurement error. The discriminator receives the pseudo-expansion sample and the real expansion sample composed of the historical calibration data sample, and determines the authenticity of the input sample. The generator and the discriminator alternately update the network parameters until the discriminator can no longer distinguish between the pseudo-expanded samples and the real expanded samples, at which point training stops, and the generator at the point where training stops is used as the pre-trained data generation model.

[0129] The expanded calibration data is comprehensively evaluated. Based on the evaluation results, the workstation-range combinations requiring supplementary data and the corresponding input feature intervals are determined. These input feature intervals are then used as supplementary conditions and input into the data generation model. Supplementary calibration data is iteratively generated until the preset data sufficiency requirements are met, resulting in the target expanded calibration dataset, including: Step 1: Count the number of samples for each station-range combination in the expanded calibration data, and mark the station-range combinations with a sample number lower than a preset threshold as the first combination to be supplemented; Step 2: For each workstation-range combination where the number of samples has reached the preset number threshold, the standard resistance value range corresponding to the workstation-range combination is divided into multiple sub-intervals. The sample distribution density in each sub-interval is calculated. Workstation-range combinations in sub-intervals where the sample distribution density is lower than the preset density threshold are marked as the second combination to be supplemented. Step 3: Take the first and second supplementary combinations as supplementary combinations, and determine the standard resistance value sub-range and ambient temperature range corresponding to the insufficient samples in each supplementary combination, so as to obtain the supplementary standard resistance value sub-range and supplementary ambient temperature range corresponding to each supplementary combination. Step 4: For each combination to be supplemented, take the midpoint value of the sub-interval of the standard resistance value to be supplemented as the representative value of the standard resistance value, take the midpoint value of the ambient temperature range to be supplemented as the representative value of the ambient temperature, and take the representative value of the standard resistance value, the representative value of the ambient temperature, the workstation number corresponding to the combination to be supplemented, and the range level corresponding to the combination to be supplemented as the target generation conditions, input them into the data generation model, and generate the expanded calibration data; Step 5: Repeat steps 1 to 4 until the number of samples for all workstation-range combinations is greater than or equal to the preset number threshold, and the sample distribution density of each workstation-range combination in each sub-interval is greater than or equal to the preset density threshold. At this point, integrate all the extended calibration data into the target extended calibration dataset.

[0130] Using pseudo-standard resistance values, pseudo-ambient temperatures, and pseudo-station numbers from the target expanded calibration dataset as input features, and using pseudo-measurement errors from the target expanded calibration dataset as supervision labels, an error prediction model is trained, including: An initial error prediction model is constructed based on a deep neural network. The initial error prediction model includes a workstation embedding layer, a feature concatenation layer, multiple fully connected hidden layers, and an output layer. The workstation embedding layer is used to map the workstation number of each calibration workstation into a trainable embedding vector. During training, the pseudo workstation number of each training sample in the target augmented calibration dataset is input into the workstation embedding layer to obtain the corresponding embedding vector. The embedding vector, the pseudo standard resistance value of the training sample, and the pseudo ambient temperature of the training sample are input into the feature splicing layer for splicing to obtain the spliced ​​feature. The spliced ​​feature is then sequentially input into the multiple fully connected hidden layers and the output layer to obtain the prediction error value of the training sample. With the training objective of minimizing the difference between the predicted error value and the pseudo-measurement error corresponding to the training sample, the network parameters of the initial error prediction model are updated until the verification convergence is obtained, thus obtaining the error prediction model.

[0131] In one embodiment, the calibration result determination and report generation module 15 is specifically used for: The difference between the corrected measured value and the corresponding standard resistance value at each calibration station is calculated as the final measurement error at each calibration station. Based on the model information of each handheld digital multimeter, the tolerance threshold corresponding to each range is obtained from the pre-stored model-range configuration mapping table, and the absolute value of the final measurement error of each calibration station is compared with the corresponding tolerance threshold. When the absolute value of the final measurement error is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is determined to be calibrated successfully in the current range. When the absolute value of the final measurement error is greater than the tolerance threshold, the standard resistance value output by the standard resistance output channel remains unchanged. The resistance measurement value displayed by the handheld digital multimeter is read again by the image sensor. The final measurement error is recalculated based on the reread resistance measurement value. The absolute value of the recalculated final measurement error is compared with the tolerance threshold again. If the comparison result is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated successfully in the current range. If the comparison result is greater than the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated unsuccessfully in the current range. Based on the calibration pass / fail results of each calibration station at each range, and the final measurement error of each calibration station, the resistance calibration results of each handheld digital multimeter are generated.

[0132] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0134] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. An eight-position parallel resistance calibration method based on a high-precision resistance network, characterized in that, The method includes: Obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range according to the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration stations respectively. The standard resistance value corresponding to each of the eight calibration stations is synchronously output to the eight calibration stations through a high-precision resistor network. The resistance measurement values ​​displayed on each handheld digital multimeter are read by the image sensor corresponding to each calibration station; Based on the historical calibration data of each calibration station, the error correction coefficient corresponding to each station is determined through data expansion and comprehensive evaluation. The resistance measurement value of each station is then corrected according to the error correction coefficient to obtain the corrected measurement value of each calibration station. The calibrated measured values ​​at each calibration station are compared with the corresponding standard resistance values, and the resistance calibration results of each handheld digital multimeter are determined based on the comparison results.

2. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 1, characterized in that, Obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range based on the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration stations respectively, including: Obtain the model information of each handheld digital multimeter to be calibrated, and query the pre-stored model-range configuration mapping table according to the model information to obtain the resistance range configuration information of each handheld digital multimeter. The resistance range configuration information includes the resistance range supported by each handheld digital multimeter and the range of each resistance range. Based on the resistance range configuration information, a calibration point is selected within the range of each resistance range to determine the standard resistance value corresponding to each range. Place the eight handheld digital multimeters to be calibrated into the eight calibration stations respectively.

3. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 1, characterized in that, The standard resistance value corresponding to each of the eight calibration stations is synchronously output to the eight calibration stations through a high-precision resistor network, including: Obtain the network topology of the high-precision resistor network, which includes eight independent standard resistor output channels. Configure the output resistance value of each standard resistor output channel according to the standard resistor value corresponding to each workstation; A synchronous trigger signal is generated by a field-programmable gate array and simultaneously sent to the eight standard resistor output channels of the high-precision resistor network, so that each standard resistor output channel outputs the configured standard resistor value to the corresponding calibration station at the same time.

4. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 1, characterized in that, The resistance measurement values ​​displayed on each handheld digital multimeter are read through the image sensor corresponding to each calibration station, including: The display panel image of the handheld digital multimeter is acquired by the image sensor corresponding to each calibration station; The display panel image is subjected to tilt correction processing to obtain a corrected display panel image; The corrected display panel image is input into a pre-trained optical character recognition model, and the output is the resistance measurement value displayed by the handheld digital multimeter.

5. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 1, characterized in that, Based on historical calibration data from each calibration station, error correction coefficients are determined for each station through data expansion and comprehensive evaluation. The resistance measurements at each station are then corrected using these error correction coefficients to obtain the corrected measurements for each calibration station, including: Acquire historical calibration data accumulated by each calibration station in previous calibrations. The historical calibration data includes historical standard resistance value, historical resistance measurement value, historical ambient temperature, historical station number, historical range setting, and corresponding historical measurement error. The historical measurement error is the difference between the historical resistance measurement value and the historical standard resistance value. The historical ambient temperature is acquired by a temperature sensor. Based on the historical standard resistance values, historical ambient temperatures, historical workstation numbers, and historical measurement ranges in the historical calibration data, expanded calibration data is generated through a pre-trained data generation model. The expanded calibration data includes pseudo standard resistance values, pseudo ambient temperatures, pseudo workstation numbers, pseudo measurement ranges, and pseudo measurement errors. A comprehensive evaluation is performed on the expanded calibration data. Based on the comprehensive evaluation results, the workstation-range combination that needs to be supplemented and the input feature interval corresponding to the workstation-range combination are determined. The input feature interval is then used as a supplementary condition and input into the data generation model to iteratively generate supplementary calibration data until the preset data sufficiency requirements are met, thereby obtaining the target expanded calibration dataset. Using the pseudo-standard resistance value, pseudo-ambient temperature, and pseudo-station number in the target expanded calibration dataset as input features, and the pseudo-measurement error in the target expanded calibration dataset as supervision label, an error prediction model is trained. Input the current ambient temperature, the standard resistance value of each calibration station, and the station number into the error prediction model, and output the predicted error value for the current calibration cycle. The negative of the predicted error value is used as the error correction coefficient for each calibration station, and the corresponding error correction coefficient is added to the resistance measurement value of each calibration station to obtain the corrected measurement value of each calibration station.

6. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 5, characterized in that, The pre-training process of the data generation model includes: Multiple sets of historical calibration data samples are obtained. Each set of historical calibration data samples includes historical standard resistance value samples, historical ambient temperature samples, historical workstation number samples, historical range samples, and corresponding historical measurement error samples. Based on conditional generative adversarial networks, an initial data generation model is constructed, which includes a generator and a discriminator. The generator generates a pseudo-expansion sample based on the historical standard resistance value sample, the historical ambient temperature sample, the historical workstation number sample, and the historical range sample. The pseudo-expansion sample includes a pseudo standard resistance value, a pseudo ambient temperature, a pseudo workstation number, a pseudo range, and a pseudo measurement error. The discriminator receives the pseudo-expansion sample and the real expansion sample composed of the historical calibration data sample, and determines the authenticity of the input sample. The generator and the discriminator alternately update the network parameters until the discriminator can no longer distinguish between the pseudo-expanded samples and the real expanded samples, at which point training stops, and the generator at the point where training stops is used as the pre-trained data generation model.

7. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 5, characterized in that, The expanded calibration data is comprehensively evaluated. Based on the evaluation results, the workstation-range combinations requiring supplementary data and the corresponding input feature intervals are determined. These input feature intervals are then used as supplementary conditions and input into the data generation model. Supplementary calibration data is iteratively generated until the preset data sufficiency requirements are met, resulting in the target expanded calibration dataset, including: Step 1: Count the number of samples for each station-range combination in the expanded calibration data, and mark the station-range combinations with a sample number lower than a preset threshold as the first combination to be supplemented; Step 2: For each workstation-range combination where the number of samples has reached the preset number threshold, the standard resistance value range corresponding to the workstation-range combination is divided into multiple sub-intervals. The sample distribution density in each sub-interval is calculated. Workstation-range combinations in sub-intervals where the sample distribution density is lower than the preset density threshold are marked as the second combination to be supplemented. Step 3: Take the first and second supplementary combinations as supplementary combinations, and determine the standard resistance value sub-range and ambient temperature range corresponding to the insufficient samples in each supplementary combination, so as to obtain the supplementary standard resistance value sub-range and supplementary ambient temperature range corresponding to each supplementary combination. Step 4: For each combination to be supplemented, take the midpoint value of the sub-interval of the standard resistance value to be supplemented as the representative value of the standard resistance value, take the midpoint value of the ambient temperature range to be supplemented as the representative value of the ambient temperature, and take the representative value of the standard resistance value, the representative value of the ambient temperature, the workstation number corresponding to the combination to be supplemented, and the range level corresponding to the combination to be supplemented as the target generation conditions, input them into the data generation model, and generate the expanded calibration data; Step 5: Repeat steps 1 to 4 until the number of samples for all workstation-range combinations is greater than or equal to the preset number threshold, and the sample distribution density of each workstation-range combination in each sub-interval is greater than or equal to the preset density threshold. At this point, integrate all the extended calibration data into the target extended calibration dataset.

8. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 5, characterized in that, Using pseudo-standard resistance values, pseudo-ambient temperatures, and pseudo-station numbers from the target expanded calibration dataset as input features, and using pseudo-measurement errors from the target expanded calibration dataset as supervision labels, an error prediction model is trained, including: An initial error prediction model is constructed based on a deep neural network. The initial error prediction model includes a workstation embedding layer, a feature concatenation layer, multiple fully connected hidden layers, and an output layer. The workstation embedding layer is used to map the workstation number of each calibration workstation into a trainable embedding vector. During training, the pseudo workstation number of each training sample in the target augmented calibration dataset is input into the workstation embedding layer to obtain the corresponding embedding vector. The embedding vector, the pseudo standard resistance value of the training sample, and the pseudo ambient temperature of the training sample are input into the feature splicing layer for splicing to obtain the spliced ​​feature. The spliced ​​feature is then sequentially input into the multiple fully connected hidden layers and the output layer to obtain the prediction error value of the training sample. With the training objective of minimizing the difference between the predicted error value and the pseudo-measurement error corresponding to the training sample, the network parameters of the initial error prediction model are updated until the verification convergence is obtained, thus obtaining the error prediction model.

9. The eight-station parallel resistance calibration method based on a high-precision resistance network according to claim 1, characterized in that, The calibrated measured values ​​at each calibration station are compared with the corresponding standard resistance values. Based on the comparison results, the resistance calibration results of each handheld digital multimeter are determined, including: The difference between the corrected measured value and the corresponding standard resistance value at each calibration station is calculated as the final measurement error at each calibration station. Based on the model information of each handheld digital multimeter, the tolerance threshold corresponding to each range is obtained from the pre-stored model-range configuration mapping table, and the absolute value of the final measurement error of each calibration station is compared with the corresponding tolerance threshold. When the absolute value of the final measurement error is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is determined to be calibrated successfully in the current range. When the absolute value of the final measurement error is greater than the tolerance threshold, the standard resistance value output by the standard resistance output channel remains unchanged. The resistance measurement value displayed by the handheld digital multimeter is read again by the image sensor. The final measurement error is recalculated based on the reread resistance measurement value. The absolute value of the recalculated final measurement error is compared with the tolerance threshold again. If the comparison result is less than or equal to the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated successfully in the current range. If the comparison result is greater than the tolerance threshold, the handheld digital multimeter corresponding to the calibration station is deemed to be calibrated unsuccessfully in the current range. Based on the calibration pass / fail results of each calibration station at each range and the final measurement error of each calibration station, the resistance calibration results of each handheld digital multimeter are generated.

10. An eight-station parallel resistance calibration device based on a high-precision resistance network, characterized in that, The method for performing the eight-station parallel resistance calibration method based on a high-precision resistance network according to any one of claims 1 to 9 includes: The range configuration and workstation deployment module is used to obtain the resistance range configuration information of each handheld digital multimeter to be calibrated, determine the standard resistance value corresponding to each range according to the resistance range configuration information, and place the eight handheld digital multimeters to be calibrated into eight calibration workstations respectively. A standard resistor synchronous output module is used to synchronously output the standard resistor value corresponding to each of the eight calibration stations through a high-precision resistor network. The measurement value image recognition and reading module is used to read the resistance measurement values ​​displayed by each handheld digital multimeter through the image sensor corresponding to each calibration station; The error correction coefficient determination and correction module is used to determine the error correction coefficient corresponding to each calibration station based on the historical calibration data of each calibration station through data expansion and comprehensive evaluation, and to correct the resistance measurement value of each station according to the error correction coefficient to obtain the corrected measurement value of each calibration station. The calibration result determination and report generation module is used to compare the corrected measured values ​​of each calibration station with the corresponding standard resistance values, and determine the resistance calibration results of each handheld digital multimeter based on the comparison results.

Images