A modular weighing system suitable for the pre-trained measurement model of electronic belt scales
By generating standardized structural labels through a modular weighing system, the problem of poor data consistency among electronic belt scale devices is solved, a unified data structure foundation and the identifiability of signal sources are achieved, and the applicability of the metrological pre-training model and the reusability of cross-device data are improved.
Patent Information
- Application Number
- CN202611337280.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-31
- Publication Date
- 2026-11-17
AI Technical Summary
In existing electronic belt scales, the inconsistent roller spacing, weighing area layout, and effective weighing length in the pre-training model lead to poor consistency of data samples between different devices, affecting the model's ability to determine the source of data differences and making it difficult to form a unified training sample.
A modular weighing system is adopted, including a mechanical weighing module, an identification and signal transmission module, a structural parameter storage module, a structural label generation module, and a measurement time series sample construction module. By uniformly storing hardware structural parameters, standardized labels are generated to eliminate interference from differences in the scale body structure and form structural label information suitable for pre-trained models.
It achieves a unified structural basis for data from different devices, improves the identifiability of weighing signal sources and the applicability of data training, reduces the impact of structural differences on model training, and enhances data consistency and reusability.
Smart Images

Figure CN123111426A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automatic weighing instrument measurement technology, and in particular to a modular weighing system suitable for a pre-trained model of electronic belt scale measurement. Background Technology
[0002] With the development of intelligent applications of electronic belt scales, long-term operational data has begun to be used for metrological status analysis and model training. For metrological pre-training models, the model not only needs weighing signals, speed signals, and flow data, but also needs to know under what weighing structural conditions these data were generated.
[0003] Current electronic belt scales focus more on the on-site adaptation of individual devices. The idler spacing, weighing zone layout, and effective weighing length may vary between different devices. Even if the data collected is of the same type, the data from different devices may correspond to different structural conditions.
[0004] When data from multiple electronic belt scales are used together to pre-train a measurement model, this structural difference can affect the consistency of the samples. The model struggles to determine whether the data discrepancies stem from changes in measurement status, or from differences in idler spacing, weighing zone length, or scale configuration.
[0005] While existing electronic belt scales offer advantages such as strong on-site adaptability, flexible installation, and mature application, they still have shortcomings in terms of the data requirements for pre-training measurement models. The main problem lies not in whether a single device can complete the measurement, but in the lack of a unified weighing structure across different devices. In particular, when the relationships between idler roller spacing, effective weighing length, and signal channels are inconsistent, it is difficult to directly form consistent training samples from long-term operational data. Summary of the Invention
[0006] To address the aforementioned problems, this invention provides a modular weighing system suitable for the pre-trained measurement model of electronic belt scales.
[0007] The present invention provides a modular weighing system suitable for the pre-training model of electronic belt scale measurement, including a mechanical weighing module: a weighing structure for mounting a preset idler spacing, which collects unit material load signals and supports multi-unit splicing to adjust the effective weighing length; Identification and signal transmission module: used to assign a unique identifier to each weighing unit and independently transmit each weighing signal, and synchronously upload the unit's operating status; Structural parameter storage module: used to uniformly store all hardware structural parameters and bind them to unit identifiers; Structural label generation module: integrates weighing signals, unit identifiers and structural parameters to generate standardized labels, and marks each segment of weighing time series data with corresponding hardware structure and operating condition information; Measurement Time Series Sample Construction Module: Used to pair and encapsulate time series weighing signals with structural labels, complete multi-device data alignment and grouping based on standardized labels, and generate a training dataset with a unified format that can distinguish hardware operating conditions; Metrology pre-training model adaptation module: used to parse the structural label features attached to the samples, eliminate the interference of scale body structural differences on training, support the model to learn general metrology rules and assist in the analysis of metrology faults and accuracy status.
[0008] In summary, the present invention has at least the following beneficial effects: 1. Facilitates the formation of a unified weighing structure foundation. The modular weighing unit of this invention has a preset idler roller spacing. During on-site installation, it can be adapted based on the structural parameters of this modular weighing unit. Therefore, the weighing data collected by different electronic belt scales at the modular weighing unit can correspond to the same or determinable idler roller spacing conditions, reducing the impact of on-site idler roller arrangement differences on subsequent model training.
[0009] 2. Improves the identifiability of weighing signal sources. This invention assigns each modular weighing unit a relatively independent weighing area and outputs a weighing signal corresponding to itself. This weighing signal can be correlated with a unit identifier or channel identifier. Therefore, it is possible to determine which modular weighing unit each weighing signal originates from, avoiding the situation where weighing data exists only as a whole-machine signal.
[0010] 3. Facilitates the formation of structural label information suitable for pre-trained models. This invention can establish a relationship between information such as the preset roller spacing, number of modules, bandwidth, installation position, effective state, and calibration parameters of the modular weighing unit and the corresponding weighing signal. This relationship can form structural label information. This structural label information can be used together with the weighing signal to construct measurement time-series samples, enabling the measurement pre-trained model to identify the structural conditions corresponding to the data.
[0011] 4. Improves the applicability of training data from different devices. This invention enables electronic belt scales to establish a data foundation with clearly defined structural parameters and identifiable signal sources during the data acquisition phase. Devices with different bandwidths or different numbers of modules can be distinguished through corresponding structural label information. This helps reduce the impact of differences in on-site structures on model training and improves the consistency and reusability of data from different devices used for pre-training measurement models. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.
[0013] Figure 1 This is a schematic diagram of the formation of structural label information and sample construction according to the present invention; Figure 2 This is a schematic diagram of the structure of a multi-roller suspended scale. Figure 3 This is a schematic diagram of the preset roller spacing of the modular weighing unit of the present invention. Detailed Implementation
[0014] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0015] The following is in conjunction with the appendix Figures 1 to 3 The present invention will be described in further detail below.
[0016] like Figure 1 As shown, a modular weighing system suitable for the pre-trained measurement model of an electronic belt scale includes a mechanical weighing module, an identification and signal transmission module, a structural parameter storage module, a structural label generation module, a measurement time series sample construction module, and a measurement pre-trained model adaptation module.
[0017] I. About the mechanical weighing module.
[0018] like Figure 2 As shown, this invention uses a multi-roller suspended scale body described in application number 202620548705.5 as an embodiment of the mechanical weighing module, and adopts as follows... Figure 3 The roller distribution shown.
[0019] II. Regarding the Identification and Signal Transmission Module.
[0020] This module assigns a unique identifier to each modular weighing unit, enables independent transmission of weighing signals generated by each unit through separate channels, and simultaneously collects and uploads all operational status information of the corresponding weighing unit. This module serves as an intermediary unit connecting the hardware weighing unit and the metrological pre-training model data link, receiving the sensor output signals from the modular weighing unit and outputting structured metrological data with complete traceability information.
[0021] This module uses two types of matching identification carriers to complete the unique identification assignment: at the hardware level, a unique unit identifier is assigned to each modular weighing unit; at the signal acquisition level, a unique channel identifier is configured for each unit signal. The two types of identifiers form a one-to-one binding relationship. Each modular weighing unit corresponds to an independent weighing area and only outputs its own weighing signal. When a signal is received, it will automatically be associated with the unit identifier or channel identifier corresponding to this unit, forming a binding logic where each signal matches a unique unit identifier.
[0022] The purpose of this step is twofold: first, to accurately locate the physical source of each weighing signal through dual-identification binding and to clarify the acquisition unit corresponding to the load data; and second, to avoid the problem of confusion and inability to distinguish acquisition units among multiple weighing signals in multi-module combined weighing scenarios, ensuring that the source of all weighing signals can be identified.
[0023] This module relies on the built-in signal interface and junction box of the modular weighing unit to establish an independent signal transmission link. After the load cell completes the unit load detection, it will output a weighing signal belonging only to this unit, which is transmitted through a dedicated junction box and independent line. The physical signal transmission channels of different units are independent of each other, and there will be no signal aliasing. A single modular weighing unit can be matched with one or more dedicated signal channels, and the number of channels is determined according to the sensor configuration inside the unit.
[0024] All data collected from all channels are bound to the corresponding unit identifier or channel identifier, enabling split transmission and split storage. At the same time, modular weighing units of the same specification have a unified preset idler spacing. The independently transmitted weighing signal naturally has a unified and determinable idler and idler spacing structural basis, and the signals output by units of the same specification on different devices have a consistent structural premise.
[0025] The purpose of this step is firstly to isolate the load signals of different weighing units by using hardware to separate the transmission, avoid mutual interference, and ensure the integrity and accuracy of the unit weighing data; secondly, by binding independent channels with unique identifiers, the signal transmission can be traced throughout the process, and any set of flow and load data can be back-matched to the corresponding modular weighing unit; finally, by relying on the independent signal of the uniformly preset idler spacing, the data baseline deviation caused by the differences in the original idler arrangement on site is eliminated, reducing the difficulty of subsequent model sample alignment.
[0026] This module, while transmitting weighing signals and unit identifiers, simultaneously collects and uploads complete operational information and structural parameters of the corresponding units, including unit installation location, effective working status, conveyor belt bandwidth, unit calibration parameters, and the number of modules used in the overall unit combination. Based on the effective weighing length calculated from the number of modules and unit parameters, this module will mark the effective status field of the unit in real time. If a weighing unit malfunctions or is out of service, an abnormal status will be marked simultaneously to distinguish the data of faulty units from those of normally functioning units. This module also uses unit identifiers and channel identifiers as core indexes to establish a one-to-one mapping relationship between independent weighing signals, unit operating status, and various structural parameters.
[0027] The purpose of this step is threefold: first, to automatically filter data based on synchronously uploaded unit status, allowing invalid measurement data from faulty units to be directly removed during model training; second, to bind structural parameters such as bandwidth, effective weighing length, and calibration parameters to weighing signals, distinguishing data differences caused by different conveyor specifications and module combinations; and third, to integrate scattered and independent signal data, hardware status, and structural parameters into a set of associated data through unified index mapping, providing a complete and comprehensive data source for generating standardized structural label information.
[0028] The weighing signal, unit identifier, unit operating status, and all structural parameters output by this module are combined to generate structural label information. This structural label information is standardized data specifically used to describe the source of the weighing signal, weighing structural conditions, and unit operating status. The label fully includes all associated fields such as preset idler roller spacing, number of modules combined, effective weighing length, conveyor bandwidth, unit installation location, effective unit operating status, signal channel correspondence, and sensor calibration parameters.
[0029] The purpose of this step is to transform the original weighing signal, which originally had no structural description, into labeled data carrying complete structural traceability information, thereby solving the problems of traditional electronic belt scales lacking structural descriptions in their data collection and the inability to conduct unified model training using data from multiple devices.
[0030] The weighing signal output by this module, combined with the generated structural label information, forms a standardized measurement time series sample. This entire set of measurement time series samples serves as the input dataset for the measurement pre-training model. The model can automatically distinguish the weighing unit, idler spacing, belt width, effective weighing length, and unit operating status corresponding to each set of data based on the unit identifier and structural label information.
[0031] The purpose of this step is twofold: first, to enable the pre-trained metrology model to autonomously identify the causes of data differences, distinguishing whether data fluctuations originate from changes in material load, equipment failure, or different weighing structure specifications of the belt scale; and second, to unify the data standards of multiple conveyors and multiple sets of electronic belt scales, improve the consistency and reusability of cross-equipment data, reduce the workload of sample alignment during model training, and enhance the model's generalization and adaptation capabilities to various metrology conditions.
[0032] III. Regarding the structural parameter storage module.
[0033] This module centrally and uniformly stores all hardware structural parameters corresponding to all modular weighing units of the electronic belt scale, and binds each set of stored hardware structural parameters with the unit identifier and channel identifier of the corresponding modular weighing unit. This module receives various structural data uploaded by the identification and signal transmission modules, providing a complete and unified parameter data source for the generation of structural label information.
[0034] The hardware parameters stored in this module are the preset roller spacing of the modular weighing unit. This preset roller spacing is a fixed structural parameter determined in advance according to the specifications of the modular weighing unit. The roller spacing value is kept uniform within the same specification unit, while different specification units correspond to different preset spacing values. This module will bind and store this spacing value with the unit identifier of the corresponding unit.
[0035] The purpose of this step is to retain uniform weighing structure benchmark parameters and record the fixed conditions of the roller arrangement inside different units. When constructing measurement time series samples later, the stored binding parameters can be used to distinguish the data baseline differences caused by the roller spacing of different units, reducing the difficulty of sample alignment when training data from multiple devices.
[0036] This module stores the effective weighing length parameter calculated from the combination of multiple modular weighing units. The effective weighing length is determined by the number of modular weighing units arranged along the conveyor belt running direction, the preset idler spacing of the units, and the arrangement of the units. This length parameter is bound to and stored with the identifiers of all units included in the entire weighing area. This module also stores the conveyor bandwidth parameter separately, binding the bandwidth information with the identifiers of all modular weighing units under the corresponding machine and archiving it.
[0037] The purpose of this step is to fully record the overall structural specifications of the entire weighing area, distinguish weighing acquisition intervals of different lengths by the effective weighing length, and distinguish different conveyor equipment specifications by the bandwidth parameter, so that the model can distinguish the weighing data generated under different weighing area sizes and different belt widths.
[0038] This module synchronously archives and stores the auxiliary hardware parameters of each modular weighing unit, including unit installation location information, load cell calibration parameters, and synchronously stores the identifier and uploaded valid unit status information. All auxiliary parameters and status data are associated with the corresponding unit identifier and stored. When a modular weighing unit malfunctions or becomes unusable, the valid status field bound to that unit identifier within the module will be updated synchronously, ensuring that parameters and unit operating status are stored in sync.
[0039] The purpose of this step is to retain complete information on the unit's installation location, calibration, and operating conditions. When the training model retrieves data, it can automatically identify failed units through the stored valid state parameters, directly filter out invalid weighing data produced by faulty units, and rely on calibration parameters to ensure the traceability of measurement data.
[0040] In this module, the complete set of hardware structural parameters bound to the unit identifier will be synchronously output as a fixed data source to the structural label information generation process. All stored binding parameters, together with the identifier and transmitted weighing signals, unit identifiers, and other data, will be integrated to generate standardized structural label information. All structural fields contained in the structural label information directly retrieve the stored hardware parameters bound to the unit identifier.
[0041] The purpose of this step is to provide a standardized and unified source of parameters for structural label information, avoid chaotic or missing parameter input, ensure that the structural parameters of the label associated with each weighing signal are accurate and complete, and enable the label information to truly reflect the hardware structural conditions of the corresponding weighing unit.
[0042] In the process of constructing measurement time-series samples and inputting them into the measurement pre-training model, the system retrieves a complete set of pre-bound hardware structural parameters from this module based on the unit identifier carried by the weighing signal. The retrieved structural parameters are then combined with the real-time weighing signal to form a complete measurement time-series sample for the model to use. The measurement pre-training model can rely on the retrieved parameters to identify the hardware structural conditions corresponding to the data, distinguishing whether data differences stem from changes in equipment structural specifications or material / equipment operating conditions.
[0043] The purpose of this step is to achieve automated matching and retrieval of weighing signals and hardware structural parameters, eliminating the need for manual input of structural information, improving the consistency and reusability of data from multiple electronic belt scales, reducing the data processing cost of model training, and enhancing the model's generalization learning ability across working conditions and devices.
[0044] IV. About the structure tag generation module.
[0045] This module integrates three core data sources and generates standardized structural label information. The three data sources are: the weighing signal output from the modular weighing unit, the bound unit identifier or channel identifier, and all hardware structural parameters uniformly archived by the structural parameter storage module and bound to the unit identifier. Finally, a unique tag is added to each continuous segment of weighing time-series data, fully annotating the corresponding weighing hardware structural conditions and real-time operating status information of the unit.
[0046] This model first completes the matching and fusion of weighing signals with unit identifiers and channel identifiers. Modular weighing units output exclusive weighing signals belonging only to their own independent weighing area. Each acquired weighing signal is pre-associated with the unit identifier or channel identifier of the corresponding modular weighing unit. After reading the real-time transmitted weighing time-series data, this module automatically matches the unique identifier code inherent in the signal, establishing a one-to-one correspondence between a single segment of time-series weighing data and the corresponding weighing unit.
[0047] The purpose of this step is to accurately pinpoint the physical source of each weighing time series data, distinguish the load data produced by different units in a multi-module combined weighing scenario, avoid the inability to trace the acquisition module after multi-channel signal aliasing, and ensure that the tag can accurately match the true source of the data.
[0048] This module retrieves the complete set of hardware structural parameters bound to the matched unit identifier from the structural parameter storage module. Key parameters include the preset idler spacing of the modular weighing unit, a fixed structural parameter predetermined at the factory, ensuring uniform spacing for units of the same specification. Simultaneously, this module retrieves the effective weighing length calculated from the number of units, the preset idler spacing, and the unit arrangement, and reads the bandwidth parameters of the accompanying conveyor. All of these hardware dimensional parameters are then integrated into the label generation process.
[0049] The purpose of this step is to solidify the basic structural conditions of the weighing area corresponding to the weighing data into the label, so that the model can distinguish the measurement data generated by equipment with different idler spacing, different weighing area length, and different belt width through the label, thereby reducing the difficulty of sample alignment during cross-equipment training.
[0050] Based on the hardware parameters, this module simultaneously retrieves the associated operating conditions and calibration parameters bound to the same unit identifier, covering the installation location of the modular weighing unit, the unit's real-time effective operating status, and the weighing sensor calibration parameters. When the corresponding weighing unit experiences abnormal conditions such as malfunction, shutdown, or disuse, the retrieved effective status fields will be simultaneously marked with an abnormality identifier and written to a tag, thus completely retaining the unit's operating condition information throughout its entire lifecycle.
[0051] The purpose of this step is to distinguish the weighing time series data produced by the normally operating unit and the faulty and disabled unit in the label. During the model training process, invalid and abnormal data can be automatically filtered out by relying on the status field in the label. At the same time, the calibration parameters are used to retain the metrological calibration benchmark to ensure that the data metrological benchmark is traceable.
[0052] This module integrates the completed weighing signals, unit identifiers, all hardware structural parameters, and unit operating condition parameters to generate standardized structural label information. The standardized structural label fully includes all fields such as preset idler spacing, number of module combinations, effective weighing length, conveyor bandwidth, unit installation position, unit effective status, signal channel correspondence, and sensor calibration parameters.
[0053] The purpose of this step is to integrate discrete signals, identifiers, and parameters into unified and standardized data labels, eliminating problems such as inconsistent parameter recording formats and missing or chaotic information, and forming standardized annotation data with unified format and complete fields.
[0054] This module binds and pairs the generated standardized structural label information with the original weighing time series data, so that each segment of weighing time series data comes with a unique and complete structural label. After binding, the weighing data is no longer an isolated load or flow value without any explanation, but a complete dataset that simultaneously carries a clear module source, hardware structure conditions, and real-time operating conditions.
[0055] The purpose of this step is to give each time series measurement data a complete structural background, so as to solve the problem that when multiple electronic belt scales are used together for training, the model cannot distinguish whether the data differences are due to material fluctuations, equipment failures or different weighing structure specifications.
[0056] The weighing time-series data, after being bound with standardized structural labels, will collectively form a metrological time-series sample. This entire sample set will directly serve as the input data source for the metrological pre-training model. When the metrological pre-training model reads the sample, it can simultaneously parse the structural label information attached to the sample to identify the roller spacing, effective weighing length, bandwidth specifications, and unit working status corresponding to each segment of weighing data.
[0057] The purpose of this step is to significantly improve the consistency and reusability of data collected from multiple sites and devices, reduce the workload of sample alignment during model training, facilitate the model to learn general measurement rules from massive amounts of cross-device data, and improve the model's generalization and adaptation capabilities to various weighing conditions.
[0058] V. Regarding the time series sample construction module.
[0059] This module receives two pre-amplifier outputs: one is the time-series weighing signal continuously acquired and output by the modular weighing unit, and the other is standardized structural label information generated by the structural label generation module. The module pairs and encapsulates the corresponding individual time-series weighing signals with their respective structural labels to generate a single basic metrological time-series sample.
[0060] After pairing and encapsulating a single time-series data entry with a structural tag, this module performs multi-device data alignment based on the various structural fields recorded within the standardized structural tag. The tag contains uniformly standardized hardware parameter fields such as preset idler spacing, effective weighing length, bandwidth, and module quantity. The module reads the tag parameter information from all samples, using the preset idler spacing as a unified benchmark to align and calibrate the measurement data from different conveyors and modular weighing units, eliminating data baseline deviations caused by inconsistencies in the original conveyor idler arrangement and weighing area dimensions.
[0061] The purpose of this step is to address the problem of significant differences in the structure of different field equipment and the difficulty in unifying training data from multiple devices in existing technologies. It reduces the difficulty of sample alignment during the training of the metrology pre-training model and facilitates the model in extracting general metrology rules from multi-source data.
[0062] This module groups all samples based on the multi-dimensional information carried by standardized structural tags. The grouping criteria include the conveyor bandwidth, the number of modular weighing unit combinations, the effective weighing length, the effective working status of the unit, and the sensor calibration parameters recorded on the tags. This module groups samples with consistent idler spacing, bandwidth, and weighing zone length into the same group. Samples from units marked as faulty or out of service are grouped separately as abnormal data, while samples collected from units operating normally are grouped into the normal measurement group.
[0063] The purpose of this step is to automatically classify and distinguish samples with different hardware specifications and operating conditions. During model training, corresponding grouped samples can be selected as needed for training. At the same time, invalid fault data is quickly isolated to avoid abnormal samples interfering with the model's learning of normal measurement features.
[0064] This module integrates all grouped samples to generate a complete training dataset with a unified overall format that clearly distinguishes various hardware configurations and operating conditions. Each measurement time series sample in the entire training dataset has a built-in matching structural label, which fully records all structural and operating condition information corresponding to the sample, such as the preset idler spacing, number of modules, effective weighing length, bandwidth, unit installation position, effective unit status, and calibration parameters.
[0065] The purpose of this step is to unify the data format standards of multiple electronic belt scales and data collected from multiple sites, break down the barriers of independent and incompatible data from different devices, and form a standardized dataset for use by the pre-trained metrology model.
[0066] VI. Regarding the quantitative pre-trained model adaptation module.
[0067] The function of this module is to parse the standardized structural label features attached to the metering time series sample, read the complete fields of the structural label bound to the sample encapsulation, and extract all structural and working condition feature information contained in the label, such as the preset idler spacing, number of modules, effective weighing length, bandwidth, unit installation position, effective status of unit, channel correspondence, calibration parameters, etc., to complete the decomposition, identification and classification of the multi-dimensional features of the label.
[0068] The metrology pre-training model adaptation module utilizes the analyzed label structural features to specifically eliminate interference caused by differences in the structure of different weighing units during model training. This module differentiates and corrects sample data through standardized structural parameters within the labels, and relies on a unified preset idler roller spacing benchmark, quantifiable effective weighing length, and bandwidth parameters to eliminate data baseline deviations caused by hardware structure, filtering out data difference interference caused by different modular weighing unit configurations.
[0069] The purpose of this step is to address the shortcomings of multi-device mixed training, which leads to high difficulty in sample alignment and difficulty in model extraction of general patterns due to the superposition of a large number of scale body structural differences, and to avoid the problem that hardware specification differences mask the changing characteristics of real measurement data.
[0070] The metrology pre-training model adaptation module relies on standardized sample data after removing interference to support the metrology pre-training model in learning metrology rules applicable to all equipment. Samples, after module calibration, uniformly possess clear and quantifiable structural annotations. Equipment data from different bandwidths, different module combinations, and different field conditions can be mixed and used in training. The model can isolate hardware-specific differences and focus on learning common metrology variation characteristics under various weighing scenarios, such as material load, belt speed, and sensor response.
[0071] The purpose of this step is to fully leverage the value of long-term operating data from multiple electronic belt scales, improve the consistency and reusability of cross-device data collection, and allow the model to break free from the limitations of a single device's specific operating conditions, forming a universally applicable measurement and identification capability.
[0072] This module synchronously accesses two types of operating condition characteristics recorded within the tags: unit validity status and calibration parameters, to assist in metering fault identification and metering accuracy status analysis. When the modular weighing unit corresponding to the structural tag is in an abnormal or disabled state, the module locates the time-series data corresponding to this type of fault sample, and simultaneously compares it with calibration parameters to determine accuracy anomalies such as sensor drift and metering zero-point offset. It also uses structural parameters such as preset idler roller spacing and effective weighing length to distinguish between hardware assembly problems and equipment malfunctions. This module ultimately generates an equipment metering status analysis report, which integrates three categories of content: a fault sample location list, metering accuracy deviation values, and fault cause classification information.
[0073] The purpose of this step is to distinguish between faults and accuracy issues, avoid misjudging inherent structural differences of the scale body as measurement faults, improve the accuracy of equipment status analysis results, and provide data support for on-site equipment operation and maintenance, calibration and verification.
[0074] The above are merely preferred embodiments of the invention and are not intended to limit the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.
Claims
1. A modular weighing system suitable for a pre-trained measurement model of an electronic belt scale, characterized in that, Includes a mechanical weighing module: a weighing structure equipped with a preset roller spacing, which collects the unit material load signal and supports multi-unit splicing to adjust the effective weighing length; Identification and signal transmission module: used to assign a unique identifier to each weighing unit and independently transmit each weighing signal, and synchronously upload the unit's operating status; Structural parameter storage module: used to uniformly store all hardware structural parameters and bind them to unit identifiers; Structural label generation module: integrates weighing signals, unit identifiers and structural parameters to generate standardized labels, and marks each segment of weighing time series data with corresponding hardware structure and operating condition information; Measurement Time Series Sample Construction Module: Used to pair and encapsulate time series weighing signals with structural labels, complete multi-device data alignment and grouping based on standardized labels, and generate a training dataset with a unified format that can distinguish hardware operating conditions; Metrology pre-training model adaptation module: used to parse the structural label features attached to the samples, eliminate the interference of scale body structural differences on training, support the model to learn general metrology rules and assist in the analysis of metrology faults and accuracy status.
2. The modular weighing system for the pre-training model of electronic belt scales according to claim 1, characterized in that, The data processing steps in the identification and signal transmission module are as follows: Match the corresponding physical unit identifier to the original weighing signal, assign channel identifiers, establish the mapping relationship between weighing signal, unit identifier, and channel identifier, and generate the split original weighing data stream; The valid status information of the unit is attached and bound to the original weighing data of the corresponding channel, and then merged according to the time sequence to generate a weighing time sequence stream with identifier and unit status. The branch weighing time stream is isolated and split according to the channel identifier, and abnormal segments of faulty units are removed to generate branch channel time weighing data streams; The data format and timing sampling interval of the sub-channel timing weighing data stream are uniformly converted, and a timestamp timing index is added. After encapsulation, a standardized labeled channel weighing timing dataset is output.
3. The modular weighing system for the pre-training model of electronic belt scales according to claim 2, characterized in that, The data processing steps in the structural parameter storage module are as follows: Each identifier in the standardized labeled channel weighing time series dataset is analyzed, and the unique identifier field is extracted and used as a unit parameter retrieval query instruction. Based on the unit parameter retrieval query instruction, match the corresponding unit identifier in the database, read the corresponding unit hardware structure parameters, and generate a unit structure parameter set; Based on all the unit identifiers of the electronic belt scale, the unit structural parameter set is bound and spliced with the assembly parameters of the electronic belt scale to generate the structural configuration parameter set of the electronic belt scale; Based on the unit identifier, the calibration parameters in the unit calibration parameter library are matched, and the calibration parameters are added to the structural configuration parameter set of the electronic belt scale, and a complete unit structural parameter data package is output.
4. The modular weighing system for the pre-training model of electronic belt scales according to claim 3, characterized in that, The data processing steps in the structure tag generation module are as follows: Extract the common unit identifier / channel identifier from the standardized labeled channel weighing time series dataset and the complete unit structure parameter data package as the association primary key, and perform time series alignment according to timestamp to establish the mapping relationship between each weighing time series record and the corresponding unit structure parameter, and generate a signal-parameter associated time series data package; Extract all structural dimension information from the signal-parameter associated time-series data packet. The all structural dimension information includes preset idler spacing, number of modules, effective weighing length, bandwidth, unit installation position, effective unit status, calibration parameters, and channel correspondence. Then, organize all structural dimension information into a unified structured field group to generate a standardized structural information field set. Extract the weighing signal and unit identifier / channel identifier from the signal-parameter associated time series data packet, and concatenate them with the standardized structural information field set to integrate the load time series data, module traceability ID, and equipment structural parameters into a single data record, generating a complete time series record set with traceability and structural fields; According to the document definition rules, the complete time series record set with traceability and structure fields is encapsulated into an auxiliary tag data package and bound to the corresponding weighing time series record to generate structure tag information; The standardized weighing time series dataset with identification on a single channel is paired and encapsulated with the structural label information, and the data storage format and sampling time series index are unified. The label-signal combination corresponding to invalid fault units is removed, and the weighing time series data packet with structural label is output.
5. The modular weighing system for the pre-training model of electronic belt scales according to claim 4, characterized in that, The data processing steps in the metrology time series sample construction module are as follows: Based on a fixed time window length, the weighing time data packet with structure tags is segmented by sliding slices, and the structure tag information bound to each segment is retained to generate segmented time-tag segment groups; Based on the structural configuration of the same equipment, the segmented time-tag fragment groups are grouped and classified to generate a set of time-tag subsets grouped by structure; Read the unit valid status field in the set of time tag subsets grouped by structure, filter out the time segments corresponding to units marked as faulty or disabled, and remove abnormal segments with signal distortion or missing timing, to generate a cleaned valid grouped time tag subset; Read the structural tag information in the effective grouped time sequence tag subset after cleaning, determine the same structure group based on the unified preset idler spacing in the structural tag information, perform sampling frequency correction and load dimension normalization conversion for the time sequence signals from different electronic belt scales in the group, and generate an aligned and normalized time sequence tag subset. Traverse all time segments and corresponding structural label information in the aligned and normalized time tag subset, bind and encapsulate each weighing time signal and its corresponding structural label information, and combine them one-to-one into a single independent training unit to generate a measurement time series sample. All the aforementioned time-series econometric samples are integrated, a unified storage index is established, and training / backup partitions are created to output a standardized and complete econometric training dataset.
6. The modular weighing system for the pre-training model of electronic belt scales according to claim 5, characterized in that, The data processing steps in the quantitative pre-trained model adaptation module are as follows: Extract the structural label information of the measurement time series sample, parse all standardized feature fields in the label, and simultaneously extract the original measurement data of the time series weighing signal. Establish the mapping relationship between a single segment of time series weighing signal and the corresponding complete set of label feature fields to generate a time series weighing dataset with label feature mapping. All time-series weighing signals in the labeled feature mapping time-series weighing dataset are subjected to hierarchical correction, and the data difference components caused by the inherent structure of the weighing body are removed to generate a standardized metrological dataset that eliminates the interference of weighing body structural differences. The standardized metrology dataset is input into the metrology pre-training model. The metrology pre-training model relies on the time-series weighing data that has been uniformly corrected within the dataset to extract general metrology features and generate a metrology model that has completed the learning of general metrology rules. The standardized metrological dataset, which eliminates interference from structural differences in the weighing platform, is input into the metrological model that has completed the learning of general metrological laws to determine the equipment status and output an equipment metrological status analysis report.
