A system and method for manufacturing quality stratified analysis and abnormality identification of an electric energy meter

Abstract

The application discloses a kind of electric energy meter manufacturing quality stratified analysis and abnormal identification system and method.The system collects the error value of each electric energy meter in multiple working conditions, the error change quantity before and after aging, CDI key test quantity and manufacturing trace field, and only with the electric energy meter in the qualified domain of enterprise internal control qualified limit as the abnormal identification object;According to the physical order of current ratio, power factor, positive and negative energy and the detection stage before and after aging, a multi-working-condition error curve is constructed;Based on the qualified domain sample, generate the qualified domain reference curve and quantile envelope;Calculate the overall residual quantity, adjacent working condition jump quantity, second-order bending abnormal quantity, envelope edge rate, pre- and post-aging drift and positive and negative asymmetry;Finally output table position reinspection, equipment calibration, process verification or component batch isolation instruction.The present application can identify all single-point qualified but curve form abnormal potential reliability risk products, realize the pre-warning and stratified disposal of electric energy meter manufacturing quality.

Description

Technical Field

[0001] This invention relates to the field of electricity meter manufacturing quality analysis technology, and in particular to an electricity meter manufacturing quality stratification analysis and anomaly identification system and method. Background Technology

[0002] Against the backdrop of the ongoing construction of smart grids and the continuous promotion of electricity consumption information collection systems, the manufacturing quality control of electricity meters, as legally mandated measuring instruments for trade settlement, has gradually shifted from simply focusing on the accuracy of metering at the factory to a comprehensive control of batch consistency, process stability, and reliability throughout the entire life cycle. In the large-scale automated production process of electricity meters, processes such as FCT functional circuit testing, high-temperature aging, factory verification, and packaging generate a large amount of testing data. Among these, multi-condition error data is the core data for evaluating the metering performance and manufacturing consistency of electricity meters.

[0003] Existing electricity meter calibration systems typically perform error tests point-by-point according to preset operating conditions. For example, relative error values ​​are obtained under different current ratios, different power factors, forward or reverse power, and before and after aging testing. Each error point is then compared with the national standard limit or the company's internal control compliance limit. If all operating conditions are within the compliance limit, the system determines the electricity meter to be qualified; if any operating condition exceeds the compliance limit, it is determined to be unqualified and transferred to the repair or scrapping process.

[0004] The currently widely adopted "discrete point determination method based on static threshold" and its supporting "automatic verification pipeline system" are described below:

[0005] (1) Existing technical methods and steps (taking the core error detection as an example):

[0006] Step 1: Operating condition traversal test. The calibration device controls the standard source to output a series of discrete specific operating conditions (e.g., voltage 220V, current from 0.05Ib, 0.1Ib, 1.0Ib to Imax, power factors of 1.0, 0.5L, 0.8C, etc.).

[0007] Step 2: Single-point error acquisition. At each operating point, the pulse signal of the tested energy meter is collected and compared with the pulse signal of the standard meter to calculate the relative error value at that operating point.

[0008] Step 3: Static Threshold Comparison. The system presets a fixed error limit range (e.g., national standard or enterprise internal control limit). The system independently compares the error value of each operating point with this limit.

[0009] Step 4: Logical Judgment and Handling:

[0010] 1) If the error values ​​at all working points are within the limit range, the product is deemed "qualified" and transferred to the packaging process.

[0011] 2) If the error at any operating point exceeds the limit, the product is deemed "unqualified" and transferred to the rework area or scrapped.

[0012] (2) Structure of the existing device: The existing system usually consists of a programmable power source, a standard energy meter, an error calculator (DSP / FPGA), a multi-channel error acquisition terminal, and a host computer management system (MTS / MES). Connection relationship: The host computer sends instructions to the programmable power source to output voltage and current, driving the meter under test and the standard meter; the error calculator simultaneously collects the pulses of both and calculates the error; finally, the data is uploaded to the host computer database for storage and comparison with the above threshold.

[0013] While the static threshold determination method based on discrete operating points can effectively intercept obviously non-conforming products, it can only answer whether a product exceeds the limit and cannot characterize the subtle changes in the overall trend, consistency, continuity, and aging stability of the multi-condition error curve. For energy meters where all single-point errors are within the enterprise's internal control acceptable limits, if their error curves exhibit abnormal shapes such as overall shift, local jumps, nonlinear bending, proximity to the acceptable boundary, increased drift before and after aging, or asymmetry between positive and negative directions, traditional systems usually still release them as ordinary qualified products, failing to identify the potential reliability risks.

[0014] In actual manufacturing processes, unstable meter clamping, output drift of calibration equipment, fluctuations in the controlled state of the high-temperature aging process, and batch differences in key components such as sampling resistors or crystal oscillators can all cause morphological changes with manufacturing significance in the multi-condition error curves within the acceptable range. These changes may not reach the level of single-point deviation, but they may correspond to metering drift or reliability degradation during subsequent use. Existing systems lack a mechanism for constructing benchmark curves and quantile envelopes for acceptable samples from the same batch or process window, and also lack the means to analyze the residual morphology according to the physical sequence of the electricity meter's metering conditions. Therefore, it is difficult to identify and handle electricity meters that are "acceptable at a single point but have abnormal curve morphology."

[0015] Meanwhile, existing verification systems typically output results at the level of product conformity or non-conformity, lacking the logic to further map the risks associated with curve patterns within the conformity range to the manufacturing handling logic at the surface, equipment, process, and component levels. When slight fluctuations occur in the same surface, the same verification equipment, the same aging process window, or the same batch of critical components, production personnel can often only conduct post-conformity investigations after the batch non-conformity occurs, making it difficult to take timely and targeted measures such as surface re-inspection, equipment calibration, process verification, or component batch isolation. Summary of the Invention

[0016] The purpose of this invention is to address the shortcomings of existing technologies by proposing a quality layering analysis and anomaly identification system and method for electricity meter manufacturing. This system aims to solve the problem that traditional single-point static threshold verification methods cannot identify curve morphology anomalies and potential reliability risks hidden in all qualified data. It also addresses the lack of risk identification within the qualified domain and the lack of manufacturing disposal basis for the four layers of meter location, equipment, process, and components in the production process.

[0017] The technical solution of this invention does not take the general electrical equipment condition monitoring process as the main technical route. Instead, it uses the multi-condition error curves of all single-point qualified energy meters in the same batch or within the same process window as qualified domain samples, constructs qualified domain benchmark curves and quantile envelopes, calculates the physical residuals of the operating conditions according to the physical order of the energy meter's metering conditions, and forms curve shape risk values ​​accordingly. These values ​​are used to identify products with potential reliability risks that are qualified at all single points but have abnormal curve shapes.

[0018] To achieve the above objectives, the present invention provides a system for stratified analysis and anomaly identification of electricity meter manufacturing quality, the system comprising:

[0019] The qualified domain data acquisition and object screening module is used to collect the multi-condition error values, error changes before and after aging, CDI key test quantities, and manufacturing traceability fields of each energy meter under the same preset operating condition sequence. Only energy meters with all items within the enterprise's internal control qualified limits are identified as qualified domain anomaly identification objects. Among them, energy meters with error values ​​or key test quantities exceeding the qualified limits at any operating condition point are processed according to the regular non-conformance process and are not included in the qualified domain curve shape risk identification process.

[0020] The operating condition serialization and error curve construction module is used to sort multiple operating condition error values ​​of the same energy meter according to the preset physical order of current ratio, power factor, forward and reverse energy, phase, and testing stages before and after aging, and form a multi-operating condition error curve.

[0021] The qualified domain baseline curve and quantile envelope construction module is used to calculate the baseline median curve, upper quantile envelope and lower quantile envelope corresponding to each preset working point based on qualified domain samples in the same batch or the same process window, forming a qualified domain curve reference to characterize the normal curve shape range of the current batch or the current process window.

[0022] The working condition physical residual calculation module is used to compare the multi-working condition error curves of the energy meter to be evaluated with the qualified domain reference curve and quantile envelope, and calculate the overall residual, adjacent working condition jump variable, second-order bending anomaly, envelope edge contact rate, drift before and after aging and forward and reverse asymmetry according to the working condition physical order.

[0023] The curve shape risk identification module is used to generate curve shape risk values ​​within the qualified domain based on at least three of the following: overall residual amount, adjacent operating condition jump variable, second-order bending anomaly amount, envelope edge contact rate, drift amount before and after aging, and forward and reverse asymmetry amount. It also identifies potential reliability risk products that are qualified at a single point but have abnormal curve shape based on the risk upper limit formed by qualified domain samples of the same batch or the same process window.

[0024] The four-layer manufacturing handling module is used to generate handling instructions for table re-inspection, equipment calibration, process verification, or component batch isolation based on the contribution type of each risk indicator in the curve shape risk value, combined with the table number, equipment number, process record, and component batch information.

[0025] Furthermore, the multi-condition error values ​​include at least the error values ​​of the energy meter under different current ratios, different power factors, positive active power, reverse active power, and different phases; the error change before and after aging includes the error difference or error change rate before and after high-temperature aging under the same condition; the manufacturing traceability fields include production barcodes, batch numbers, meter position numbers, calibration equipment numbers, process timestamps, tooling fixture numbers, key component batch numbers, and environmental temperature and humidity records.

[0026] Furthermore, the CDI key test quantities serve as auxiliary data acquisition fields for qualified domain object screening and four-layer manufacturing processing, including FCT key point voltage, sampling loop resistance, daily timing results, communication response time, voltage and current stability of calibration equipment, meter contact status, and process test quantities related to the manufacturing quality of the electricity meter; the CDI key test quantities are only used as auxiliary data for qualified domain screening and manufacturing processing, and are not used as independent anomaly identification algorithms.

[0027] Furthermore, the qualified domain samples are energy meter samples that are all within the enterprise's internal control qualified limits within the same batch, the same verification equipment group, the same process window, or the same production time window; when the process parameters, verification equipment status, environmental status, or production batch are changed, the corresponding qualified domain benchmark curve and quantile envelope are reconstructed.

[0028] Furthermore, the qualified domain reference curve is formed by the median, truncated mean, or robust mean of each operating point; the upper quantile envelope and lower quantile envelope are formed by the upper and lower quantiles of each operating point, and can also be extended based on the quantiles by combining the interquartile range or median absolute deviation to characterize the statistical envelope range of the normal qualified curve within the same batch or the same process window.

[0029] Furthermore, the physical residual under operating conditions is not the distance between disordered state data, but a curve residual formed based on the current multiplier increase relationship, power factor switching relationship, forward and reverse energy correspondence relationship and the correspondence relationship before and after aging; it is used to characterize the overall position, local continuity, curvature change, boundary proximity, aging stability and forward and reverse symmetry of the error curve of the energy meter to be evaluated relative to the qualified domain reference curve.

[0030] Furthermore, the four-layer manufacturing processing module includes a surface position layer processing unit, an equipment layer processing unit, a process layer processing unit, and a component layer processing unit; the surface position layer processing unit is used to generate surface position re-inspection, surface clamp inspection, or single surface position pause instructions; the equipment layer processing unit is used to generate calibration instructions for verification equipment, standard source inspection, or equipment shutdown verification instructions; the process layer processing unit is used to generate aging process verification, process parameter review, or tightened sampling inspection instructions; and the component layer processing unit is used to generate batch isolation instructions for key components, rework investigation instructions, or related batch retesting instructions.

[0031] On the other hand, the present invention also provides a method for quality stratification analysis and anomaly identification in the manufacture of electricity meters, comprising the following steps:

[0032] S1, collect the multi-condition error values, error changes before and after aging, CDI key test quantities and manufacturing traceability fields of each energy meter under the same preset operating condition sequence, and judge the pass / fail limits of each operating condition error value and key test quantity.

[0033] S2, only include energy meters whose preset working condition error values ​​and key test quantities are within the enterprise's internal control qualification limits into the qualification domain abnormal identification object set, and energy meters whose any indicator exceeds the limit are handled according to the conventional non-qualification process;

[0034] S3, according to the preset physical sequence of current ratio, power factor, forward and reverse power, phase and aging stages, constructs multiple error values ​​of the qualified domain abnormal identification object into a multi-condition error curve.

[0035] S4 generates the qualified region baseline curve, upper quantile envelope, and lower quantile envelope based on qualified region samples within the same batch or process window.

[0036] S5 compares the multi-condition error curves of the energy meter to be evaluated with the qualified domain reference curve and quantile envelope, and calculates the overall residual, adjacent condition jump variable, second-order bending anomaly, envelope edge contact rate, drift before and after aging and forward and reverse asymmetry.

[0037] S6. Generate curve shape risk values ​​within the qualified domain based on the above risk indicators, and identify potential reliability risk products that are qualified at a single point but have abnormal curve shapes.

[0038] S7 determines the disposal objects at the tabletop, equipment, process, or component level based on the contribution type of the curve shape risk indicator and the manufacturing traceability field, and outputs tabletop re-inspection, equipment calibration, process verification, or component batch isolation instructions.

[0039] The beneficial effects of this invention include:

[0040] (1) Under the premise that all single-point errors meet the enterprise's internal control qualification limits, it can further identify abnormal curve morphology within the qualification range, and make explicit potential reliability risks such as overall offset, local jump, nonlinear bending, boundary closeness, aging drift and forward and reverse asymmetry that are difficult to detect by traditional verification systems.

[0041] (2) By establishing the curve reference under the current manufacturing conditions through the qualified domain benchmark curve and quantile envelope of the same batch or the same process window, it no longer relies on the generalization processing of disordered multidimensional state data, and can more directly reflect the physical properties and manufacturing consistency of the multi-condition error curve of the energy meter.

[0042] (3) By analyzing the correlation between current ratio, power factor, forward and reverse electrical energy and aging stages through the physical residual of the operating conditions, we can avoid splitting multiple operating points into unrelated discrete indicators and improve the sensitivity of identifying hidden risks in the qualified domain.

[0043] (4) By using four layers of manufacturing processes—the surface layer, equipment layer, process layer, and component layer—the risk of curve morphology is linked to the re-inspection, calibration, verification, and isolation measures on the production site, which is conducive to preventive measures and process optimization before batch problems occur. Attached Figure Description

[0044] 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.

[0045] Figure 1 This is a schematic diagram illustrating the risk of error curve morphology under multiple operating conditions within the qualified domain;

[0046] Figure 2 This is an overall structural block diagram of a system for stratified analysis and anomaly identification of manufacturing quality of electricity meters according to the present invention;

[0047] Figure 3 This is the main flowchart of a method for stratified analysis and anomaly identification of manufacturing quality of electricity meters according to the present invention;

[0048] Figure 4This is a schematic diagram of the construction of the benchmark curve and quantile envelope of the qualified region for the same batch or the same process window;

[0049] Figure 5 This is a schematic diagram of a four-layer manufacturing process based on the risk profile of curves: surface layer, equipment layer, process layer, and component layer. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are merely illustrative and not intended to limit the invention.

[0051] Example 1: Overall System Architecture

[0052] like Figure 2 As shown, this embodiment provides a quality stratification analysis and anomaly identification system for electricity meter manufacturing. The system includes a qualified domain data acquisition and object screening module, a working condition serialization and error curve construction module, a qualified domain benchmark curve and quantile envelope construction module, a working condition physical residual calculation module, a curve morphology risk identification module, and a four-layer manufacturing handling module.

[0053] The qualified domain data acquisition and object screening module collects data from the FCT test bench, high-temperature aging chamber, automated calibration line, meter position detection unit, calibration equipment operation logs, and production management system. The collected data includes multi-condition error values ​​for each energy meter under the same preset operating condition sequence, error changes before and after aging, key CDI test values, and manufacturing traceability fields. Manufacturing traceability fields include at least the production barcode, batch number, meter position number, calibration equipment number, process timestamp, and batch number of key components.

[0054] This module first performs an internal control compliance check on all data points of each electricity meter. If any preset operating condition error value or key test quantity exceeds the compliance limit, the electricity meter is processed according to the standard non-compliance procedure. If all preset operating condition error values ​​and key test quantities are within the internal control compliance limits, the electricity meter is identified as an anomaly in the compliance domain and enters the subsequent curve morphology risk identification process. Through this precondition, the system's identification targets are limited to electricity meters that have been judged as qualified by traditional verification systems but may have hidden quality risks.

[0055] The operating condition serialization and error curve construction module sorts the error values ​​according to the physical relationships of the operating conditions. These physical relationships include the relationship of current ratio changing from small to large, the power factor switching relationship, the correspondence between forward and reverse electrical energy, the phase correspondence, and the correspondence between the testing stages before and after aging. Through unified sorting, multiple discrete error points of the same energy meter are organized into multi-operating condition error curves with physical meaning.

[0056] The qualified domain baseline curve and quantile envelope construction module selects qualified domain samples from the same batch or within the same process window, and calculates the baseline median, upper quantile, and lower quantile values ​​at each preset operating point, thereby forming the qualified domain baseline curve, upper quantile envelope, and lower quantile envelope. The same process window can be determined by the same calibration equipment group, the same aging furnace batch, the same production time window, or the same combination of process parameters.

[0057] The operating condition physical residual calculation module compares the multi-operating condition error curves of the energy meter to be evaluated with the qualified domain reference curve point by point to obtain the residual curve; at the same time, it judges whether the curve is touching the edge, locally exceeding the normal qualified curve envelope, or continuously biased to one side based on the upper quantile envelope and lower quantile envelope.

[0058] The curve morphology risk identification module calculates the overall residual, adjacent operating condition jump variables, second-order bending anomaly, envelope edge contact rate, drift before and after aging, and forward and reverse asymmetry based on the residual curve and quantile envelope relationship, and synthesizes them into a curve morphology risk value within the qualified domain. This risk value is used to characterize whether the error curve of the energy meter under evaluation still has manufacturing quality risks even if all individual points are qualified.

[0059] The four-layer manufacturing handling module generates handling instructions based on the contribution type of the curve shape risk indicator and the manufacturing traceability field. When the risk is mainly manifested as curve jumps or envelope edge contact under the same table position, it outputs table position re-inspection or table position pause instructions; when the risk is mainly manifested as the overall residual of multiple tables under the same equipment increasing in the same direction, it outputs equipment calibration instructions; when the risk is mainly manifested as drift before and after aging concentrated in the same process window, it outputs process verification instructions; when the risk is strongly correlated with the batch of key components or the key test quantity of FCT, it outputs component batch isolation or rework investigation instructions.

[0060] Example 2: Main Flow of the Method for Identifying Curve Shape Risks within the Qualified Domain

[0061] like Figure 3 As shown in the figure, this embodiment provides a method for stratified analysis and anomaly identification of electricity meter manufacturing quality:

[0062] Step one: The on-site testing system collects the multi-condition error values, error changes before and after aging, key CDI test values, and manufacturing traceability fields for each energy meter under the same preset operating condition sequence. The preset operating condition sequence is determined by the enterprise's verification procedures and remains consistent within the same batch or the same process window.

[0063] Step two involves screening the collected data for compliance. The system checks each item on the energy meter under evaluation to determine whether its multi-condition error values, error changes before and after aging, and key test quantities are within the company's internal control compliance limits. Only when all items are qualified is the energy meter included in the compliance domain for anomaly identification; otherwise, it is processed according to the standard non-conforming product procedure.

[0064] Step 3: Construct multi-condition error curves for the energy meter to be evaluated according to the physical sequence of operating conditions. This step transforms the originally discrete error points into a continuous curve structure based on current, power factor, direction, and aging stage, providing input for subsequent morphological risk analysis.

[0065] Step four: Select qualified samples from the same batch or within the same process window to construct the qualified region baseline curve and quantile envelope. If there are significant changes in the production batch, calibration equipment, aging process parameters, or environmental conditions, the system will redetermine the qualified region sample range and update the curve reference.

[0066] Step 5: Calculate the physical residual of the energy meter to be evaluated relative to the benchmark curve of the qualified domain, and further calculate the overall residual, adjacent operating condition jump variable, second-order bending anomaly, envelope edge contact rate, drift before and after aging, and forward and reverse asymmetry.

[0067] Step six: Generate a curve shape risk value within the qualified range based on the above indicators. If this risk value is higher than the risk limit formed by qualified samples in the same batch or process window, the energy meter is identified as a product with potential reliability risks due to single-point qualification but abnormal curve shape.

[0068] Step 7: Based on the contribution type of the risk indicator of the curve shape and the manufacturing traceability field, perform four-level manufacturing processing, output the instructions for table re-inspection, equipment calibration, process verification or component batch isolation, and write the processing results back to the quality database for subsequent identification of similar risks.

[0069] Example 3: Construction of Qualified Domain Reference Curve and Quantile Envelope

[0070] like Figure 4 As shown, in one embodiment, the system uses energy meters within the same batch or process window whose entire range is within the enterprise's internal control acceptable limits as the acceptable sample set. Assume each electricity meter has... Error curves are generated under preset operating conditions. ,in Indicates the sample number. Indicates the operating condition number.

[0071] For the For each operating condition point, the system calculates all samples in the qualified domain sample set. The median is used as the reference value for the baseline curve at that operating point. Simultaneously calculate the upper quantile. and lower quantile This forms an upper quantile envelope and a lower quantile envelope. The upper and lower quantiles can be selected as 90% and 10% quantiles, respectively, or 95% and 5% quantiles can be selected according to the company's quality control requirements.

[0072] When the sample size is small or there are slight outliers, the system can use robust statistical methods to construct the envelope. For example, the envelope boundary can be formed by adding or subtracting a certain multiple of the median absolute deviation from the median, or the upper and lower quantiles can be corrected using the interquartile range. The baseline curve and quantile envelope formed in this way can reflect the typical position and reasonable fluctuation range of the error curve of a normal and qualified electricity meter under current manufacturing conditions.

[0073] When the number of qualified samples is lower than the preset number, the system will not perform curve morphology risk assessment, or will use the benchmark curve of the previous stable process window as a temporary reference. When the calibration status of the verification equipment, aging conditions, ambient temperature and humidity, or key process parameters change significantly, the system will re-divide the process window and reconstruct the qualified region benchmark curve and quantile envelope.

[0074] Example 4: Calculation of Risk Values ​​for Physical Residuals and Curve Shape under Operating Conditions

[0075] For the electricity meter to be evaluated Let its multi-condition error curve be... The benchmark curve for the qualified region is Then the first The physical residuals of each operating point are:

[0076]

[0077] Based on this, the system calculates multiple curve shape risk indicators according to the physical sequence of operating conditions. The overall residual value R is used to characterize the overall positional deviation of the curve under evaluation relative to the benchmark curve in the qualified domain, and can be obtained by weighting the absolute values ​​of the residuals of each operating condition according to the operating condition weights. The adjacent operating condition jump variable J is used to characterize whether the residual changes between adjacent current ratios, adjacent power factors, or adjacent directional operating conditions are abnormally increased. The second-order bending anomaly C is used to characterize whether there are abnormal bends or nonlinear distortions in the residual curve.

[0078] The envelope contact ratio E characterizes the proportion of operating points in the curve under evaluation that are close to the upper or lower quantile envelope boundaries, as well as the length of continuous points close to the same side of the envelope. The drift before and after aging D characterizes whether the error change under the same operating condition before and after high-temperature aging exceeds the normal drift range of qualified samples in the same batch. The forward and reverse asymmetry H characterizes whether the difference between the forward and reverse electrical energy errors increases abnormally.

[0079] The curve shape risk value Q can be generated by the following formula:

[0080]

[0081] in: These are the weights of the corresponding risk indicators, which can be set according to the company's internal control focus, the types of historical quality problems, or process expert rules, or adjusted according to the stability of risk indicators in the same batch of qualified samples.

[0082] The system determines the risk ceiling based on the Q-value distribution calculated from the qualified domain samples themselves. For example, the upper quantile of the Q-values ​​of the qualified domain samples can be used as the risk ceiling, or the median plus a certain multiple of the median absolute deviation can be used to determine the risk ceiling. If the Q-value of the electricity meter to be evaluated is higher than the risk ceiling, and all single-point errors of the electricity meter are still within the enterprise's internal control acceptable limits, then it is identified as a curve-shaped risk product within the qualified domain.

[0083] Example 5: Four-layer manufacturing processing rules

[0084] like Figure 5 As shown, after identifying the risk of curve shape within the qualified domain, the system performs four-layer manufacturing processing based on the risk indicator type and manufacturing traceability field: table level, equipment level, process level, and component level.

[0085] I. Meter Position Layer Handling. When the risk is concentrated in the form of consecutive increases in adjacent operating condition jumps, increased envelope contact rate, or sudden changes in local residuals at the same meter position, and other meters on the same calibration equipment do not exhibit similar phenomena, the system determines that the risk is mainly related to the meter position clamping, contact status, or the stability of the single meter position sampling channel. The system outputs a meter position re-inspection, meter clamp contact check, wiring terminal tightening check, or a single meter position suspension online command; for associated energy meters that have already been tested by this meter position, the system marks them as a batch with questionable meter positions and recommends re-inspection.

[0086] II. Equipment-level Handling. When multiple meters corresponding to the same calibration equipment or the same standard source simultaneously exhibit an overall increase in residuals in the same direction, with the curves closely approximating the upper or lower quantile envelopes, and showing a consistent trend across different meters, the system determines that the risk is primarily related to the overall output stability of the calibration equipment. The system should issue commands for equipment calibration, standard voltage and current output stability checks, power amplifier status checks, or equipment shutdown verification, and mark the batch of energy meters tested during the abnormal period.

[0087] III. Process-Level Handling. When the risk is mainly manifested as an increase in drift before and after aging, or occurs concentrated in the same aging furnace, the same FCT period, the same factory inspection period, or the same environmental fluctuation range, the system determines that the risk is mainly related to the stability of process execution or the controlled status of the process window. The system outputs instructions for aging condition verification, FCT test parameter review, inspection environment confirmation, process cycle time check, or stricter sampling inspection, and marks the relevant products as batches with questionable processes.

[0088] IV. Component-Level Handling. When the risk mainly manifests as abnormal curve bending under specific power factor conditions, increased forward and reverse asymmetry, or abnormal residual differences at specific load points, and is related to batch information of sampling resistors, crystal oscillators, metering chips, PCB soldering, or other key components, the system determines that the risk is primarily related to the manufacturing quality of key components inside the energy meter. The system outputs instructions for component batch isolation, rework investigation, retesting of related batches, or traceability of key component supply batches.

[0089] Furthermore, when multiple levels simultaneously meet the disposal conditions, the system can execute coordinated disposal based on the contribution of risk indicators and the scope of production impact. For example, when both the surface level and the equipment level have high risks, surface isolation and equipment verification are prioritized; when both the process level and the component level have high risks, related batch retesting is prioritized, and component-level rework analysis is arranged simultaneously. Through the above mechanism, the system forms a closed-loop quality control process of "qualified domain risk identification—level judgment—targeted disposal—result write-back—rule optimization".

[0090] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A system for stratified analysis and anomaly identification in the manufacturing quality of electricity meters, characterized in that, The system includes: The qualified domain data acquisition and object screening module is used to collect the multi-condition error values ​​of the electricity meter under the same preset operating condition sequence, the error change before and after aging, the CDI key test quantity and manufacturing traceability field, and the electricity meter whose full range is within the enterprise's internal control qualified limit is used as the qualified domain anomaly identification object. The operating condition serialization and error curve construction module is used to sort multiple operating condition error values ​​of the energy meter based on the preset physical sequence of the detection stage, and form a multi-operating condition error curve. The qualified domain baseline curve and quantile envelope construction module is used to calculate the baseline median curve, upper quantile envelope and lower quantile envelope corresponding to each preset working point based on the qualified domain samples, forming a qualified domain curve reference to characterize the range of normal curve shape. The working condition physical residual calculation module is used to compare the multi-working condition error curves of the energy meter to be evaluated with the qualified domain reference curve and quantile envelope, and calculate the overall residual, adjacent working condition jump variable, second-order bending anomaly, envelope edge contact rate, drift before and after aging and forward and reverse asymmetry according to the working condition physical order. The curve shape risk identification module is used to generate curve shape risk values ​​within the qualified domain based on at least three items from the results of the working condition physical residual calculation module, and to identify potential reliability risk products that are qualified at a single point but have abnormal curve shapes based on the risk upper limit formed by the qualified domain samples. The four-layer manufacturing handling module is used to handle risks based on information about products with identified potential reliability risks.

2. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, The multi-condition error values ​​include at least the error values ​​of the energy meter under different current ratios, different power factors, positive active power, reverse active power, and different phases; the error change before and after aging includes the error difference or error change rate before and after high-temperature aging under the same condition; the manufacturing traceability fields include production barcode, batch number, meter position number, calibration equipment number, process timestamp, tooling fixture number, key component batch number, and environmental temperature and humidity records.

3. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, The CDI key test quantities are used as auxiliary data acquisition fields for qualified domain object screening and four-layer manufacturing processing. They include FCT key point voltage, sampling loop resistance, daily timing results, communication response time, voltage and current stability of calibration equipment, meter contact status, and process test quantities related to the manufacturing quality of the electricity meter. The CDI key test quantities are only used as auxiliary data for qualified domain screening and manufacturing processing and are not used as independent anomaly identification algorithms.

4. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, In the operating condition serialization and error curve construction module, multiple operating condition error values ​​of the same energy meter are sorted according to the preset physical order of current ratio, power factor, forward and reverse energy, phase, and detection stages before and after aging.

5. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, The qualified domain samples are energy meter samples that are all within the enterprise's internal control qualified limits within the same batch, the same calibration equipment group, the same process window, or the same production time window; when the process parameters, calibration equipment status, environmental status, or production batch are changed, the corresponding qualified domain benchmark curve and quantile envelope are reconstructed.

6. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, The qualified range baseline curve is formed by the median, truncated mean, or robust mean of each operating point; the upper quantile envelope and lower quantile envelope are formed by the upper and lower quantiles of each operating point, and can also be extended based on the quantiles by combining the interquartile range or median absolute deviation to characterize the statistical envelope range of the normal qualified curve within the same batch or the same process window.

7. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, The physical residual under the operating condition is not the distance between disordered state data, but a curve residual formed based on the current rate increase relationship, power factor switching relationship, positive and negative energy correspondence relationship and aging stage correspondence relationship; It is used to characterize the overall position, local continuity, curvature change, boundary proximity, aging stability, and forward and reverse symmetry of the error curve of the energy meter under evaluation relative to the qualified domain reference curve.

8. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 1, characterized in that, The four-layer manufacturing processing module is used to generate processing instructions for table re-inspection, equipment calibration, process verification, or component batch isolation based on the contribution type of each risk indicator in the curve shape risk value, combined with the table number, equipment number, process record, and component batch information.

9. The system for stratified analysis and anomaly identification of electricity meter manufacturing quality according to claim 8, characterized in that, The four-layer manufacturing processing module includes a surface position processing unit, an equipment processing unit, a process processing unit, and a component processing unit. The surface position processing unit is used to generate surface position re-inspection, surface clamp inspection, or single surface position pause instructions. The equipment processing unit is used to generate calibration instructions for verification equipment, standard source inspection, or equipment shutdown verification instructions. The process processing unit is used to generate aging process verification, process parameter review, or tightened sampling inspection instructions. The component processing unit is used to generate batch isolation, rework investigation, or related batch retest instructions for key components.

10. A method for quality stratification analysis and anomaly identification of electricity meters based on the system described in any one of claims 1-9, characterized in that, The method specifically includes the following steps: S1, collect the multi-condition error values, error changes before and after aging, CDI key test quantities and manufacturing traceability fields of each energy meter under the same preset operating condition sequence, and judge the pass / fail limits of each operating condition error value and key test quantity. S2, only include energy meters whose preset working condition error values ​​and key test quantities are within the enterprise's internal control qualification limits into the qualification domain abnormal identification object set, and energy meters whose any indicator exceeds the limit are handled according to the conventional non-qualification process; S3, according to the preset physical sequence of current ratio, power factor, forward and reverse power, phase and aging stages, constructs multiple error values ​​of the qualified domain abnormal identification object into a multi-condition error curve. S4 generates the qualified region baseline curve, upper quantile envelope, and lower quantile envelope based on qualified region samples within the same batch or process window. S5 compares the multi-condition error curves of the energy meter to be evaluated with the qualified domain reference curve and quantile envelope, and calculates the overall residual, adjacent condition jump variable, second-order bending anomaly, envelope edge contact rate, drift before and after aging and forward and reverse asymmetry. S6. Generate curve shape risk values ​​within the qualified domain based on the above risk indicators, and identify potential reliability risk products that are qualified at a single point but have abnormal curve shapes. S7 determines the disposal objects at the tabletop, equipment, process, or component level based on the contribution type of the curve shape risk indicator and the manufacturing traceability field, and outputs tabletop re-inspection, equipment calibration, process verification, or component batch isolation instructions.

Images