Intelligent electric energy meter full life cycle fault root cause analysis method and system
By dynamically adjusting the kernel weights of a multi-core support vector machine in the smart meter production environment, the problem of static weights being unable to adapt to seasonal temperature drift and equipment aging is solved, enabling accurate classification of fault types and root cause tracing, thereby improving production efficiency and service life.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGYIN ZHONGHE POWER METER
- Filing Date
- 2026-04-15
- Publication Date
- 2026-05-12
AI Technical Summary
Existing multi-core support vector machine algorithms in smart meter production environments suffer from misclassification of fault types and inaccurate tracing of the root cause of sudden heat changes due to the inability of static kernel weights to adapt to seasonal temperature drift and aging of heating equipment.
By calculating the temperature deviation in the workshop and the aging compensation factor of the heating equipment, the kernel weights of the multi-core support vector machine are dynamically adjusted, the target discriminant function is reconstructed, and the root cause of the smart energy meter failure is traced.
It improves the accuracy of fault type classification and the robustness of root cause tracing in smart energy meters, enhances the real-time nature of full life cycle analysis, and can promptly and accurately pinpoint rework nodes caused by sudden heat changes, thereby improving production efficiency and service life.
Smart Images
Figure CN122020339A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of knowledge-based model computing devices. More specifically, this invention relates to a method and system for root cause analysis of faults throughout the entire lifecycle of a smart energy meter. Background Technology
[0002] With the continuous deepening and popularization of smart grid construction, the long-term operational reliability of smart meters, as the core terminal equipment for data acquisition and metering, is of paramount importance in complex field environments. Since the production and manufacturing of smart meters involves many complex and precise automated full life cycle processes such as surface mount, dual in-line package, and wave soldering, printed circuit boards often undergo multiple manual resoldering and reflow processes during actual assembly line production due to misjudgment by automatic instruments or substandard test parameters. This irregular and repeated rework will cause the tiny electronic components on the printed circuit board to be subjected to multiple non-standard local thermal shocks, resulting in latent thermal stress damage. This damage may appear normal during factory testing, but after long-term full-load operation, it is very easy to cause serious failures such as pad peeling and communication module offline.
[0003] Currently, multi-core support vector machine (SVM) algorithms are commonly used to mine the correlation between discrete rework text and continuous temperature and power curves for classification. However, after the model training is completed, the weight distribution ratio between the kernel functions in the existing multi-core SVM algorithm is static and fixed. In actual automated production lines, the reference temperature of the workshop's macroscopic environment will drift significantly with the seasons. For example, in the extremely cold winter environment, heat dissipates very quickly, while in the high-temperature summer environment, heat can easily accumulate in the deep substrate. Static kernel weights are completely unable to perceive and adapt to this macroscopic seasonal drift, resulting in the rigidity of the decision hyperplane in the high-dimensional space partitioning. A benign rework curve that is safe in winter may evolve into malignant data that causes thermal stress damage in summer. At the same time, this static solidification mechanism cannot detect the fatal damage caused by the hidden deep heat energy superposition when facing the phenomenon of power surge caused by the micro-aging of the underlying heating equipment. This easily leads to serious classification misjudgments when the algorithm processes historical batch operation data across seasons and months, and thus cannot accurately trace the real thermodynamic root cause of the on-site failure of smart meters throughout their entire life cycle from the complex production network. Summary of the Invention
[0004] To address the technical problem that existing multi-core support vector machine algorithms have statically fixed weights for each kernel function, making them difficult to adapt to seasonal temperature drift and aging heating equipment in smart meter production workshops, leading to misclassification of fault types and inability to accurately trace the root cause of sudden heat changes, this invention provides solutions in the following aspects.
[0005] In a first aspect, the present invention provides a method for analyzing the root cause of faults in a smart meter throughout its entire lifecycle, comprising: collecting the full lifecycle operation record data of the smart meter, preprocessing it to obtain a multimodal operation dataset of the smart meter; determining the initial kernel weight adjustment coefficient of the smart meter based on the difference between the workshop ambient temperature and the standard workshop reference temperature in the multimodal dataset; determining the equipment aging compensation factor of the smart meter based on the difference between the average actual operating output power of each heating device in the multimodal dataset and its rated standard power under factory health conditions; determining the optimized kernel weight adjustment coefficient of the smart meter by combining the initial kernel weight adjustment coefficient and the equipment aging compensation factor; determining the optimized weight of each kernel function of the smart meter using the optimized kernel weight adjustment coefficient, and reconstructing the target discriminant function of the multi-core support vector machine of the smart meter; classifying the fault types of the smart meter using the target discriminant function; judging the difference between the actual operating output power of each heating device and the rated standard power using a preset danger judgment threshold based on the classification results; tracking the rework time node of the smart meter generating heat mutation; and locking the root cause of the fault.
[0006] This invention highlights the impact of macroscopic environmental drift by calculating the initial kernel weight adjustment coefficient and utilizing workshop temperature deviation; it analyzes the risk of power surge in heating equipment by calculating equipment aging compensation factors; it enhances the classification robustness of multi-kernel support vector machines in complex production environments by combining the two to determine the optimized kernel weight adjustment coefficient and adaptively adjusts the weights of each kernel function; and it achieves fault root cause locking by reconstructing the target discriminant function and performing reverse path tracing based on the classification results, thereby improving the accuracy and traceability efficiency of fault analysis throughout the entire life cycle of smart energy meters.
[0007] Preferably, the method for obtaining the initial kernel weight adjustment coefficient is as follows: calculate the absolute value of the difference between the workshop ambient temperature and the standard workshop reference temperature in the multimodal dataset to obtain a first value; calculate the ratio between the first value and the standard workshop reference temperature to obtain a second value; apply the natural exponential function to the second value; and use the result as the initial kernel weight adjustment coefficient.
[0008] This invention evaluates the initial kernel weight adjustment coefficient by analyzing the ambient temperature of the workshop and the reference temperature of the standard workshop. The initial kernel weight adjustment coefficient reflects the degree of seasonal ambient temperature drift, so that when the deviation between the ambient temperature of the workshop and the reference temperature of the standard workshop is large, a larger initial kernel weight adjustment coefficient is calculated, and when the deviation between the ambient temperature of the workshop and the reference temperature of the standard workshop is small, a smaller initial kernel weight adjustment coefficient is calculated. This provides a reliable environmental perception basis for subsequent optimization of the kernel weight adjustment coefficient calculation.
[0009] Preferably, the method for obtaining the equipment aging compensation factor is as follows: calculate the difference between the average actual operating output power of each heating device in the multimodal dataset and its rated standard power under factory health conditions to obtain a third value; calculate the ratio between the third value and the rated standard power of each heating device under factory health conditions to obtain a fourth value; sum the fourth values corresponding to all heating devices, and use the result as the equipment aging compensation factor.
[0010] This invention evaluates the equipment aging compensation factor by analyzing the average actual operating output power of each heating device and its rated standard power under factory healthy conditions. The equipment aging compensation factor reflects the hidden deep thermal radiation risk caused by the micro-aging of the heating device. This results in smart meters using severely aged heating devices calculating a larger equipment aging compensation factor, while smart meters using healthy heating devices calculate a smaller equipment aging compensation factor. This provides an aging compensation analysis basis for optimizing the calculation of the core weight adjustment coefficient.
[0011] Preferably, the method for obtaining the optimized kernel weight adjustment coefficient is as follows: applying the natural exponential function to the equipment aging compensation factor to obtain a fifth value, calculating the product of the fifth value and the equipment aging compensation factor, calculating the sum of the product and the initial kernel weight adjustment coefficient, and using the obtained result as the optimized kernel weight adjustment coefficient.
[0012] This invention constructs a composite function that linearly superimposes the exponential gain term of the equipment aging compensation factor and the initial kernel weight adjustment coefficient, thereby achieving the evaluation of the optimized kernel weight adjustment coefficient. The optimized kernel weight adjustment coefficient reflects the combined effects of environmental drift and equipment aging, resulting in a larger optimized kernel weight adjustment coefficient for smart meters under the superposition of dual risks, while a smaller optimized kernel weight adjustment coefficient is calculated for smart meters under normal conditions. This provides a dynamic adjustment basis for the calculation of the weights of each kernel function after optimization.
[0013] Preferably, the method for obtaining the optimized weights of each kernel function is as follows: apply the natural exponential function to the optimized kernel weight adjustment coefficient, use the optimized kernel weight adjustment coefficient after applying the natural exponential function as the exponential term of the initial weight of each kernel function, use the result as the sixth value, calculate the ratio between the sixth value of each kernel function and the sum of the sixth values of all kernel functions, and use the result as the optimized weight of each kernel function.
[0014] This invention constructs a normalization function to optimize the exponential term of the kernel weight adjustment coefficient, thereby enabling the evaluation of the weights of each kernel function after optimization. This normalization term reflects the intensity of the intervention of environmental and aging risks on different feature kernels, which significantly amplifies the kernel function weights corresponding to real-time heating temperature data and actual operating output power data, while relatively compressing the kernel function weights corresponding to rework process sequence text data. This provides an adaptive weight basis for the reconstruction of the target discriminant function.
[0015] Preferably, the reconstruction of the target discriminant function of the multi-core support vector machine of the smart energy meter includes: substituting the optimized weights of each kernel function into the initial target discriminant function, replacing the initial weights with the optimized weights of each kernel function of the smart energy meter, and completing the reconstruction of the target discriminant function of the multi-core support vector machine of the smart energy meter.
[0016] Preferably, the step of tracking the rework time node of the heat mutation of the smart energy meter and locking the root cause of the fault includes: in response to the classification result being severe thermal damage, reverse path tracing is performed based on the kernel function corresponding to the actual operating output power data of the heating equipment corresponding to all rework processes of the smart energy meter; in response to the difference between an actual operating output power data of the heating equipment corresponding to a rework process and the rated standard power being greater than a preset danger judgment threshold, the time node of the actual operating output power data of the heating equipment corresponding to the rework process is the rework time node of the heat mutation of the smart energy meter, thereby locking the root cause of the fault.
[0017] Preferably, the collection of the full life cycle circulation record data of the smart energy meter includes: collecting the full life cycle circulation record data of the smart energy meter through an IoT sensor cluster deployed in various production workshops, which includes text data of all rework process sequences of the smart energy meter, actual operating output power data and real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart energy meter, and workshop ambient temperature data during the production period of the smart energy meter.
[0018] Preferably, the preprocessing to obtain the multimodal operation dataset of the smart energy meter includes: based on the unique medium access control address identification code burned into the smart energy meter, performing timestamp alignment operations on the collected text data of all rework process sequences of the smart energy meter, the actual operating output power data and real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart energy meter, and the workshop ambient temperature data during the production period of the smart energy meter, to construct the multimodal operation dataset of the smart energy meter.
[0019] Secondly, the present invention provides a smart energy meter full life cycle fault root cause analysis system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned smart energy meter full life cycle fault root cause analysis method is implemented.
[0020] By adopting the above technical solution, a computer program is generated from the above-mentioned method for analyzing the root cause of faults throughout the entire life cycle of a smart energy meter, and stored in a memory for loading and execution by a processor. This allows for the creation of terminal devices based on the memory and processor, making them convenient to use.
[0021] The beneficial effects of this invention are as follows: This invention solves the technical problem that static weights in existing multi-core support vector machine algorithms are prone to misclassification in the smart energy meter production environment by introducing a dynamic kernel weight adjustment mechanism based on the fusion of workshop temperature drift and heating equipment aging.
[0022] This invention establishes the intrinsic mapping relationship between the initial kernel weight adjustment coefficient and the equipment aging compensation factor by analyzing workshop temperature deviation and equipment power deviation. On this basis, the two are further integrated to construct an optimized kernel weight adjustment coefficient, which tightly couples the indicators reflecting the macro-environmental impact with the indicators reflecting the micro-aging risk, thereby achieving accurate identification of seasonal drift and latent thermal radiation.
[0023] This invention calculates optimized weights for each kernel function, matching the complexity of its actual operating conditions, for each smart meter. Through adaptive reconstruction of the target discriminant function, it achieves accurate matching between the decision hyperplane and production data. For smart meters with severe environmental impact and high aging risk, the kernel function weights corresponding to power and temperature characteristics are amplified to improve the sensitivity of detecting severe thermal damage. For smart meters in normal conditions, static weights are maintained to prevent overfitting. Ultimately, this invention improves the accuracy of smart meter fault type classification and the robustness of root cause tracing, enhances the real-time performance of full lifecycle analysis, and can promptly and accurately pinpoint rework nodes caused by sudden thermal changes, providing a solid technical guarantee for improving the productivity and extending the service life of smart meters. Attached Figure Description
[0024] Figure 1 This is a flowchart illustrating a method for analyzing the root cause of faults in a smart energy meter throughout its entire lifecycle, as described in this invention. Figure 2 This is a comparison chart showing the accuracy of existing technologies and the present invention in classifying faults in smart energy meters. Detailed Implementation
[0025] 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, not all, of the embodiments of the present invention. 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.
[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0027] This invention discloses a method for root cause analysis of faults throughout the entire lifecycle of smart energy meters, referring to... Figure 1 This includes steps S001-S005: S001: Collect the entire lifecycle data of the smart energy meter, preprocess it, and obtain the multimodal operation dataset of the smart energy meter.
[0028] Specifically, IoT sensor clusters deployed in various production workshops collect data on the entire lifecycle of smart meters. This data includes textual data of all rework process sequences for the smart meters, actual operating output power data and real-time heating temperature data of the heating equipment corresponding to each rework process, and workshop ambient temperature data during the smart meter production period. In this embodiment, the collection frequency for the actual operating output power data of the heating equipment corresponding to each rework process, the real-time heating temperature data, and the workshop ambient temperature data during the smart meter production period is set to 10 times per second. In other embodiments, the implementer can set this frequency according to the actual situation. Based on the unique media access control address identification code burned into the smart meter, the collected textual data of all rework process sequences for the smart meters, the actual operating output power data of the heating equipment corresponding to each rework process, the real-time heating temperature data, and the workshop ambient temperature data during the smart meter production period are timestamped to construct a multimodal operation dataset for the smart meters.
[0029] S002: Determine the initial kernel weight adjustment coefficient of the smart energy meter based on the difference between the workshop ambient temperature and the standard workshop reference temperature in the multimodal dataset.
[0030] It should be noted that the statically fixed parameters in existing multi-core support vector machine models are difficult to adapt to the seasonal drift of the macroscopic environment thermodynamic baseline in the production workshop. This results in the model being unable to effectively perceive the interference and damage caused by changes in the external environment temperature, leading to inaccurate final fault assessment results. According to thermodynamic heat transfer theory, the greater the absolute deviation between the workshop ambient temperature and the standard workshop reference temperature, the more severe the interference of the macroscopic environment on the heat dissipation and thermal stress accumulation of the circuit board. Therefore, this invention combines the difference between the workshop ambient temperature and the standard workshop reference temperature in the multimodal dataset to calculate the initial kernel weight adjustment coefficient of the smart energy meter, which is used to analyze the degree of drift of the macroscopic environment thermodynamic baseline and provide a physical environment perception basis for subsequent kernel weight adjustment.
[0031] Specifically, the initial kernel weight adjustment coefficients satisfy the expression: ; In the formula, For the first The initial kernel weight adjustment coefficient for each smart energy meter. For the first Average workshop ambient temperature data during the production period of each smart energy meter The reference temperature for the standard workshop. It is a natural exponential function. To obtain an absolute value, the standard workshop reference temperature is set to 25 degrees Celsius in this embodiment. In other embodiments, the implementer can set it according to the actual situation.
[0032] In the formula, The larger the value, the more likely it is to be the first. The more severe the deviation of the workshop ambient temperature from the standard workshop reference temperature during the season in which the smart meter is produced, the more serious the deviation. The larger the initial kernel weight adjustment coefficient of a smart energy meter, the stronger the intervention in the kernel function weight related to processing real-time heating temperature data.
[0033] To more clearly illustrate the effect and calculation process of the initial kernel weight adjustment coefficient, the following example will demonstrate this: In one embodiment of the present invention, when the production period of the smart energy meter coincides with the spring and autumn seasons with suitable climates, the macroscopic thermodynamic baseline of the production workshop is extremely stable, and the collected and calculated data is... The average workshop ambient temperature during the production period of each smart energy meter is 25 degrees Celsius. ; In another embodiment of the invention, when the smart energy meter is being produced during the hot summer months or under harsh conditions where the constant temperature air conditioning is not turned on, the first... The average workshop ambient temperature during the production period of each smart energy meter soared to 35 degrees Celsius. ; The analysis of the two scenarios of workshop ambient temperature data above shows that when the workshop ambient temperature drifts significantly, the calculated initial kernel weight adjustment coefficient will increase significantly exponentially from the baseline value of 1. This is consistent with the physical fact that the greater the absolute deviation of the ambient temperature difference, the more severe the interference on the heat dissipation of the circuit board, enabling the algorithm to adaptively perceive the deterioration of the macro environment.
[0034] S003: Based on the difference between the average actual operating output power of each heating device in the multimodal dataset and its rated standard power under factory health conditions, determine the device aging compensation factor of the smart energy meter. Combine the initial kernel weight adjustment coefficient and the device aging compensation factor to determine the optimized kernel weight adjustment coefficient of the smart energy meter.
[0035] It should be noted that during the rework process in smart meter manufacturing, the ceramic heating wires inside heating equipment such as wave soldering or hot air presses undergo subtle micro-aging over time. To maintain the real-time heating temperature set on the process interface, the heating equipment adaptively and significantly increases its actual power output. This results in the deep substrate inside the smart meter's circuit board still implicitly absorbing excessive and destructive heat energy, even under normal conditions where the workshop ambient temperature is suitable and the real-time heating temperature curve appears to be within acceptable limits. This causes deep heat accumulation damage that is difficult to detect with the naked eye and conventional surface temperature measurement methods, thus reducing the comprehensiveness and accuracy of the assessment of severe thermal damage. According to the theory of thermal radiation and energy penetration, changes in workshop ambient temperature and real-time heating temperature are not equivalent to the actual accumulation state of deep energy inside the smart meter. The surge in the actual operating output power of the heating equipment directly corresponds to the physical process of the heating equipment injecting highly penetrating thermal stress into the smart meter. Therefore, this invention determines the equipment aging compensation factor of the smart energy meter based on the difference between the average actual operating output power of each heating device in the multimodal data set and its rated standard power under factory health conditions. This factor characterizes the risk of each heating device injecting latent damaging thermal radiation into the circuit board due to micro-aging. Combining the initial kernel weight adjustment coefficient and the equipment aging compensation factor, the optimized kernel weight adjustment coefficient of the smart energy meter is determined. This constructs a dual dynamic adjustment mechanism that combines workshop ambient temperature analysis and heating equipment aging compensation to characterize the latent deep thermal damage risk of the smart energy meter.
[0036] Specifically, the equipment aging compensation factor satisfies the following expression: ; In the formula, For the first Equipment aging compensation factor for smart energy meters For the first The total number of heating devices corresponding to each smart energy meter is consistent with the total number of rework processes. For the first The first smart energy meter The average actual operating output power of each heating device For the first The first smart energy meter The rated standard power of each heating device under factory-safe, healthy condition. To take the absolute value.
[0037] In the formula, The larger the value, the more likely it is to be the first. The first smart energy meter In order to overcome its severe aging and degradation, the heating device consumes more and more redundant electrical energy, which is converted into hidden and deadly deep heat radiation that penetrates into the circuit board of the smart meter. Therefore, the first The device aging compensation factor of the first smart energy meter is used to compensate for the aging of the first smart energy meter. The first smart energy meter The real-time heating temperature of each heating device and the ambient temperature of the workshop are numerically lower than expected.
[0038] To more clearly illustrate the effect and calculation process of the equipment aging compensation factor, the following example will demonstrate this: First, let's assume the... The rated power of the two heating devices corresponding to the two rework processes of the smart energy meter is 2000w and 500w respectively. In one embodiment of the present invention, when both heating devices are in good working order immediately after leaving the factory, their average actual operating output power is 2000W and 500W respectively. ; In another embodiment of the invention, when the ceramic heating wires inside both heating devices age, and these two heating devices are forced to significantly increase their output power in order to maintain the real-time heating temperature, the average actual operating output power surges to 2400W and 600W respectively. ; The comparison between the two states of the heating equipment above shows that the aging compensation factor of the equipment in a healthy state is 0. However, when the heating equipment consumes redundant electrical energy to overcome its own micro-aging, the aging compensation factor will accumulate and amplify proportionally. Through this calculation process, the illusion of meeting the heating temperature standard can be identified, and the destructive deep heat radiation risk hidden in the bottom layer of the equipment can be analyzed and identified.
[0039] Specifically, the optimized kernel weight adjustment coefficient satisfies the expression: ; In the formula, For the first The optimized kernel weight adjustment coefficient for a smart energy meter. For the first The initial kernel weight adjustment coefficient for each smart energy meter. For the first Equipment aging compensation factor for smart energy meters It is a natural exponential function.
[0040] In the formula, The larger the value, the more likely it is to be the first. The more severe the deviation of the workshop ambient temperature from the standard workshop reference temperature during the season in which the smart meter is produced, the more serious the deviation. The larger the optimization kernel weight adjustment coefficient of a smart energy meter, the better. The larger the value, the more likely it is to be the first. In order to overcome its severe aging and degradation, the heating device of a smart meter consumes more and more redundant electrical energy, which is then converted into deep heat radiation that penetrates into the circuit board of the smart meter. Therefore, the first... The more the initial kernel weight adjustment coefficient of a smart energy meter needs to be compensated, the more the first... The larger the optimization kernel weight adjustment coefficient of a smart energy meter, the better.
[0041] To more clearly illustrate the effect and calculation process of optimizing the kernel weight adjustment coefficient, the following example will demonstrate this: In one embodiment of the present invention, when the smart energy meter is in the optimal normal state of ideal workshop temperature and heating equipment is healthy, according to the aforementioned calculation logic, the initial kernel weight adjustment coefficient is 1, and the equipment aging compensation factor is 0. ; In another embodiment of the present invention, when the smart energy meter is in a dangerous state of abnormally high temperature in the workshop and aging heating equipment, according to the aforementioned calculation logic, the initial kernel weight adjustment coefficient is 1.49, and the equipment aging compensation factor is 0.4, then... ; The analysis of the two combined operating conditions of the smart energy meter above shows that when facing the superposition of dual thermophysical risks, the optimized kernel weight adjustment coefficient achieves a doubling amplification through the combined effect of the exponential and product terms, ensuring that the subsequent multi-core support vector machine algorithm can obtain sufficiently strong intervention when reconstructing the weights.
[0042] S004: By optimizing the kernel weight adjustment coefficient, determine the optimized weights of each kernel function of the smart energy meter, and reconstruct the target discriminant function of the multi-kernel support vector machine of the smart energy meter.
[0043] It should be noted that after obtaining the optimized kernel weight adjustment coefficient, this invention uses the optimized kernel weight adjustment coefficient to complete the adaptive optimization of the initial weight of the kernel function of the multi-kernel support vector machine algorithm, so that the optimized kernel weight can enhance the sensitivity of the multi-kernel support vector machine algorithm to capture smart energy meters with severe thermal damage, and construct a more accurate target discrimination function.
[0044] Specifically, the optimized weights of each kernel function satisfy the expression: ; In the formula, For the first The first smart energy meter Weights optimized by each kernel function For the first The first smart energy meter The initial weights of each kernel function, For the first The total number of kernel functions for a smart energy meter. , No. The kernel function of the first smart energy meter and the first The sample features of each smart energy meter have a one-to-one correspondence. The kernel functions for each smart meter are: the kernel function corresponding to the text data of all rework process sequences of the smart meter; the kernel function corresponding to the actual operating output power data of the heating equipment corresponding to all rework processes of the smart meter; and the kernel function corresponding to the real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart meter. Using the natural exponential function, in this embodiment, the initial weight of the kernel function corresponding to the text data of all rework process sequences of the smart energy meter is set to 0.2, and the initial weights of the kernel functions corresponding to the actual operating output power data of the heating equipment corresponding to all rework processes of the smart energy meter and the real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart energy meter are both set to 0.4. In other embodiments, implementers can set these values according to actual conditions. For example, when the overall service life of the heating equipment in the production workshop is long and the sensitivity requirement for capturing deep hidden thermal damage inside the circuit board is high, the initial weight of the kernel function corresponding to the actual operating output power data can be appropriately increased to improve the algorithm's judgment of the micro-aging of the heating equipment. When the rework flow path of the smart energy meter is complex or the compliance monitoring requirements for surface conventional thermal shock are high, the initial weight of the kernel function corresponding to the rework process sequence text data or the real-time heating temperature data can be appropriately increased to improve the algorithm's tracking and evaluation efficiency of process topology trajectory or surface heat accumulation.
[0045] In the formula, The larger the value, the more likely it is to be the first. The more severe the deviation of the workshop ambient temperature from the standard workshop reference temperature during the production period of the smart meter, the more significant the deviation. The more severe the aging and degradation of the heating equipment in a smart meter, the more serious the degradation. The first smart energy meter The larger the weight of each optimized kernel function, the stronger the intervention in the kernel function weights related to processing real-time heating temperature data and actual operating output power data can be through exponential adjustments.
[0046] To more clearly illustrate the effect and calculation process of the weights after optimization by each kernel function, the following examples will demonstrate this: First, assume that the initial weight of the kernel function corresponding to the text data of all rework process sequences of the smart energy meter is set to 0.2, and the initial weights of the kernel functions corresponding to the actual operating output power data of the heating equipment corresponding to all rework process sequences of the smart energy meter and the real-time heating temperature data of the heating equipment corresponding to all rework process sequences of the smart energy meter are both set to 0.4; since the minimum theoretical value of the optimization kernel weight adjustment coefficient is 1 under ideal optimal normal operating conditions, the kernel weight adjustment coefficient is set to 0.4 as an exponential term. The minimum is approximately 2.718; In one embodiment of the present invention, when the smart energy meter is in a normal, benign state, the calculated... When the integer approximation is 3, the optimized weights of each kernel function are as follows: the optimized weight of the kernel function corresponding to the text data of all rework process sequences of the smart energy meter is approximately 5.88%, and the optimized weights of the kernel functions corresponding to the actual operating output power data of the heating equipment corresponding to all rework processes of the smart energy meter and the real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart energy meter are approximately 47.06%. In another embodiment of the present invention, when the smart energy meter is in a state of high risk of severe thermal damage, it is assumed that the calculated exponential term... When the value is increased to 4, the optimized weights of each kernel function are as follows: the optimized weight of the kernel function corresponding to the text data of all rework process sequences of the smart energy meter is approximately 3.03%, and the optimized weights of the kernel functions corresponding to the actual operating output power data of the heating equipment corresponding to all rework processes of the smart energy meter and the real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart energy meter are approximately 48.48%. Analysis of the weight optimization calculation process reveals that, due to the exponential term... If the value is always greater than 1, the algorithm will compress text sequence features with smaller initial weights, while amplifying power and temperature features with larger initial weights; and when the electricity meter is in a harsh environment, the exponential term will be affected. As the weights are increased further, the polarization effect of this weight allocation will be further enhanced, causing the model's decision focus to shift sharply towards power and temperature characteristics, completely breaking the rigidity of static weights and improving the sensitivity of intercepting latent thermal damage.
[0047] Specifically, the objective discriminant function of the multi-core support vector machine of the smart energy meter is reconstructed, including: The optimized weights of each kernel function of the smart energy meter are substituted into the initial target discriminant function, and the optimized weights of each kernel function of the smart energy meter are used to replace the initial weights, thus completing the reconstruction of the target discriminant function of the multi-kernel support vector machine of the smart energy meter.
[0048] S005: The reconstructed target discriminant function is used to classify the fault types of smart energy meters. Based on the classification results, a preset danger judgment threshold is used to determine the difference between the actual operating output power and the rated standard power of each heating device. The rework time node of the smart energy meter generating heat mutation is tracked to lock the root cause of the fault.
[0049] Specifically, track the rework time points when smart meters generate sudden changes in heat to pinpoint the root cause of the fault, including: The fault types of smart energy meters are classified using the reconstructed target discriminant function to obtain classification results. In this embodiment, the classification results are set to two types: malignant thermal damage and benign thermal damage. In other embodiments, implementers can set the types according to the actual implementation situation. In response to the classification result being severe thermal damage, reverse path tracing is performed based on the kernel function corresponding to the actual operating output power data of the heating equipment for all rework processes of the smart energy meter. If the difference between the actual operating output power data of the heating equipment for a rework process and its rated standard power exceeds a preset danger threshold, then the time node of that actual operating output power data is considered the rework time node where the smart energy meter experiences a sudden heat change, thus locking in the root cause of the fault. In this embodiment, the danger threshold is set to 1.5 times the difference between the actual operating output power data of the heating equipment for a rework process and its rated standard power. In other embodiments, implementers can set this threshold according to the actual implementation situation. For example, when the accuracy requirement for tracing deep, hidden thermal damage of the smart energy meter is high, the danger threshold can be appropriately reduced to improve the sensitivity of capturing small, sudden heat change nodes at the bottom layer. When the requirements for anti-interference capability and reducing false alarm rate in batch detection of high-heat rework workshops are high, the danger threshold can be appropriately increased to improve the efficiency of identifying the core fatal fault root cause.
[0050] like Figure 2 As shown in the figure, this graph compares the accuracy of existing technologies and the present invention in classifying faults in smart energy meters. The horizontal axis represents the month in which the smart energy meters were manufactured, and the vertical axis represents the classification accuracy. The solid line marked with an asterisk represents the classification accuracy of the present invention, while the dotted line marked with a cross represents the classification accuracy of the existing technology. This graph verifies the technical principle of the target discriminant function reconstruction mechanism. It is evident that in specific months when the macroscopic environmental temperature fluctuates significantly in the middle of the year and when the underlying heating equipment experiences deep microscopic aging at the end of the year, the existing technology suffers a sharp drop in accuracy, falling to 70%, because its kernel weights are in a static and fixed state, making it difficult to perceive the damage caused by the accumulation of hidden deep heat energy. In contrast, the present invention, by adaptively adjusting the weights of each kernel function, can stably maintain a high classification accuracy of over 98% under various complex operating conditions throughout the year. This result fully demonstrates that the present invention can enhance the classification robustness of the algorithm, thus providing a solid guarantee for accurately tracing the rework time nodes of smart energy meters caused by sudden heat changes.
[0051] This invention also discloses a smart energy meter full life cycle fault root cause analysis system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a smart energy meter full life cycle fault root cause analysis method according to the present invention is implemented.
[0052] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.
Claims
1. A method for root cause analysis of faults throughout the entire life cycle of a smart energy meter, characterized in that, include: Collect the entire lifecycle operation record data of smart energy meters, perform preprocessing, and obtain the multimodal operation dataset of smart energy meters; The initial kernel weight adjustment coefficient of the smart energy meter is determined based on the difference between the workshop ambient temperature and the standard workshop reference temperature in the multimodal dataset. Based on the difference between the average actual operating output power of each heating device in the multimodal dataset and its rated standard power under factory health conditions, the device aging compensation factor of the smart energy meter is determined. Combined with the initial kernel weight adjustment coefficient and the device aging compensation factor, the optimized kernel weight adjustment coefficient of the smart energy meter is determined. Using the optimized kernel weight adjustment coefficient, the optimized weights of each kernel function of the smart energy meter are determined, and the target discriminant function of the multi-kernel support vector machine of the smart energy meter is reconstructed; The target discrimination function is used to classify the fault types of smart energy meters. Based on the classification results, a preset danger judgment threshold is used to determine the difference between the actual operating output power of each heating device and the rated standard power. The rework time node of the smart energy meter generating heat mutation is tracked to lock the root cause of the fault.
2. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The method for obtaining the initial kernel weight adjustment coefficient is as follows: calculate the absolute value of the difference between the workshop ambient temperature and the standard workshop reference temperature in the multimodal dataset to obtain a first value; calculate the ratio between the first value and the standard workshop reference temperature to obtain a second value; apply the natural exponential function to the second value; and use the result as the initial kernel weight adjustment coefficient.
3. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The method for obtaining the equipment aging compensation factor is as follows: calculate the difference between the average actual operating output power of each heating device in the multimodal dataset and its rated standard power under factory health conditions to obtain a third value; calculate the ratio between the third value and the rated standard power of each heating device under factory health conditions to obtain a fourth value; sum the fourth values corresponding to all heating devices, and use the result as the equipment aging compensation factor.
4. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The method for obtaining the optimized kernel weight adjustment coefficient is as follows: apply the natural exponential function to the equipment aging compensation factor to obtain a fifth value, calculate the product of the fifth value and the equipment aging compensation factor, calculate the sum of the product and the initial kernel weight adjustment coefficient, and use the result as the optimized kernel weight adjustment coefficient.
5. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The method for obtaining the optimized weights of each kernel function is as follows: apply the natural exponential function to the optimized kernel weight adjustment coefficient, use the optimized kernel weight adjustment coefficient after applying the natural exponential function as the exponential term of the initial weight of each kernel function, use the result as the sixth value, calculate the ratio between the sixth value of each kernel function and the sum of the sixth values of all kernel functions, and use the result as the optimized weight of each kernel function.
6. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The reconstruction of the target discriminant function of the multi-core support vector machine of the smart energy meter includes: substituting the optimized weights of each kernel function into the initial target discriminant function, replacing the initial weights with the optimized weights of each kernel function of the smart energy meter, and completing the reconstruction of the target discriminant function of the multi-core support vector machine of the smart energy meter.
7. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The method of tracking the rework time node of the heat mutation of the smart energy meter and locking the root cause of the fault includes: in response to the classification result of severe thermal damage, reverse path tracing is performed based on the kernel function corresponding to the actual operating output power data of the heating equipment corresponding to all rework processes of the smart energy meter; in response to the difference between the actual operating output power data of the heating equipment corresponding to a rework process and the rated standard power being greater than the preset danger judgment threshold, the time node of the actual operating output power data of the heating equipment corresponding to the rework process is the rework time node of the heat mutation of the smart energy meter, thereby locking the root cause of the fault.
8. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The collection of full lifecycle circulation record data of smart energy meters includes: collecting full lifecycle circulation record data of smart energy meters through IoT sensor clusters deployed in various production workshops. This data includes text data of all rework process sequences of smart energy meters, actual operating output power data and real-time heating temperature data of heating equipment corresponding to all rework processes of smart energy meters, and workshop ambient temperature data during the production period of smart energy meters.
9. The method for root cause analysis of faults throughout the entire life cycle of a smart energy meter according to claim 1, characterized in that, The preprocessing process yields a multimodal operation dataset for the smart energy meter, which includes: based on the unique medium access control address identification code burned into the smart energy meter, performing timestamp alignment operations on the collected text data of all rework process sequences of the smart energy meter, the actual operating output power data and real-time heating temperature data of the heating equipment corresponding to all rework processes of the smart energy meter, and the workshop ambient temperature data during the production period of the smart energy meter, to construct the multimodal operation dataset for the smart energy meter.
10. A system for analyzing the root cause of faults throughout the entire lifecycle of a smart energy meter, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for analyzing the root causes of faults throughout the entire life cycle of a smart energy meter according to any one of claims 1-9.