Fault detection system and method for electronic component

Abstract

The invention discloses a fault detection system and method for an electronic component, and relates to the technical field of fault detection of the electronic component, and the method comprises the steps: filtering the noise and missing values of a sensor through multi-source high-speed data collection and online preprocessing, and packaging the filtered noise and missing values into a micro-batch data set; a fault is judged in real time and parameters are incrementally updated through a lightweight intelligent fault detection model, then the lightweight intelligent fault detection model is closely integrated with an industrial control system, millisecond-level closed-loop protection is executed by using a multi-level risk function, and feedback is returned; and finally, uploading historical records and mark information to a cloud end, and realizing full-life-cycle management by applying big data analysis and model iteration. Through the technologies of wavelet filtering, incremental learning, cross-level data integration and the like, fault detection delay can be effectively shortened, shutdown loss can be reduced, guarantee is provided for predictive maintenance, the safety and efficiency of a production line are greatly improved, and therefore the requirements of industrial scenes for fault recognition and real-time response can be met.

Description

Technical Field

[0001] This invention relates to the field of electronic component fault detection technology, specifically to a fault detection system and method for electronic components. Background Technology

[0002] In today's highly automated and fast-paced industrial production environments, various production lines (such as automated assembly lines, continuous process chemical lines, or high-precision machining lines) often acquire real-time data from multiple sources, including temperature, pressure, vibration, and current, at millisecond-level sampling frequencies. Meanwhile, to ensure efficient production and stable quality, companies typically deploy multiple monitoring devices and sensor networks to rigorously monitor critical equipment. However, given the high-speed operation of production lines and frequent environmental disturbances, the inability to quickly identify potential fault signs often leads to minor malfunctions rapidly escalating into equipment overheating before they are even detected. Furthermore, millisecond-level data streams must be analyzed in high-fidelity real-time and rapidly integrated with the control system to minimize the risk of fault propagation and maintain normal production rhythm.

[0003] Currently, the core challenge in real-time fault detection in the industrial sector lies in how to rapidly and accurately analyze massive amounts of noisy multi-source sensor data, and deeply integrate the detection results with industrial control systems under conditions of zero or extremely low latency. Traditional monitoring mostly employs alarm mechanisms based on fixed thresholds, which are ill-suited to abnormal patterns caused by equipment aging, environmental changes, or fluctuations in operating conditions. Furthermore, transmitting massive amounts of data to the cloud for complex model analysis can easily lead to a loss of real-time performance due to network bandwidth and latency limitations. In addition, different production lines use different sensor and control system standards, posing significant challenges to interface compatibility and data alignment management. If the detection results cannot accurately identify faults and trigger corresponding emergency plans or warnings within milliseconds, localized faults can quickly escalate into wider equipment failures, resulting in a significant decline in production efficiency and safety. Therefore, a comprehensive solution is urgently needed that integrates rapid on-site analysis, noise filtering, dynamic model updates, and cloud-based big data collaborative optimization to fundamentally solve the problem of efficient real-time fault detection of multi-source sensor data in industrial environments, and achieve true millisecond-level closed-loop response through tight integration with the control system.

[0004] Therefore, the present invention provides a fault detection system and method for electronic components. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a fault detection system and method for electronic components. By relying on multi-source high-speed data acquisition and online preprocessing, sensor noise and missing values ​​are filtered and packaged into micro-batch datasets. A lightweight intelligent fault detection model then determines faults in real time and incrementally updates parameters. Subsequently, it is tightly integrated with the industrial control system, utilizing multi-level risk functions to execute millisecond-level closed-loop protection and provide feedback. Finally, historical records and marking information are uploaded to the cloud, and big data analysis and model iteration are used to achieve full lifecycle management, providing assurance for predictive maintenance and significantly improving production line safety and efficiency. This meets the needs of industrial scenarios for fault identification and real-time response, thus solving the technical problems described in the background art.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A method for fault detection of electronic components, comprising, When high-frequency sampling triggers a surge of multi-source data streams, wavelet filtering and missing value placeholders are applied to the original sensor sequence to remove pseudo data and mark transient anomalies. Interpolation is performed on the late data after alignment delay to generate micro-batch encapsulation to carry sensor ID, timestamp and transient suspicious flags. The edge side uses kernel mapping and accelerated transformation to extract feature vectors, and calls incremental machine learning algorithms to output fault labels and fault confidence in real time, and corrects the model parameters of the fault detection model to adapt to environmental disturbances. The on-site PLC or DCS executes high-priority actions or issues medium-level warnings based on multi-level risk indices, and feeds back the actual execution records and fault labels to the fault detection model, and optimizes the model parameters based on the actual fault confirmation process; After collecting information on micro-batch packaging, fault marking, fault confidence, and multi-level risk index, the cloud or back-end system collects multi-source data and uses aggregation operations to mine fault patterns. The optimized fault detection model and maintenance strategy are then distributed to the edge side, and the model is iterated under the comparison of fault labels.

[0007] Preferably, after acquiring the original sensor sequence, it is aligned by timestamp, and the feature coefficients at multiple scales are extracted by discrete wavelet transform. When the energy proportion of the target scale segment is consistently higher than expected, it is marked as high-frequency interference and suppression is performed. Transient anomalies are identified by wavelet energy index, which is used to detect anomalies in high-energy components, and transient anomaly flags are attached.

[0008] Preferably, the obtained reconstructed signal sequence is divided into multiple micro-batch datasets. After recording metadata at the micro-batch level, wavelet energy index and corresponding flags are retained for the corresponding data points within the micro-batch datasets, and then the micro-batch encapsulation is generated.

[0009] Preferably, after simple interpolation or skipping of missing values, all sensor channels in the micro-batch dataset are obtained and concatenated with the wavelet energy index to form an initial feature vector sequence. A customizable kernel function is introduced to map the initial feature vector sequence to obtain the dimensionality-reduced feature representation. An accelerated transformation index is introduced to dynamically adjust the dimensionality reduction intensity in the kernel mapping process, and the output feature vector is obtained.

[0010] Preferably, based on the feature vector, a lightweight fault detection model is used to quickly predict whether a fault exists and its severity, and the model parameters are incrementally updated in small batches or online based on real feedback. Add a fault confidence level to the model output layer. When the fault confidence level exceeds the preset confidence threshold, an emergency control strategy is executed quickly. If the fault confidence level does not exceed the preset confidence threshold, only a warning or recording is selected.

[0011] Preferably, the acquired detection results are transmitted to the on-site PLC or DCS; and a multi-level risk index of fault confidence and current system critical status is introduced. Based on the landing points of the multi-level risk index in different threshold ranges, the industrial control system executes the corresponding level of protection or alarm, and issues action commands directly to the production line execution equipment through the PLC or DCS.

[0012] Preferably, after the PLC or DCS transmits back the actual actions and operating status data of the system, it pairs the executed actions with the fault detection results and records the newly generated system status. If the previous fault labeling is biased or the fault level is judged insufficiently or excessively, the real label will be used to correct the lightweight fault detection model; if multiple responses for the same fault type are not effective, the relevant parameters of the multi-level risk index will be fine-tuned by the relationship between the multi-level risk index and the final process result.

[0013] Preferably, fault detection records accumulated on field edge devices or industrial PCs are transmitted to the cloud or factory back-end system on a regular or real-time basis; a global aggregation operator is constructed to summarize massive historical data and key field parameters, and a visualization panel is built in the cloud or factory back-end system after data aggregation is completed.

[0014] Preferably, the fault detection model is updated based on the remaining service life of the equipment or the fatigue of key components; if there are fault modes or evolution trends of existing modes, they are recorded and incorporated into the corresponding maintenance strategy library. If there is room for optimization in the fault detection model used, the improved version can be packaged and distributed to the field edge devices; and a tiered maintenance plan can be developed by combining the equipment's remaining service life prediction results and fault mode analysis.

[0015] A fault detection system for electronic components includes, The data acquisition unit, when high-frequency sampling triggers the influx of multi-source data streams, applies wavelet filtering and missing value placeholders to the original sensor sequence to remove pseudo data, marks transient anomalies, and performs interpolation operations on the aligned delayed data to generate micro-batch packages to carry sensor ID, timestamp, and transient suspicious flags. The feature extraction unit extracts feature vectors at the edge using kernel mapping and accelerated transformation, and calls an incremental machine learning algorithm to output fault labels and fault confidence in real time, and corrects the model parameters of the fault detection model to adapt to environmental disturbances. The early warning unit, on-site PLC or DCS, executes high-priority actions or issues medium-level early warnings based on multi-level risk indices, and feeds back the actual execution records and fault labels to the fault detection model, and optimizes the model parameters based on the actual fault confirmation process; The feedback iteration unit collects micro-batch packaging, fault marking, fault confidence, and multi-level risk index information. The cloud or back-end system collects multi-source data and uses aggregation operations to mine fault patterns. The optimized fault detection model and maintenance strategy are then distributed to the edge side, and the model iteration is performed based on fault label comparison.

[0016] (III) Beneficial Effects This invention provides a fault detection system and method for electronic components, which has the following beneficial effects: By establishing multi-source high-speed data acquisition, lightweight intelligent inference, industrial control closed-loop, and cross-level data integration between the industrial site and the cloud, fault detection and response can be completed within milliseconds. Furthermore, continuous iterative optimization through large-scale historical data analysis can significantly bring the following benefits: By employing multi-source data synchronization and wavelet filtering methods, spurious data can be effectively removed and transient anomalies can be marked, with output buffering. This method transmits data downstream, significantly reducing noise interference while ensuring timing alignment; and it obtains feature vectors by employing kernel mapping and accelerated transformation. Furthermore, it utilizes incremental machine learning algorithms to achieve real-time fault diagnosis and confidence building, ensuring high detection accuracy and adaptability even under the influence of multiple factors such as equipment aging and changes in operating conditions.

[0017] Utilizing multi-level risk indices and execution actions By triggering emergency plans or issuing early warnings in industrial control systems, the potential time window for fault propagation to downtime is significantly shortened; and feedback loops adjust model parameters. Gain continuous correction opportunities to gradually reduce false alarms and false negatives; The system collects and controls records from the field and uploads them to the cloud or factory backend. It then uses aggregation operators to mine fault modes and equipment health trends on a large scale. Finally, it distributes the optimized fault detection model or maintenance strategy to the edge to achieve full lifecycle management through cloud-edge collaboration.

[0018] Overall, this solution can complete fault identification and closed-loop response within milliseconds, meeting the stringent real-time requirements of industrial sites. Furthermore, it can continuously improve the system's adaptability to new fault modes or long-cycle equipment wear through cross-level data integration and online incremental learning, achieving significant improvements in both safety and stability. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the fault detection method for electronic components of the present invention; Figure 2 This is a schematic diagram of the fault detection system for electronic components of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This invention provides a fault detection method for electronic components, including: Step 1: When high-frequency sampling triggers a surge of multi-source data streams, process the original sensor sequence. Wavelet filtering and missing value placeholders are used to remove spurious data and mark transient anomalies. Interpolation is then performed on the aligned and delayed data to generate micro-batch encapsulation. To carry sensor ID, timestamp, and transient suspicious flags ; Step one includes the following: Step 101: Multi-source data alignment and wavelet coefficient-based depth filtering: The raw sensor sequence was acquired. ,in Indicates sensor number , Indicates the first Each sampling time , and The number of sensors and the length of the sampling sequence are respectively determined; all sensor data are aligned by timestamp, and missing values ​​or severely distorted data points are initially marked. For each original sensor sequence To extract the characteristic coefficients at multiple scales, a discrete wavelet transform is performed. Let the selected wavelet basis functions be... Then for In scale With displacement The wavelet coefficients on can be expressed as: ; In the formula: For sensors In scale and displacement Wavelet coefficients on; For wavelet basis functions used in multi-scale decomposition (such as Morlet or other functions suitable for fault feature extraction); This is a scale parameter, and its numerical range can be selected based on the sampling frequency and the actual fault characteristic frequency band. These are displacement parameters, corresponding to translation on the time axis; For the sampling time, satisfying , The sampling interval; By appropriately reconstructing wavelet coefficients and filtering out low-amplitude high-frequency coefficients, high-frequency noise can be reduced while preserving the transient characteristics of the fault as much as possible. The noise components are determined based on the distribution pattern of wavelet coefficients at each scale. When the energy proportion of the target scale segment is consistently higher than expected, it is marked as high-frequency interference and suppression is performed. After performing wavelet filtering, a wavelet energy index is defined to detect high-energy component anomalies. Quickly identify transient anomalies: ; in: For at any time For sensors The wavelet energy index; The selected set of key scales is used to extract the scales that best characterize the failure modes; In wavelet transform, with time... The corresponding displacement; To enhance the contribution of high-energy components to anomalies; When wavelet energy index If a significant surge or a substantial difference from the historical baseline is observed within a certain sliding window, the data at that moment is initially marked as a transient suspicious value and a transient anomaly flag is attached. ; In practice, it is used through multi-scale decomposition of discrete wavelets and wavelet energy index. The introduction of this feature effectively preserves short-term fault symptoms, and multi-sensor time alignment ensures the comparability of different physical quantities at the same moment, providing stable and synchronous input conditions for the lightweight inference in the subsequent step two, with the help of transient anomaly flags. By making suspected fault signs explicit at the earliest stage, a reference label is provided for the incremental learning of the subsequent fault detection model; the wavelet energy index is used. By introducing an online filtering stage, the system can capture fault transient characteristics more sensitively while maintaining its time-frequency localization capability.

[0022] Step 102, Micro-batch Packaging and Timing Marking:

[0023] The reconstructed signal sequence obtained after discrete wavelet transform denoising, resampling alignment, missing value placement, and transient anomaly labeling in step 101 is... The dataset is divided into multiple micro-batch datasets within a time window of a certain length. Each microbatch dataset It can be represented as ,in This represents a micro-batch index; necessary metadata, such as sensor numbers, is recorded at the micro-batch level. Time range Data tagging, etc.; If a transient suspicious flag is added at some point in step 101 Then it is necessary to ensure that in the micro-batch dataset The original energy index is retained internally for these data points. And corresponding flags; this allows for secondary screening in the reasoning model of step two to identify the actual probability of failure, and adds a completeness indicator. This indicates the micro-batch dataset The percentage of valid data; After completing the above packaging, the micro-batch data will be uniformly written into the micro-batch package. : ; in These are the start and end points of a micro-batch; next, these micro-batch packages... The lightweight fault detection model will directly read and call the data in step two. When using it, the preprocessed results are packaged into manageable micro-batch packages. This ensures that subsequent algorithms can quickly access feature tags and necessary contextual information, enabling targeted interpolation or focused analysis during later inference, which can improve the accuracy of fault detection. This is achieved by leveraging high-dimensional wavelet energy exponents and transient suspicious markers. It can promptly mark transient anomalies, thereby improving adaptability to equipment aging or environmental changes.

[0024] Step 2: Extract feature vectors from the edge side using kernel mapping and accelerated transformation. It also calls incremental machine learning algorithms to output fault markers in real time. and confidence level And correct the model parameters of the fault detection model to adapt to environmental disturbances such as equipment aging and sensor drift; Step two includes the following: Step 201: Temporal representation and fast dimensionality reduction based on feature kernel mapping: From micro-batch packaging Reading micro-batch datasets And corresponding tagging information, including transient suspicious tags. Wavelet energy index and the percentage of valid data Wait, for each differential batch dataset For the data, first perform simple interpolation or skip missing values; then obtain the micro-batch dataset. All sensor channels After the wavelet energy index The sequences are concatenated to form an initial feature vector sequence. ; If the current dataset is a micro-batch Include The number of key moments selected for the road sensor is Then, after expanding in the time dimension, we get a length of The temporal concatenation vector is used to retain transient suspicious flags. Location index; To balance real-time performance and strong discriminative power, a customizable kernel function is introduced. For the initial feature vector sequence Perform mapping to obtain the dimensionality-reduced feature representation. Its basic form can be written as: ; in This represents a nonlinear mapping operation based on a kernel function, containing multiple orthogonal basis components to highlight fault-sensitive characteristics and suppress redundant noise; for example, if a composite kernel is selected... ,but It is the feature mapping corresponding to the composite kernel; Adjustable weights for kernel components; can be combined with wavelet energy information. Differentiated weighting is applied to sensor time series to avoid the weakening of transient anomalies by ordinary linear dimensionality reduction; To rapidly perform dimensionality reduction and enhance the reliability of missing data in real-time industrial environments, an accelerated transformation exponent is introduced. Used to dynamically adjust the dimensionality reduction level during kernel mapping: ; In the formula: These are adjustable hyperparameters for the model, used to balance the relationship between the missing rate and the complexity of dimensionality reduction; For micro-batch packaging The percentage of valid data; When the proportion of valid data When the value is low, there are more missing values, accelerating the transformation index. This will also decrease, indicating a need for a more conservative dimensionality retention strategy to avoid excessive dimensionality reduction leading to the loss of key information; when the proportion of effective data... At higher levels, the acceleration conversion index Correspondingly, a more aggressive projection depth can be used in the dimensionality mapping to obtain a more concise feature representation; The feature vector output in the final step 201 can be represented as: ; in, This indicates an element-wise or component-wise weighted operation; eigenvector This will serve as the direct input for real-time fault inference in the subsequent step 202; When used, it is achieved through nonlinear kernel mapping. Transforming multi-source time-series data into a more discriminative feature space allows for the full exploitation of wavelet energy indices. With transient suspicious signs Potential information; aided by the accelerated transformation index Dynamically controlling the dimensionality reduction depth can effectively reduce the impact of missing values ​​on the dimensionality reduction effect and maximize the compression of redundant dimensions when the data integrity is high, thus balancing real-time performance and accuracy; an acceleration transformation exponent is introduced. As a dynamic adjustment coefficient, it is combined with the proportion of effective data. Flexible control of kernel mapping can meet the real-time requirements of fluctuating industrial environments.

[0025] Step 202, Real-time Fault Reasoning and Incremental Update: In the eigenvector Based on this, a lightweight classification or regression model (such as an improved random forest or incremental deep network) is used to quickly predict whether a fault exists and its severity. Let the model parameters of the fault detection model be... For the current microbatch dataset eigenvectors within Prediction results It can be represented as: ; in This represents a decision function, and the output can be a failure probability, risk score, or specific failure label. If only relying on prediction results Binary or multi-class classification often struggles to distinguish between the different levels of high-risk faults and minor anomalies. Therefore, a fault confidence level is added to the model's output layer. The definition is as follows: ; in, For micro-batch packaging The corresponding fault detection results include the fault confidence level, which is used to quantify the fault risk at the current moment (or the current micro-batch packaging). The range of values ​​is within Smooth, monotonic activation functions (such as the Logistic function) are used to map the internal weighted sum to the confidence interval; The feature vector is a compact representation obtained after kernel mapping and accelerated transformation of information such as multi-source time series data, time-frequency energy index, and missing value placeholders. This is a square matrix used to characterize the coupling relationship between different feature channels and fault modes, typically... Combined with exponential mapping For eigenvectors Applying nonlinear transformations amplifies or suppresses fault symptoms in a high-dimensional space, thereby improving the accuracy and discriminative power of confidence assessment. For control matrix The scaling factor in exponential mapping. It can be positive, negative, or zero; This refers to matrix exponentiation, i.e.: ; To address different index channels (i.e., different...) The weighting coefficients for linear superposition of the transformation branches can be any real number. This represents the number of principal activation units selected by the model in the output layer. When the fault confidence When the confidence threshold is exceeded, it indicates that the current micro-batch dataset... The system state it represents is highly likely to be at high risk, requiring rapid implementation of emergency control strategies; if the fault confidence level... If the preset confidence threshold is not exceeded, you can choose to issue a warning or record the information. To adapt to aging equipment or changes in operating conditions, incremental or online learning is required in the industrial field. The core logic is: If a genuine fault label or a further confirmed fault sample is detected, its labeling information can be sent back to step 202 to adjust the model parameters. A simplified form for performing small-batch or online incremental updates can be written as: ; in, For learning rate, This represents a composite loss function, consisting of fault classification focus loss and confidence regression smoothing loss. The true labels obtained in subsequent stages or after a slight delay are used to fine-tune or correct the current model; If the model adopts a tree structure, the subtree can be expanded or the leaf node statistics can be updated as needed when high-risk fault samples appear; if a neural network is used, the weights can be fine-tuned based on mini-batch gradient descent or adaptive optimization algorithms, thereby maintaining the model's sensitivity to system changes without significantly increasing the computational load. In practice, the lightweight design and prediction workflow for micro-batch datasets ensure rapid inference, meeting the stringent real-time requirements of industrial scenarios; through fault confidence... Quantitatively characterizing fault risk facilitates the flexible triggering of control strategies at different levels when obtaining detection results. A fault confidence level is introduced into the model output layer to assess fault severity. This helps industrial control systems perform precise scheduling based on different risk levels.

[0026] Step 3: On-site PLC or DCS based on multi-level risk index Perform high-priority actions such as emergency shutdown and redundancy switchover, or issue a moderate warning, and record the actual execution. and fault labels The data is fed back to the fault detection model to form a closed-loop protection system, and the model parameters are optimized based on the actual fault confirmation process. Step three includes the following: Step 301: Generation of control instructions based on multi-level risk indices: Receive the test results from step two. ,in Indicates the current microbatch dataset Or the prediction result of the time slice, that is, the fault type or fault flag. This indicates the fault confidence level, representing the severity of the fault. The above results are transmitted to the field PLC or DCS via real-time communication protocols such as OPCUA, Modbus, or Industrial Ethernet. To enable precise control over different fault levels and system states, a multi-level risk index is introduced. Taking into account the confidence level of the fault and the current critical state of the system (For example, temperature, pressure, flow rate, and process parameters mentioned above that are not directly used for fault detection but are crucial for control). An example definition is as follows: ; Indicates micro-batch packaging At the corresponding time or time window, combined with the fault confidence level Compared with the current process status The resulting risk assessment amount; The fault confidence level of the fault detection model; This represents the key process state vector of the industrial system at the current micro-packaging batch stage, which may include different physical quantities such as temperature, pressure, flow rate, and torque. It is typically set as follows: for Column vectors; For matrix exponentiation, it means: ; To transfer the matrix exponent from Points to The result, in form, is The matrix, as Changes with the changes; To control the weight matrix, it is usually... Dimension, and It shares the same matrix dimension as the integral term; For trace operation, for Sum of the diagonal elements of a matrix; Based on multi-level risk index By determining the threshold values ​​within different ranges, the industrial control system can decide to execute the corresponding level of protection or alarm, as follows: If multi-level risk index Exceeding the high-risk threshold Immediately trigger emergency shutdown, backup equipment switching, or other highest-level protection measures: if there are multiple risk indices Between the median threshold With high risk threshold In between, an early warning notification is generated for the operator, and the system enters a protective load reduction or redundancy preheating mode; if there are multiple risk indices... Less than the low risk threshold If it is, it is considered a normal or minor risk, and only needs to be recorded or mildly intervened; Control commands are executed via PLC or DCS, directly issuing action commands to the production line equipment, such as valve opening adjustment, motor speed adjustment, and cooling system startup, achieving millisecond-level connection from fault detection to actual action.

[0027] In use, leveraging real-time communication protocols and the hardware characteristics of on-site PLCs / DCS, fault model outputs can be rapidly converted into specific action instructions, shortening the potential time window for fault propagation and equipment damage, and increasing the fault confidence level. By embedding matrix exponentiation and trace operations, a more feature-rich multi-level risk index can be formed. Based on the system status, high-risk, medium-risk, and low-risk can be distinguished.

[0028] Step 302: Feedback on execution results and incremental strategy iteration: After executing the control command, the PLC / DCS will send back the actual actions of the system (such as adjusting the valve opening, switching the standby equipment, etc.) and the subsequently collected operating status data to this step 302; The actions will be recorded using timestamps. With fault detection results Perform pairing and record the newly generated system state. Save, among It reflects the key process parameters within several sampling periods after the execution of control; If, after a period of observation or manual inspection, the previously identified fault markers are found... If there are deviations or insufficient or excessive assessments of fault levels, the true labeling needs to be corrected. Feedback is sent to the second step (especially step 202) to perform parameter updates in order to correct the judgment capability of the lightweight fault detection model.

[0029] If multiple responses to the same fault type are ineffective (such as multiple instances of unintended or delayed shutdowns), multi-level risk indices can be recorded locally or in the cloud. The relationship with the final process result allows for further fine-tuning of the weight matrix described in step 301. Benchmark matrix B or corresponding risk threshold To reduce the risk of failure in similar scenarios later; this optimization process can be completed on a local edge device (industrial PC), or it can be transmitted to the cloud or factory back-end system via network for large-scale data analysis (corresponding to step 4), and then the adjusted configuration is distributed to the field to achieve iterative optimization integrating cloud and edge; to ensure consistency with the labeling information in the previous steps, any new labels produced in step 302 need to be assigned... and system status Plus the same batch of index micro-batch Or timestamp labels to avoid data silos.

[0030] By comparing multiple risk indices Calculated values ​​vs. actual execution results The system determines whether the control actions are timely and effective, and improves subsequent responses accordingly, enabling rapid evaluation of real-time control performance. If an actual fault label is confirmed or a response failure is detected, the fault detection algorithm in the second step can be updated, and the threshold or weight matrix in step 301 of the third step can be corrected, forming a closed loop of detection-decision-relearning. Industrial sites often have variable operating conditions. Through the incremental feedback in step 302, the system can adapt to long-term changes such as the access of new equipment, changes in production formulas, or equipment aging, and can be differentiated in different industrial processes, thereby maintaining efficient and accurate fault detection and control capabilities.

[0031] Step 4: Collect micro-batch packaging and fault marking data Fault confidence After obtaining multi-level risk index information, the cloud or backend system collects multi-source data and uses aggregation operations to uncover complex fault modes. The optimized fault detection model and maintenance strategy are then distributed to the edge, and the fault tags are updated accordingly. Compare the evolution of the execution model; Step four includes the following: Step 401: Multi-source data aggregation and visualization analysis: Fault detection records accumulated on field edge devices or industrial PCs (such as...) Fault diagnosis Fault confidence Multi-level risk index and the actual execution information of the control system. (etc.) are transmitted periodically or in real time to the cloud or factory back-end system to complete the big data collection; Uploaded data is stored in a distributed database and encapsulated according to micro-batch processing. Establish a search index for process equipment IDs (such as sensor IDs or PLC numbers); in addition to fault diagnosis information and control response data, external data sources such as equipment maintenance history, operator comments, and equipment operating condition change records can also be integrated. Construct a global aggregation operator that integrates massive historical data with key on-site parameters. In terms of form, it can be written as: ; In the formula: A set of indexes for historical micro-batch encapsulations or time slices; To include current micro-batch packages The multidimensional matrix representation merged in the cloud (e.g., combining...) (and information such as the status of external devices); This is a weighting factor, related to fault severity or data freshness; It is a global coupling matrix used to measure the collaborative patterns of information across multiple devices and time periods during cross-batch analysis; The control matrix used for weighting large-scale historical data can be optimized offline in the cloud or specified manually; This is the matrix trace operation, used to compress the result of a composite matrix into a scalar; Such high-order aggregation can amplify the commonalities of failure modes, suppress noise, and provide a global data foundation for the predictive maintenance strategy in step 402 in large-scale samples.

[0032] In cloud-based or factory back-end systems, visual dashboards can be built to display real-time and historical trends of each production line or equipment, including failure frequency, downtime, and risk level distribution; specific indicators (such as...) can also be analyzed. , It can perform heat map or time series overlay analysis; it provides interactive exploration functions for operators or managers, enabling in-depth mining and comparison of specific time periods or fault types, laying a cognitive foundation for subsequent maintenance and scheduling decisions; When used, the detection results and control logs scattered on the edge side and the local control system are transformed into a large-scale queryable data asset, avoiding information silos and supporting long-term retention. By integrating external data with fault detection results, a richer context can be formed in the cloud, which helps to accurately locate the root cause of the fault and identify potential hazards. Step 402, Predictive Maintenance and Model Iteration throughout the entire lifecycle: After big data collection is completed in step 401, a comprehensive approach using machine learning, signal processing, or domain knowledge (such as equipment health modeling) is used to deeply explore historical failure cases, failure precursor characteristics, and control response effects. With the remaining service life of the equipment or the fatigue of key components as the target, long-term fault detection models can be trained or updated in the cloud, such as those based on time-series recursive networks, graph models, or other advanced algorithms suitable for engineering fields. When a new fault mode is discovered (such as a combination of sensor waveforms that has never been seen before at the edge) or the evolution trend of an existing mode (such as faults gradually occurring more frequently in a specific temperature range), it can be recorded and included in the corresponding maintenance strategy library. If the cloud analysis results indicate that the fault detection model (including feature extraction and inference algorithms) used in the second step has significant room for optimization, the improved version can be packaged and distributed to the field edge equipment; combined with the equipment's remaining service life prediction results and fault mode analysis, a tiered maintenance plan can be developed, such as replacing vulnerable parts in advance, scheduling maintenance periods, and optimizing spare parts inventory. Unlike the real-time closed-loop control in step three, which focuses on emergency protection, predictive maintenance here leans towards medium- to long-term planning. This avoids unnecessary downtime or excessive repairs, thereby improving equipment utilization. When in use, leveraging the accumulation of data throughout the entire lifecycle, it is possible not only to better predict risks before failures occur, but also to extend equipment life and reduce downtime losses. The combination of cloud-based big data analysis and edge incremental updates allows the fault detection model to continuously adapt to changes in the industrial environment, reducing false alarms, missed alarms, and delayed alarms. Predictive maintenance reduces unnecessary expenses caused by regular large-scale overhauls and avoids repeated emergency shutdowns caused by relying solely on real-time control. By directly linking the in-depth fault mode analysis and equipment remaining life prediction in the cloud to the update of the fault detection model on the edge side, it achieves seamless integration of macro-level big data and local micro-batch detection, breaking through the limitations of traditional processing that is only done locally or only in the cloud.

[0033] Please see Figure 2 This invention provides a fault detection system for electronic components, comprising, The data acquisition unit, when high-frequency sampling triggers the influx of multi-source data streams, applies wavelet filtering and missing value placeholders to the original sensor sequence to remove pseudo data, marks transient anomalies, and performs interpolation operations on the aligned delayed data to generate micro-batch packages to carry sensor ID, timestamp, and transient suspicious flags. The feature extraction unit extracts feature vectors at the edge using kernel mapping and accelerated transformation, and calls an incremental machine learning algorithm to output fault labels and fault confidence in real time, and corrects the model parameters of the fault detection model to adapt to environmental disturbances. The early warning unit, on-site PLC or DCS, executes high-priority actions or issues medium-level early warnings based on multi-level risk indices, and feeds back the actual execution records and fault labels to the fault detection model, and optimizes the model parameters based on the actual fault confirmation process; The feedback iteration unit collects micro-batch packaging, fault marking, fault confidence, and multi-level risk index information. The cloud or back-end system collects multi-source data and uses aggregation operations to mine fault patterns. The optimized fault detection model and maintenance strategy are then distributed to the edge side, and the model iteration is performed based on fault label comparison.

[0034] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0035] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0036] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0037] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0038] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method of detecting a failure of an electronic component, characterized by: Comprising, When high-frequency sampling triggers multi-source data flow influx, wavelet filtering and missing value placeholder are applied to the original sensor sequence to remove false data, mark transient anomalies, and perform interpolation on delayed data to generate micro-batch packages to carry sensor ID, timestamp, and transient suspicious flag; The edge side uses kernel mapping and acceleration transformation to extract feature vectors, and calls incremental machine learning algorithm to output fault label and fault confidence in real time, and corrects the model parameters of the fault detection model to adapt to environmental disturbances; The field PLC or DCS executes high-priority actions or issues moderate early warnings based on multi-level risk indexes, and records actual execution and fault labels back to the fault detection model, and optimizes model parameters based on real fault confirmation process; After collecting micro-batch packages, fault labels, fault confidence, and multi-level risk index information, the cloud or background system aggregates multi-source data and mines fault patterns with the help of aggregation operations, and then issues the optimized fault detection model and maintenance strategy to the edge side, and performs model iteration under fault label comparison.

2. The electronic component fault detection method of claim 1, wherein: After obtaining the original sensor sequence, align it by timestamp, extract feature coefficients at multiple scales by discrete wavelet transform, and mark high-frequency interference and perform suppression when the energy proportion of the target scale segment is continuously higher than expected; Transient anomalies are identified by wavelet energy index for detecting high-energy component anomalies, and a transient anomaly flag is added.

3. The electronic component fault detection method of claim 2, wherein: The reconstructed signal sequence is divided into multiple micro-batch data sets, and after recording metadata at the micro-batch level, the wavelet energy index and corresponding flag are retained for the corresponding data points within the micro-batch data set, and the micro-batch package is generated by summarizing.

4. The electronic component fault detection method of claim 3, wherein: After simple interpolation or skip processing of missing values, all sensor channels in the micro-batch data set are obtained and spliced with the wavelet energy index to form an initial feature vector sequence; A self-definable kernel function is introduced to map the initial feature vector sequence to obtain a reduced feature representation, and an acceleration transformation index is introduced to dynamically adjust the dimensionality reduction strength in the kernel mapping process, and the output feature vector.

5. The electronic component fault detection method of claim 4, wherein: Based on the feature vector, a lightweight fault detection model is used to quickly predict whether there is a fault and its severity, and the model parameters are updated incrementally in small batches or online based on real feedback; A fault confidence is added to the model output layer, and when the fault confidence exceeds the preset confidence threshold, an emergency control strategy is quickly executed, and if the fault confidence does not exceed the preset confidence threshold, only a warning or record is selected.

6. The electronic component fault detection method of claim 5, wherein: The obtained detection results are transmitted to the field PLC or DCS; And the introduction of fault confidence and the current system key state of multi-level risk index, according to the multi-level risk index in different threshold interval, industrial control system executes the corresponding level of protection or alarm, and through the PLC or DCS execution, directly to the production line equipment to issue action instruction.

7. The electronic component fault detection method according to claim 6, wherein: After the PLC or DCS returns the actual action and running state data of the system, the execution action and the fault detection result are paired, and the newly generated system state is: If the previous fault label has deviation or is insufficient or excessive in judging the fault level, the real label is used to correct the lightweight fault detection model; if the response effect for the same fault type is not good for multiple times, the relationship between the multi-level risk index and the final process result is used to fine-tune the related parameters of the multi-level risk index.

8. The electronic component fault detection method according to claim 7, wherein: The fault detection records accumulated on the field edge device or industrial PC are transmitted to the cloud or factory background system periodically or in real time; the global aggregation operator for comprehensive measurement of massive historical data and field key parameters is constructed, and after data aggregation, a visual panel is constructed in the cloud or factory background system.

9. The electronic component fault detection method according to claim 8, wherein: The fault detection model is updated with the device remaining service life or key component fatigue as the target; if there is a fault mode or an evolution trend of the existing mode, it is recorded and included in the corresponding maintenance strategy library; If there is optimization space for the used fault detection model, the improved version can be packaged and distributed to the field edge device; Combined with the prediction result of the device remaining service life and the fault mode analysis, a hierarchical maintenance plan is developed.

10. A fault detection system for electronic components, characterized in that: Including, The data acquisition unit, when high-frequency sampling triggers multi-source data flow to rush in, wavelet filtering and missing value placeholder are used on the original sensor sequence to remove false data and mark transient anomalies, and interpolation operation is performed on the delayed late data to generate micro-batch packaging to carry sensor ID, timestamp and transient suspicious flag; The feature extraction unit extracts feature vectors on the edge side using kernel mapping and acceleration transformation, and calls incremental machine learning algorithm to output fault label and fault confidence in real time, and corrects the model parameters of the fault detection model to adapt to environmental disturbance; The early warning unit executes high-priority action or issues moderate early warning according to the multi-level risk index, and returns the actual execution record and fault label to the fault detection model, and optimizes the model parameters according to the real fault confirmation process; The feedback iteration unit collects micro-batch packaging, fault label, fault confidence, multi-level risk index information, and the cloud or background system collects multi-source data and uses aggregation operation to mine fault mode, and then distributes the optimized fault detection model and maintenance strategy to the edge side, and executes model iteration under the fault label comparison.

Images