High-end equipment industry big data analysis method and system

Abstract

The invention relates to the technical field of industrial big data analysis, and discloses a high-end equipment industrial big data analysis method and system, and the method comprises the steps: receiving sensor data flows and real-time operation parameters of different sensors of high-end equipment, and aggregating the sensor data flows of the different sensors into a physical event packet; determining an expected environmental noise feature based on the real-time operating parameters; performing signal separation processing on the physical event packet according to expected environmental noise characteristics, and extracting potential abnormal components independent of environmental noise; performing cross-correlation analysis on the potential abnormal components in a preset time window, and calculating a comprehensive fault confidence coefficient according to a preset physical association rule; and when the comprehensive fault confidence exceeds a preset threshold, generating a fault early warning. According to the method, through multi-source data fusion, noise suppression and dynamic confidence evaluation, the transformation from post-maintenance to pre-warning is realized, and the accuracy and efficiency of predictive maintenance of high-end equipment are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of industrial big data analysis, and more specifically, to a method and system for analyzing big data in the high-end equipment industry. Background Technology

[0002] Modern high-end equipment generates a large amount of diverse data during operation. This data often doesn't align in time, making it difficult for analysis systems to accurately determine if equipment malfunctions, thus affecting the timeliness of maintenance. These sensors are strategically deployed in key areas of the machining center to collect massive amounts of heterogeneous data in real time during equipment operation. To effectively fuse and analyze this data from different physical quantities and sampling frequencies, the system employs a rigorous data processing logic, the core of which is ensuring precise alignment of data from different sources along the timeline.

[0003] However, in real-world long-term operating environments, these sensors, distributed across various critical parts of the machining center, while all connected to the same data acquisition network via industrial Ethernet, each possess independent internal timing circuits. These timing circuits typically rely on quartz crystal oscillators to provide a time reference. During long-term continuous operation, due to various practical factors, the frequencies of these internal timing circuits will experience complex and non-linear deviations. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention discloses a big data analysis method and system for high-end equipment manufacturing, aiming to solve the problem of inaccurate and untimely fault warnings caused by time misalignment of sensor data streams in existing big data analysis for high-end equipment manufacturing.

[0005] The technical solution of the present invention is as follows: In a first aspect, this invention discloses a big data analysis method for high-end equipment manufacturing, comprising: It receives sensor data streams and real-time operating parameters from different sensors in high-end equipment, and aggregates the sensor data streams from different sensors into physical event packets. Based on real-time operating parameters, the expected environmental noise characteristics are determined; Based on the expected environmental noise characteristics, the physical event packets are processed for signal separation to extract potential anomalous components independent of environmental noise. Cross-correlation analysis is performed on potential anomalous components within a preset time window, and the comprehensive fault confidence is calculated based on preset physical correlation rules. When the overall fault confidence exceeds a preset threshold, a fault warning is generated.

[0006] This invention effectively solves the problem of inaccurate and untimely fault warnings caused by time misalignment of sensor data in the prior art. By using multi-source data fusion, noise separation, and cross-correlation analysis, it improves the accuracy of fault identification and the timeliness of warnings, thereby avoiding unplanned downtime and high maintenance costs.

[0007] Furthermore, the steps for receiving sensor data streams and real-time operating parameters from different sensors in high-end equipment include: The edge processing unit deployed next to the high-end equipment receives sensor data streams from different sensors, including vibration signals, current signals, temperature signals and pressure signals from the high-end equipment. The equipment control unit of the high-end equipment receives real-time operating parameters, including machining step type, spindle speed, feed rate, depth of cut and machining material type.

[0008] Based on this, the steps to aggregate sensor data streams from different sensors into physical event packets include: The system uses a synchronization network to adaptively correct the clock skew of sensor data streams from different sensors, and aggregates the corrected sensor data streams into physical event packets according to preset physical association rules.

[0009] Furthermore, the steps for determining the expected environmental noise characteristics based on real-time operating parameters include: Based on real-time operating parameters, a preset mapping relationship between operating parameters and noise features is queried to determine the environmental noise baseline characteristics of sensor data streams from different sensors corresponding to the real-time operating parameters, which are then used as the expected environmental noise characteristics.

[0010] Furthermore, based on the expected environmental noise characteristics, the steps of performing signal separation processing on the physical event packets to extract potential anomalous components independent of environmental noise include: Independent component analysis or blind source separation techniques are used to separate physical event packets into multiple source signal components; The separated source signal components are compared with the expected environmental noise characteristics to identify and extract components that deviate from the environmental noise as potential anomalous components.

[0011] As a technological improvement, the steps for performing cross-correlation analysis on potential anomalous components within a preset time window include: Identify sensor data streams from multiple sensors associated with potential anomalous components; Within a preset time window that includes potential anomalous components, calculate the cross-correlation function between the sensor data streams of multiple sensors; The time delay relationship of sensor data streams from multiple sensors is determined based on the time offset corresponding to the maximum value of the cross-correlation function.

[0012] To improve the solution, the steps for calculating the comprehensive fault confidence based on preset physical association rules include: Determine whether the time delay relationship conforms to the preset physical association rules; If the time delay relationship conforms to the physical association rule, the comprehensive fault confidence of the potential anomalous components is calculated according to the preset confidence calculation mechanism.

[0013] For specific situations, when the overall fault confidence exceeds a preset threshold, the steps for generating a fault warning include: When the overall fault confidence exceeds a preset threshold, the corresponding fault mode is matched according to the overall fault confidence. Based on the fault mode, generate corresponding fault warnings.

[0014] As a further improvement, the steps for generating corresponding fault warnings based on the identified fault modes include: The system queries a pre-defined set of fault modes and maintenance experience associations to generate fault warnings that include affected components, recommended inspection steps, necessary tool and spare parts information, and maintenance personnel skill suggestions.

[0015] Secondly, this invention also discloses a high-end equipment industry big data analysis system, comprising: The data receiving and processing module is used to receive sensor data streams and real-time operating parameters from different sensors of high-end equipment, and to aggregate the sensor data streams from different sensors into physical event packets. The feature determination module is used to determine the expected environmental noise characteristics based on real-time operating parameters; The anomaly extraction module is used to perform signal separation processing on physical event packets based on the expected environmental noise characteristics, and to extract potential abnormal components independent of environmental noise. The fault confidence calculation module is used to perform cross-correlation analysis on potential abnormal components within a preset time window, and calculate the comprehensive fault confidence based on preset physical correlation rules. The early warning module is used to generate a fault warning when the overall fault confidence exceeds a preset threshold.

[0016] This invention provides a system-level solution through modular design, which automates the processes of data reception, feature determination, anomaly extraction, fault confidence calculation, and fault early warning. This effectively supports the implementation of the above methods and improves the intelligence level of fault diagnosis for high-end equipment.

[0017] In summary, this invention provides a method and system for big data analysis in high-end equipment manufacturing. The method receives sensor data streams and real-time operating parameters from different sensors in high-end equipment and aggregates them into physical event packets, effectively solving the problem of ineffective fusion of multi-source heterogeneous data in existing technologies. Based on this, the invention determines the expected environmental noise characteristics based on real-time operating parameters and uses these characteristics to perform signal separation processing on the physical event packets, thereby accurately extracting potential abnormal components independent of environmental noise. This step overcomes the interference of environmental noise on abnormal signal identification in traditional methods, significantly improving the sensitivity and accuracy of anomaly detection. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a big data analysis method for high-end equipment manufacturing provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the structure of a high-end equipment industry big data analysis system provided in an embodiment of the present invention.

[0020] Labeling Explanation: 210, Data Receiving and Processing Module; 220, Feature Determination Module; 230, Anomaly Extraction Module; 240, Fault Confidence Calculation Module; 250, Early Warning Module. Detailed Implementation

[0021] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some, not all, of the embodiments of this invention. The components of this invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0022] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] During long-term operation, traditional high-end equipment experiences a continuous and worsening nonlinear clock frequency deviation between data streams from different sensors due to a combination of factors, including inherent differences in the timing circuits within the sensors, fluctuations in ambient temperature, and changes in power supply voltage. This time misalignment makes it difficult for data fusion analysis to accurately reflect the true causal relationships of physical events. Consequently, the accuracy and confidence of predictive models in identifying early failure modes decrease significantly, and they may even fail to identify the true fault signals masked by the time misalignment. This leads to missed optimal maintenance opportunities, resulting in unplanned downtime and high repair costs.

[0024] Firstly, please see Figure 1 This invention proposes a big data analysis method for high-end equipment manufacturing, including: S1. Receive sensor data streams and real-time operating parameters from different sensors of high-end equipment, and aggregate the sensor data streams from different sensors into physical event packets. S2. Determine the expected environmental noise characteristics based on real-time operating parameters; S3. Based on the expected environmental noise characteristics, perform signal separation processing on the physical event packet to extract potential abnormal components independent of environmental noise; S4. Perform cross-correlation analysis on potential abnormal components within a preset time window, and calculate the comprehensive fault confidence based on preset physical correlation rules; S5. When the overall fault confidence exceeds the preset threshold, a fault warning is generated.

[0025] Among them, "sensor data stream" refers to the data sequence continuously collected and transmitted by various sensors deployed on high-end equipment. "Real-time operating parameters" refers to the operating condition information of high-end equipment at a specific moment. "Physical event package" refers to the aggregation of sensor data streams from different sensors, after time synchronization and preliminary processing, according to preset physical association rules, forming a data set that can characterize a specific physical event. "Expected environmental noise characteristics" refers to the expected noise pattern in the environment where the high-end equipment is located under specific operating parameters. "Potential anomaly components" refer to signal components separated from the physical event package, independent of environmental noise, that may indicate equipment failure. "Cross-correlation analysis" is a signal processing technique used to measure the similarity of two signals at different time delays, thereby revealing their causal relationship or time synchronization. "Physical association rules" refers to a set of rules established based on the physical structure, working principle, and fault mechanism of high-end equipment, used to determine whether the correlation between different sensor data conforms to a preset fault mode. "Comprehensive fault confidence" refers to a quantitative indicator that assesses the probability or severity of a fault after analyzing potential anomaly components. "Fault warning" refers to a notification or alarm issued by the system when the overall fault confidence level reaches a certain level, indicating that the equipment may fail.

[0026] This invention proposes a big data analysis method for high-end equipment manufacturing. Its overall working principle lies in constructing a closed-loop intelligent analysis process from data acquisition to fault early warning, aiming to solve the problem of fault identification caused by data time misalignment and environmental noise interference in existing technologies. First, by receiving sensor data streams from different sensors of high-end equipment and real-time operating parameters, a comprehensive data foundation is provided for subsequent analysis. Specifically, aggregating sensor data streams from different sensors into physical event packets—this step, through preliminary event correlation and time synchronization of heterogeneous data—lays the foundation for subsequent signal separation and effectively alleviates the problem of inconsistent timestamps in the original data.

[0027] Secondly, determining the expected environmental noise characteristics based on real-time operating parameters is one of the key innovations of this invention. By matching or predicting the current operating conditions with preset noise characteristics, the system can dynamically understand the normal noise background in the current environment. This step allows subsequent anomaly detection to move beyond simply looking for absolute changes in the signal, and instead look for deviations in the signal relative to the current normal noise background, thereby significantly improving the sensitivity and accuracy of anomaly detection.

[0028] Next, the physical event packet undergoes signal separation processing based on the expected environmental noise characteristics to extract potential anomalous components independent of environmental noise. This step utilizes advanced signal processing techniques to decompose the mixed signal in the physical event packet into multiple source signals, and by comparing them with the expected environmental noise characteristics, accurately identifies and isolates those anomalous signals that truly indicate a fault. This effectively avoids interference from environmental noise in fault identification, allowing the system to focus on analyzing genuine anomalous information.

[0029] Subsequently, cross-correlation analysis is performed on potential anomaly components within a preset time window, and the overall fault confidence is calculated based on preset physical correlation rules. This step verifies the physical rationality of the abnormal signals by analyzing the time delay relationship between data streams from different sensors, combined with the physical structure and fault propagation mechanism of high-end equipment. For example, if a vibration anomaly first appears in the bearing and then on the motor housing, and the time delay matches the propagation path of a bearing fault, then its fault confidence will be significantly improved. This combination of multi-source data cross-correlation analysis and physical correlation rules greatly enhances the reliability and accuracy of fault diagnosis, avoiding false alarms that may be caused by a single sensor anomaly.

[0030] Finally, when the overall fault confidence level exceeds a preset threshold, a fault warning is generated. This step transforms the quantified fault confidence level into actionable warning information and matches the corresponding fault mode based on the confidence level to generate a detailed warning that includes affected components, recommended inspection procedures, necessary tool and spare parts information, and maintenance personnel skill suggestions. This not only achieves early warning of faults but also provides specific maintenance guidance, thus achieving a seamless transition from "problem discovery" to "problem resolution," significantly improving the predictive maintenance capabilities of high-end equipment.

[0031] The big data analysis method for high-end equipment industry proposed in this invention demonstrates significant technological progress and innovation in solving the problem of early fault identification and early warning for high-end equipment compared with existing technologies. Traditional methods often struggle to effectively handle the time misalignment problem of multi-source heterogeneous sensor data and lack robustness to environmental noise, resulting in low fault identification accuracy and high false alarm rate, thus missing the optimal maintenance opportunity.

[0032] This invention effectively overcomes the challenges of time misalignment, environmental noise interference, and inaccurate fault identification faced by existing technologies in big data analysis of high-end equipment industry by introducing adaptive data aggregation, dynamic noise feature recognition, advanced signal separation, and a fault confidence calculation mechanism that combines multi-source data cross-correlation analysis with physical association rules. It significantly improves the accuracy, timeliness, and guidance of fault early warning, and provides strong technical support for predictive maintenance of high-end equipment.

[0033] Specifically, the steps for receiving sensor data streams and real-time operating parameters from different sensors in high-end equipment include: The edge processing unit deployed next to the high-end equipment receives sensor data streams from different sensors, including vibration signals, current signals, temperature signals and pressure signals from the high-end equipment. The equipment control unit of the high-end equipment receives real-time operating parameters, including machining step type, spindle speed, feed rate, depth of cut and machining material type.

[0034] Edge processing units are deployed alongside high-end equipment, primarily for near-source data acquisition and preliminary processing. This deployment effectively reduces data transmission latency and alleviates the load on the central server, ensuring real-time and efficient data acquisition. Sensor data streams from different sensors directly reflect the operational status of the high-end equipment. Specifically, vibration signals can reveal wear, loosening, or imbalance in mechanical components; current signals can reflect motor load, electrical faults, or abnormal energy consumption; temperature signals can indicate overheating, cooling system failure, or abnormal friction; and pressure signals are commonly used to monitor the stress conditions in hydraulic systems, pneumatic systems, or during processing. These multi-dimensional data collectively constitute a comprehensive understanding of the operational status of the high-end equipment.

[0035] Furthermore, the equipment control unit is the core control system of high-end equipment, providing precise real-time operating parameters. These parameters are crucial for understanding the workload and operating conditions of high-end equipment. The type of machining step determines the equipment's operating mode and expected behavior; spindle speed and feed rate directly affect the cutting process and machining efficiency; depth of cut and material type are closely related to machining load and tool wear. These parameters provide important contextual information for subsequent determination of environmental noise characteristics and identification of potential anomalous components.

[0036] The solution of this invention, by deploying an edge processing unit next to high-end equipment, achieves real-time and efficient acquisition of data streams from various sensors, including vibration, current, temperature, and pressure signals, ensuring the richness and timeliness of the data source. Simultaneously, it acquires precise real-time operating parameters such as machining step type, spindle speed, feed rate, depth of cut, and material type through the equipment control unit, providing crucial background information for subsequent data analysis. This data reception mechanism ensures that the acquired data is not only comprehensive but also closely correlated with the actual operating status of the equipment, laying a solid foundation for accurately distinguishing normal operating noise from potential abnormal signals.

[0037] In some embodiments of the present invention described above, a method is proposed to aggregate sensor data streams from different sensors into physical event packets. However, in actual industrial big data analysis, due to potential clock skew between different sensors, directly aggregating sensor data streams may lead to inconsistent data timestamps, thereby affecting the accuracy of the physical event packets and reducing the reliability of subsequent fault analysis.

[0038] In response, this invention further proposes a step for aggregating sensor data streams from different sensors into a physical event packet, including: The system uses a synchronization network to adaptively correct the clock skew of sensor data streams from different sensors, and aggregates the corrected sensor data streams into physical event packets according to preset physical association rules.

[0039] Specifically, a synchronization network can be understood as a distributed or centralized clock synchronization mechanism, aiming to ensure the consistency of sensor data streams from different sensors in the time dimension. For example, technologies such as Network Time Protocol (NTP) or Precision Time Protocol (PTP) can be used to achieve high-precision time synchronization between sensors. Adaptive clock skew correction refers to the system's ability to dynamically adjust and compensate for real-time sensor clock differences to eliminate time inconsistencies caused by sensor clock drift or network transmission delays. Its purpose is to ensure that all sensor data is precisely aligned to a unified time base before aggregation. Predefined physical association rules can refer to rules predefined based on the structure, working principle, and fault modes of high-end equipment, used to describe the physical correlation between different sensor data. For example, when the spindle vibrates abnormally, it may be accompanied by fluctuations in spindle motor current and an increase in bearing temperature. These correlations can be defined as physical association rules to guide data aggregation. Through these rules, time-corrected related sensor data streams can be logically combined into event packages with actual physical meaning, such as a "spindle abnormality event package" or a "feed system fault event package."

[0040] The present invention effectively solves the clock skew problem that may exist between different sensor data streams by introducing a synchronization network and adaptive clock skew correction. Specifically, the synchronization network ensures that all sensor data streams have a unified reference in the time dimension, while the adaptive clock skew correction further compensates for the clock drift of the sensors themselves or the network transmission delay, making the corrected sensor data streams highly aligned in time. It is precisely because of this precise time alignment that the subsequent process of aggregating sensor data streams into physical event packets according to preset physical association rules can accurately reflect the real physical state and events of high-end equipment. Thus, data misjudgment or information loss caused by time misalignment can be avoided, laying a solid data foundation for subsequent anomaly component extraction and fault diagnosis.

[0041] Specifically, the steps for determining the expected environmental noise characteristics based on real-time operating parameters include: Based on real-time operating parameters, a preset mapping relationship between operating parameters and noise features is queried to determine the environmental noise baseline characteristics of sensor data streams from different sensors corresponding to the real-time operating parameters, which are then used as the expected environmental noise characteristics.

[0042] The mapping database between operating parameters and noise characteristics can be understood as a pre-established database or lookup table that stores typical environmental noise characteristics corresponding to different sensors (such as vibration sensors, current signals, temperature signals, and pressure signals) under different operating parameters of high-end equipment (such as machining step type, spindle speed, feed rate, depth of cut, and type of machining material). This mapping database can be constructed and continuously updated through historical data analysis, experimental testing, or expert experience. Environmental noise baseline characteristics refer to the inherent noise patterns or spectral characteristics in sensor data caused by environmental factors (such as equipment vibration, surrounding mechanical noise, and electrical interference) when high-end equipment is in normal working condition under specific operating parameters.

[0043] The present invention, by querying a pre-defined mapping database of operating parameters and noise characteristics, can accurately obtain the environmental noise baseline characteristics matching the current operating conditions of the high-end equipment based on its real-time operating parameters. This method of determining noise characteristics based on operating parameters avoids the inaccuracies that may arise from simply using a fixed noise model or real-time noise estimation, because the environmental noise patterns and intensities generated by high-end equipment vary significantly under different operating parameters. This approach ensures that the determined expected environmental noise characteristics accurately reflect the noise situation under the current operating conditions, providing an accurate benchmark for subsequent signal separation processing.

[0044] Specifically, the steps described above for performing signal separation processing on physical event packets based on expected environmental noise characteristics to extract potential anomalous components independent of environmental noise include: Independent component analysis or blind source separation techniques are used to separate physical event packets into multiple source signal components; The separated source signal components are compared with the expected environmental noise characteristics to identify and extract components that deviate from the environmental noise as potential anomalous components.

[0045] Independent Component Analysis (ICA) or Blind Source Separation (BSS) are advanced signal processing methods whose core objective is to separate statistically independent source signals from mixed signals. In the context of big data analysis in high-end equipment manufacturing, these techniques can effectively decouple various mixed signals (e.g., normal equipment operation signals, environmental noise signals, and potential anomalous signals) from sensor data streams. Specifically, ICA seeks a linear transformation by maximizing the statistical independence of the output signals, thereby converting the observed mixed signals into mutually independent components. Blind Source Separation, on the other hand, recovers the original source signals from the observed mixed signals without prior knowledge of the source signal characteristics and mixing process. These techniques are particularly suitable for processing multi-sensor data, capable of identifying and extracting weak anomalous signals hidden within complex background noise.

[0046] Furthermore, separating the physical event packet into multiple source signal components refers to processing the aggregated physical event packet using the aforementioned Independent Component Analysis (ICA) or Blind Source Separation (BSS) techniques. The physical event packet contains raw data from different sensors that may be interrelated; this data represents a mixture of various physical phenomena, including normal equipment operation, environmental interference, and potential anomalies. By applying ICA or BSS, these mixed signals are decomposed into a series of statistically independent components, each representing a potential physical source or process.

[0047] Furthermore, comparing the separated source signal components with the expected environmental noise characteristics involves evaluating these components after obtaining multiple source signal components to distinguish which components are environmental noise and which are signals related to equipment operation or potential anomalies. The expected environmental noise characteristics are determined based on real-time operating parameters and provide the patterns or spectral characteristics that environmental noise may exhibit in sensor data under normal operating conditions. By comparing each separated source signal component with this expected environmental noise characteristic, the similarity or deviation of each component from known environmental noise can be quantified.

[0048] Therefore, identifying and extracting components that deviate from ambient noise as potential anomalies involves filtering out source signal components that significantly differ from the expected ambient noise characteristics based on the comparison results described above. These components that significantly deviate from ambient noise are more likely to represent abnormal states or precursors to faults within the equipment, as they cannot be explained by simple environmental interference. In this way, genuine anomaly signals can be effectively separated from complex background noise, providing cleaner and more indicative data for subsequent fault diagnosis.

[0049] The present invention employs independent component analysis or blind source separation techniques to effectively decompose complex physical event packets into multiple independent source signal components. These source signal components may include normal equipment operation signals, environmental noise, and potential abnormal signals. By accurately comparing these separated source signal components with pre-determined expected environmental noise characteristics, components that do not conform to the environmental noise pattern can be identified. This comparison mechanism enables the system to accurately filter out interference caused by environmental factors, thereby ensuring that the extracted potential abnormal components are truly independent of environmental noise, thus avoiding the misjudgment of environmental noise as equipment malfunction.

[0050] Specifically, the steps for performing cross-correlation analysis on potential anomalous components within a preset time window include: Identify sensor data streams from multiple sensors associated with potential anomalous components; Within a preset time window that includes potential anomalous components, calculate the cross-correlation function between the sensor data streams of multiple sensors; The time delay relationship of sensor data streams from multiple sensors is determined based on the time offset corresponding to the maximum value of the cross-correlation function.

[0051] The sensor data stream that identifies multiple sensors associated with potential anomalous components refers to the system's process of identifying sensors that may have a causal or spatial proximity relationship with the anomalous event after detecting a potential anomalous component, based on preset physical association rules or sensor topology. For example, if the potential anomalous component originates from the vibration signal of a bearing, it will associate other vibration sensors, temperature sensors, or lubricating oil pressure sensors near that bearing.

[0052] Furthermore, within a preset time window that includes potential anomaly components, the cross-correlation function between the sensor data streams of multiple sensors is calculated. The preset time window is a time period whose length can be set based on factors such as the operating characteristics of the high-end equipment, the fault propagation speed, and the data sampling frequency, aiming to capture relevant data before and after an anomaly occurs. The cross-correlation function is a mathematical tool that measures the similarity of two signals under different time delays. By calculating the cross-correlation function between different sensor data streams, their interdependence and temporal lead or lag can be quantified.

[0053] Therefore, the time delay relationship of the sensor data streams from multiple sensors can be determined based on the time offset corresponding to the maximum value of the cross-correlation function. The maximum value of the cross-correlation function usually represents the strongest correlation between two signals, and the time offset corresponding to this maximum value directly reflects the time delay of one signal relative to another. For example, if the signal of sensor A has a positive time offset relative to the signal of sensor B, it indicates that the event of sensor A occurred later than the event of sensor B. In this way, the temporal order and delay magnitude between different sensor data streams can be accurately identified.

[0054] The present invention, through the aforementioned steps, enables in-depth analysis of potential anomalous components independent of environmental noise. Specifically, by identifying the sensor data streams of multiple sensors associated with potential anomalous components, the analysis's relevance and comprehensiveness are ensured. Subsequently, calculating the cross-correlation function between these sensor data streams within a preset time window reveals the temporal mutual influence and correlation between different sensor signals. The maximum value of the cross-correlation function and its corresponding time offset directly provide information on the time delay of anomalous signals propagating between different physical locations or different types of sensors. This time delay relationship is crucial for understanding the propagation path of faults, locating fault sources, and determining fault types. For example, by analyzing the time delay between vibration and temperature signals, it can be inferred whether mechanical vibration first causes a temperature rise or whether temperature anomalies first affect mechanical components.

[0055] In some embodiments of the present invention described above, after performing cross-correlation analysis on potential anomalous components within a preset time window, the comprehensive fault confidence is calculated according to preset physical association rules. However, in practical applications, if the calculation is based solely on preset physical association rules, the time delay relationship revealed by the cross-correlation analysis may not be fully utilized, resulting in insufficient accuracy in the fault confidence calculation or failure to effectively eliminate false alarms. If the above problems are not addressed, the accuracy and reliability of fault warnings may be affected. Therefore, the present invention further proposes that when calculating the comprehensive fault confidence, the time delay relationship should first be determined to conform to the preset physical association rules, thereby improving the accuracy and reliability of the fault confidence calculation.

[0056] To address this, the present invention further proposes that the steps for calculating the comprehensive fault confidence based on preset physical association rules include: Determine whether the time delay relationship conforms to the preset physical association rules; If the time delay relationship conforms to the physical association rule, the comprehensive fault confidence of the potential anomalous components is calculated according to the preset confidence calculation mechanism.

[0057] Specifically, determining whether the time delay relationship conforms to the preset physical association rules involves comparing the time delay relationship of sensor data streams from multiple sensors, obtained through cross-correlation analysis, with pre-defined rules describing the physical connections and interactions between internal components of high-end equipment. For example, if a component's failure typically causes an anomaly in its upstream sensor signal before that of its downstream sensor signal, and has a specific time delay, then this delay relationship should conform to the corresponding physical association rules. The purpose is to ensure that the identified anomaly patterns are physically plausible, thereby avoiding false alarms generated based on signal patterns that do not conform to physical laws.

[0058] If the time delay relationship conforms to the physical association rule, the comprehensive fault confidence of the potential anomalous components is calculated according to a preset confidence calculation mechanism. The preset confidence calculation mechanism can be understood as an algorithm or model that quantifies the correlation strength between potential anomalous components and specific fault modes. For example, methods such as Bayesian networks, fuzzy logic, expert systems, or machine learning models can be used to comprehensively consider factors such as the amplitude, duration, frequency characteristics of the potential anomalous components, and their conformity with the physical association rule to calculate a numerical value representing the probability of a fault. Its purpose is to provide a quantitative and reliable basis for subsequent fault early warning.

[0059] This invention effectively solves the problems of inaccurate fault confidence calculation or false alarms that may exist in traditional methods by introducing a judgment on the conformity of time delay relationships with preset physical association rules before calculating the overall fault confidence score. Specifically, cross-correlation analysis can reveal the time delay relationships between different sensor signals, and these delay relationships often contain key information about fault propagation paths and mechanisms. By comparing these time delay relationships with preset physical association rules, physically unreasonable abnormal patterns can be filtered out, ensuring that only anomalies that conform to actual physical laws are further evaluated. Therefore, when the time delay relationship conforms to the physical association rules, the overall fault confidence score is calculated based on the preset confidence score calculation mechanism, ensuring that the calculated confidence score is more physically meaningful and reliable, thereby significantly improving the accuracy of fault early warning.

[0060] Traditional big data analysis methods for high-end equipment manufacturing may only provide a general warning signal, such as "an anomaly detected" or "potential fault present," when detecting potential faults and generating early warnings. While this general warning information can alert operators to the equipment status, it often lacks detailed information such as the specific fault type, fault location, or recommended maintenance measures. If these issues are not addressed, operators still need to invest significant time and effort in further diagnosis and troubleshooting after receiving the warning, which not only reduces the efficiency of fault response but may also delay optimal maintenance, thereby increasing equipment downtime and maintenance costs. To address this, this invention proposes a more refined fault warning generation mechanism that generates fault warnings with higher guidance value by performing fault pattern matching on comprehensive fault confidence.

[0061] The steps for generating a fault warning when the overall fault confidence exceeds a preset threshold include: When the overall fault confidence exceeds a preset threshold, the corresponding fault mode is matched according to the overall fault confidence. Based on the fault mode, generate corresponding fault warnings.

[0062] Specifically, when the calculated overall fault confidence score exceeds a preset threshold, it indicates a potential fault in the high-end equipment. At this point, the system will no longer simply generate a generalized warning, but will further match the corresponding fault mode based on the overall fault confidence score. "Matching the corresponding fault mode" can be understood as the system associating the current abnormal characteristics (reflected by the overall fault confidence score) with known fault types with specific physical meanings by querying a preset fault knowledge base or rule set. For example, this matching process can be based on machine learning models (such as classifiers) to analyze the overall fault confidence score and related potential abnormal components, or it can map the confidence score to specific fault modes through predefined threshold ranges and rules. The purpose is to transform abstract abnormal signals into understandable and concrete fault descriptions.

[0063] Furthermore, once a specific fault mode is identified, the system will generate a corresponding fault warning based on that mode. "Generating a corresponding fault warning" means that the warning information will no longer be a simple "abnormality," but rather an alert containing specific details such as the clear fault type, potentially affected components, and suggested inspection directions. For example, if the matched fault mode is "bearing wear," the generated warning will clearly state "high risk of spindle bearing wear," and may include suggested inspection steps or maintenance recommendations. The purpose is to provide operators with timely, accurate, and guiding information so that they can quickly take targeted maintenance measures.

[0064] The present invention effectively solves the problem of overly generalized fault warning information in traditional solutions by introducing a fault mode matching step after the comprehensive fault confidence exceeds a preset threshold. Specifically, when the comprehensive fault confidence reaches the warning level, the system no longer directly issues a general warning. Instead, it uses this confidence as input, combined with preset fault knowledge or models, to identify the fault mode that best matches the current abnormal state. This process enables the system to not only discover anomalies from massive sensor data streams but also understand their essence, classifying them into specific physical fault types. Subsequently, based on the identified fault modes, the system can generate highly customized fault warnings, transforming abstract confidence into concrete and actionable fault descriptions. It is precisely this logical progression from "discovering anomalies" to "understanding anomalies" and then to "specific warnings" that enables the present invention to provide more insightful and practical warning information.

[0065] In some embodiments of the present invention described above, when the overall fault confidence exceeds a preset threshold, a corresponding fault warning is generated based on the matched fault mode. However, simply generating a general fault warning may not be sufficient to provide sufficiently detailed and actionable information, causing maintenance personnel to still need to spend time on further diagnosis and preparation when actually handling the fault, thereby affecting the maintenance efficiency and downtime of high-end equipment.

[0066] In response, the present invention further proposes the following steps for generating corresponding fault warnings based on the identified fault modes: querying a preset set of fault modes and maintenance experience associations, and generating fault warnings that include affected components, recommended inspection steps, necessary tool and spare parts information, and maintenance personnel skill suggestions.

[0067] Specifically, the pre-defined set of fault modes and maintenance experience associations can be understood as a structured database or knowledge base, storing the mapping relationships between various known fault modes and their corresponding detailed maintenance information. This set is typically built based on historical fault data, expert experience, maintenance manuals provided by equipment manufacturers, and actual maintenance records. Its purpose is to transform abstract fault modes into specific, executable maintenance instructions. Affected components refer to the mechanical, electrical, or control components in high-end equipment that are directly or indirectly affected under a specific fault mode. For example, when the fault mode "abnormal spindle vibration" is detected, affected components may include the spindle bearing, spindle motor, or tool clamping mechanism. Recommended inspection steps refer to the diagnostic and inspection procedures that the system suggests maintenance personnel should follow for the identified fault mode to quickly locate the root cause of the fault and verify the fault mode. For example, for a fault of "spindle bearing overheating," recommended inspection steps might include checking the lubricating oil level, measuring the bearing temperature, and listening for abnormal noises. Necessary tool and spare parts information refers to the specific tool and replacement parts list required to complete troubleshooting and repair. For example, replacing a bearing may require specialized disassembly tools, a new bearing, and the corresponding lubricant. Maintenance personnel skills recommendations refer to the level of professional skills or qualifications required to complete specific maintenance tasks in order to ensure the quality and safety of maintenance work. For example, some complex electrical faults may require personnel with senior electrical engineer qualifications to handle them.

[0068] The solution of this invention proactively queries a preset set of associations between fault modes and maintenance experience when generating fault warnings, thereby expanding single fault mode information into multi-dimensional and highly actionable maintenance guidance information. It is precisely because of this association query mechanism that the system can automatically match and extract affected components closely related to the identified specific fault mode, recommended inspection steps, necessary tool and spare parts information, and maintenance personnel skill suggestions. Thus, the generated fault warning is no longer just a simple alarm, but a comprehensive and instructive maintenance task package.

[0069] Secondly, see Figure 2 The present invention also discloses a big data analysis system for high-end equipment manufacturing, the system comprising: The data receiving and processing module 210 is used to receive sensor data streams and real-time operating parameters from different sensors of high-end equipment, and to aggregate the sensor data streams from different sensors into physical event packets. The feature determination module 220 is used to determine the expected environmental noise features based on real-time operating parameters; The anomaly extraction module 230 is used to perform signal separation processing on the physical event packet based on the expected environmental noise characteristics, and extract potential abnormal components independent of environmental noise. The fault confidence calculation module 240 is used to perform cross-correlation analysis on potential abnormal components within a preset time window, and calculate the comprehensive fault confidence based on preset physical association rules. The early warning module 250 is used to generate a fault warning when the overall fault confidence exceeds a preset threshold.

[0070] This invention proposes a high-end equipment industrial big data analysis system, aiming to solve problems such as sensor data time misalignment, environmental noise interference, and inaccurate fault identification during the long-term operation of traditional high-end equipment. Through modular design, the system achieves an automated and intelligent process from data acquisition, preprocessing, anomaly detection, and fault early warning. Specifically, the data receiving and processing module 210 efficiently integrates multi-source heterogeneous sensor data to form a unified physical event package; the feature determination module 220 dynamically identifies environmental noise characteristics under current operating conditions, laying the foundation for accurate anomaly detection; the anomaly extraction module 230 uses advanced signal processing technology to extract the true potential anomaly components from complex signals; the fault confidence calculation module 240 combines cross-correlation analysis and physical correlation rules to conduct reliability assessments of the anomaly components; finally, the early warning module 250 generates timely and detailed fault warnings based on the assessment results. Through the collaborative work of these modules, this system can significantly improve the accuracy and timeliness of early fault identification in high-end equipment, thereby effectively supporting predictive maintenance and reducing unplanned downtime and maintenance costs.

[0071] The core innovation of this invention lies in its modular and adaptive data processing and analysis architecture. First, the data receiving and processing module 210 receives sensor data streams and real-time operating parameters from different sensors in high-end equipment, and aggregates the sensor data streams into physical event packets. This module performs preliminary integration and eventification of heterogeneous data at the data source level, laying the foundation for subsequent refined analysis. Compared with the simple timestamp alignment or fixed correction models in existing technologies, the aggregation mechanism of this system is more flexible and adaptable.

[0072] Secondly, this invention utilizes a feature determination module 220 to determine the expected environmental noise characteristics based on real-time operating parameters. This innovation enables the system to dynamically adapt to different operating conditions and environmental changes. Traditional systems typically employ static noise models or thresholds, which are insufficient to handle complex and variable environmental noise. In contrast, this system, by querying a pre-defined mapping library of operating parameters and noise features or using a machine learning model for prediction, can obtain a noise baseline matching the current operating conditions in real time, thereby more accurately distinguishing between normal fluctuations and abnormal signals during signal separation.

[0073] Furthermore, this invention utilizes the anomaly extraction module 230 to perform signal separation processing on physical event packets based on expected environmental noise characteristics, extracting potential anomalous components independent of environmental noise. This represents a significant breakthrough in signal processing for this system. By employing independent component analysis or blind source separation techniques, this system can effectively separate source signals mixed within physical event packets and, combined with expected environmental noise characteristics, accurately identify those anomalous signals that truly indicate faults. This is significantly superior to anomaly detection based on fixed thresholds or simple filtering in traditional systems, which are often susceptible to interference from environmental noise, resulting in missed or false alarms.

[0074] Finally, this invention utilizes the fault confidence calculation module 240 to perform cross-correlation analysis on potential abnormal components within a preset time window, and calculates the comprehensive fault confidence based on preset physical correlation rules. This represents an innovation in the fault diagnosis decision-making level of this system. By analyzing the time delay relationship between different sensor data streams and combining it with the physical correlation rules of high-end equipment, this system can verify the physical rationality of abnormal signals from multiple dimensions, thereby calculating a more convincing comprehensive fault confidence. The introduction of this multi-source data fusion and physical constraints greatly improves the reliability and accuracy of fault judgment. When the comprehensive fault confidence exceeds a preset threshold, the early warning module 250 can generate a detailed fault warning, including affected components, recommended inspection steps, etc., providing timely and specific guidance for maintenance personnel.

[0075] In summary, this invention effectively overcomes the challenges of time misalignment, environmental noise interference, and inaccurate fault identification faced by existing technologies in big data analysis of high-end equipment industries by introducing adaptive data aggregation, dynamic noise feature recognition, advanced signal separation, and a fault confidence calculation mechanism that combines multi-source data cross-correlation analysis with physical association rules. It significantly improves the accuracy, timeliness, and guidance of fault early warning, providing strong technical support for predictive maintenance of high-end equipment.

[0076] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A big data analysis method for high-end equipment manufacturing, characterized in that, include: Receive sensor data streams and real-time operating parameters from different sensors of high-end equipment, and aggregate the sensor data streams from the different sensors into physical event packets; Based on the real-time operating parameters, the expected environmental noise characteristics are determined; Based on the expected environmental noise characteristics, the physical event packet is subjected to signal separation processing to extract potential abnormal components independent of environmental noise; Cross-correlation analysis is performed on the potential anomalous components within a preset time window, and the comprehensive fault confidence is calculated according to preset physical correlation rules. When the overall fault confidence exceeds a preset threshold, a fault warning is generated.

2. The method for big data analysis in high-end equipment manufacturing according to claim 1, characterized in that, The steps of receiving sensor data streams and real-time operating parameters from different sensors of high-end equipment include: The sensor data streams from the different sensors are received by an edge processing unit deployed next to the high-end equipment, wherein the sensor data streams from the different sensors include vibration signals, current signals, temperature signals and pressure signals from the high-end equipment. The real-time operating parameters are received by the equipment control unit of the high-end equipment, including the machining step type, spindle speed, feed rate, depth of cut, and type of machining material.

3. The method for big data analysis in high-end equipment manufacturing according to claim 1, characterized in that, The step of aggregating the sensor data streams from the different sensors into a physical event packet includes: The sensor data streams from the different sensors are adaptively clock-biased by a synchronization network, and the corrected sensor data streams are aggregated into physical event packets according to preset physical association rules.

4. The method for big data analysis in high-end equipment manufacturing according to claim 1, characterized in that, The step of determining the expected environmental noise characteristics based on the real-time operating parameters includes: Based on the real-time operating parameters, a preset mapping relationship library between operating parameters and noise features is queried to determine the environmental noise baseline features of the sensor data streams of different sensors corresponding to the real-time operating parameters, which are then used as the expected environmental noise features.

5. The method for big data analysis in high-end equipment manufacturing according to claim 1, characterized in that, The step of performing signal separation processing on the physical event packet based on the expected environmental noise characteristics to extract potential anomalous components independent of environmental noise includes: The physical event packet is separated into multiple source signal components using independent component analysis or blind source separation techniques. The separated source signal components are compared with the expected environmental noise characteristics to identify and extract components that deviate from the environmental noise as potential anomalous components.

6. The method for big data analysis in high-end equipment manufacturing according to claim 1, characterized in that, The step of performing cross-correlation analysis on the potential anomalous components within a preset time window includes: Determine the sensor data streams of multiple sensors associated with the potential anomalous components; Within a preset time window that includes the potential anomalous components, calculate the cross-correlation function between the sensor data streams of the multiple sensors; The time delay relationship of the sensor data streams of the multiple sensors is determined based on the time offset corresponding to the maximum value of the cross-correlation function.

7. The method for big data analysis in high-end equipment manufacturing according to claim 6, characterized in that, The step of calculating the comprehensive fault confidence based on preset physical association rules includes: Determine whether the time delay relationship conforms to the physical association rules in the preset physical association rules; If the time delay relationship conforms to the physical association rule, the comprehensive fault confidence of the potential abnormal component is calculated according to the preset confidence calculation mechanism.

8. The method for big data analysis in high-end equipment manufacturing according to claim 1, characterized in that, The step of generating a fault warning when the overall fault confidence exceeds a preset threshold includes: When the overall fault confidence exceeds a preset threshold, the corresponding fault mode is matched according to the overall fault confidence. Based on the described fault mode, a corresponding fault warning is generated.

9. The method for big data analysis in high-end equipment manufacturing according to claim 8, characterized in that, The step of generating a corresponding fault warning based on the identified fault mode includes: The system queries a preset set of associations between fault modes and maintenance experience to generate a fault warning that includes affected components, recommended inspection steps, necessary tool and spare parts information, and maintenance personnel skill suggestions.

10. A high-end equipment manufacturing big data analysis system, characterized in that, The system includes: The data receiving and processing module is used to receive sensor data streams and real-time operating parameters from different sensors of high-end equipment, and to aggregate the sensor data streams from the different sensors into physical event packets. The feature determination module is used to determine the expected environmental noise features based on the real-time operating parameters; An anomaly extraction module is used to perform signal separation processing on the physical event packet based on the expected environmental noise characteristics, and extract potential anomaly components independent of environmental noise; The fault confidence calculation module is used to perform cross-correlation analysis on the potential abnormal components within a preset time window, and calculate the comprehensive fault confidence based on preset physical correlation rules. The early warning module is used to generate a fault warning when the overall fault confidence exceeds a preset threshold.

Images