Industrial equipment fault diagnosis method and system based on multi-parameter statistical significance

Abstract

The invention provides an industrial equipment fault diagnosis method and system based on multi-parameter statistical significance, and relates to the technical field of industrial internet and predictive maintenance, and the method comprises the steps: obtaining sensor time sequence data of industrial equipment; based on the time sequence data of the sensor, calculating to obtain a plurality of fault measurement indexes reflecting the health state of the equipment, and forming a time sequence of each index; aiming at the time sequence of each fault measurement index, executing linear regression analysis to obtain a slope representing a trend direction, a value representing trend statistical significance and a correlation coefficient representing trend stability; performing joint judgment on the slope, the value and the correlation coefficient with a preset condition to identify a deterioration trend with statistical significance; and when the deterioration trend is identified, generating and outputting fault early warning information. According to the method, through combined judgment of slope-P value-correlation coefficient, the accuracy of fault trend diagnosis is improved, and false alarms caused by random fluctuation are reduced.

Description

Technical Field

[0001] This invention relates to the fields of industrial internet and predictive maintenance technology, and in particular to a method and system for diagnosing industrial equipment faults based on multi-parameter statistical significance. Background Technology

[0002] In industrial production environments, large rotating machinery, such as main drive systems, fans, and pump sets, are core assets ensuring production continuity, especially large industrial equipment operating under unpredictable conditions. Effective condition monitoring and predictive maintenance of this equipment are crucial for avoiding unplanned downtime, reducing maintenance costs, and ensuring production safety. However, existing fault diagnosis technologies face significant bottlenecks in practical applications.

[0003] Currently, the most common fault alarm method in the industrial field is based on setting fixed alarm thresholds for the absolute values ​​of key operating parameters (such as vibration and temperature). When the monitored value exceeds the preset upper or lower limit, the system triggers an alarm. While this method is simple and intuitive, it reveals its limitations under certain operating conditions (such as non-steady-state conditions).

[0004] First, this method is prone to generating a large number of false alarms. The operating conditions of industrial equipment (such as load, speed, power, etc.) are dynamically changing, and fluctuations in these normal operating conditions (i.e., non-steady-state conditions) will cause drift in the monitoring signal reference value. For example, when the equipment load increases (condition 1), its vibration amplitude will naturally increase accordingly; when the equipment load decreases (condition 2), its vibration amplitude will naturally decrease accordingly. If a fixed absolute threshold is set too small, the drift in the monitoring signal reference value caused by these fluctuations in normal operating conditions (i.e., non-steady-state conditions) will be easily misjudged as equipment failure, resulting in a large number of false alarms. Therefore, traditional absolute thresholds are prone to false alarms. Such frequent false alarms not only interfere with the normal work of maintenance personnel, but more seriously, they gradually erode their trust in the alarm system, leading to "alarm fatigue."

[0005] Secondly, this method leads to a significant lag in early warning. For example, to avoid frequent false alarms, the threshold may be set too high. In this case, even if the equipment malfunctions, the absolute value threshold, being set too high, will not trigger an alarm, resulting in a severe lag in early warning. Absolute value alarms can only be triggered when the fault has progressed to a considerably serious stage and the signal characteristics have deteriorated significantly. This forces maintenance work into a passive "retroactive repair" mode, failing to fully realize the core value of predictive maintenance: "prevention is better than cure." Over time, when the equipment is in an early or initial stage of failure, the high absolute value threshold may prevent the alarm from being triggered, failing to provide maintenance personnel with early warning information. Therefore, such faults may be unconsciously ignored, leading to missed opportunities for optimal handling, which will inevitably result in more serious equipment damage or even catastrophic accidents.

[0006] To address the drawbacks of static thresholds, the industry has recognized the importance of analyzing data trends. However, existing trend analysis methods are often overly simplistic and lack scientific verification mechanisms. A common practice is to determine whether equipment condition is deteriorating simply by judging whether the slope of the trend line showing the monitoring indicator changes over time is positive. This "method relying solely on slope judgment" also suffers from insufficient reliability.

[0007] Real-world industrial data is filled with random noise and transient, irregular fluctuations, especially under non-steady-state conditions. These factors may create a seemingly upward trend in the short term, but this trend does not indicate a real, sustained deterioration in the equipment's health. Relying solely on the slope of the trend can easily lead to misinterpreting these "pseudo-trends" caused by random factors as signs of failure, resulting in unreliable diagnostic conclusions.

[0008] Therefore, there is an urgent need for a more intelligent and robust diagnostic method that can effectively distinguish between real fault development trends and random data fluctuations, so as to achieve early and accurate warnings. Summary of the Invention

[0009] One objective of this invention is to provide a fault diagnosis method and system for industrial equipment based on multi-parameter statistical significance, in order to solve the problems pointed out in the background art. The purpose of this invention is to overcome the limitations of static threshold alarms under dynamic operating conditions in existing technologies (which are prone to generating a large number of false alarms or severely delayed early warnings) and the shortcomings of traditional trend analysis methods (which are too simplistic, lack scientific verification mechanisms, and easily misjudge random fluctuations as fault symptoms), and to provide a fault diagnosis method and system that can accurately and automatically determine the deterioration trend of the health status of industrial equipment at an early stage.

[0010] The core technical idea of ​​this invention lies in proposing and implementing a set of trend diagnosis rules based on multi-parameter joint judgment. This rule no longer relies solely on the direction of the trend (i.e., the slope), but innovatively introduces two key statistical indicators: P-value and correlation coefficient (CorrelationCoefficient) This method is used to rigorously quantify and verify the effectiveness of data deterioration trends. A potential fault trend must satisfy either directionality (slope greater than zero) and statistical significance (p-value sufficiently small) or directionality (slope greater than zero), statistical significance (p-value relatively small), and stability (correlation coefficient sufficiently strong) to be confirmed as a real and noteworthy fault warning signal.

[0011] In a first aspect, embodiments of the present invention provide an industrial equipment fault diagnosis method based on multi-parameter statistical significance, comprising: Acquire timing data from sensors in industrial equipment; Based on the time-series data from the sensors, multiple fault measurement indicators reflecting the health status of the equipment are calculated, and time series of each indicator are formed. For each fault metric's time series, perform linear regression analysis to obtain the slope representing the trend direction and the statistical significance of the trend. Values, and correlation coefficients that characterize trend stability; The slope, The values ​​and correlation coefficients are combined with preset conditions to identify statistically significant deterioration trends; When the deterioration trend is identified, a fault warning message is generated and output.

[0012] Optionally, the joint judgment includes: determining whether the following conditions are met simultaneously: The slope is greater than zero; The The value is less than the preset significance level threshold; The absolute value of the correlation coefficient is greater than the preset correlation strength threshold.

[0013] Optionally, the significance level threshold is 0.05, and the correlation strength threshold is 0.5.

[0014] Optionally, the fault measurement indicators include root mean square value, high-frequency bearing fault indicators, kurtosis indicators, gear sideband fault characteristic indicators, or dynamic balance fault indicators.

[0015] Optionally, the fault measurement index includes kurtosis, which is sensitive to impact components. Kurtosis is a dimensionless index for measuring rolling bearing faults and gear faults. The high-frequency bearing fault index is used to calculate the defect coefficient of the rolling bearing based on the high-frequency signal. The calculation process includes: performing high-pass filtering on the signal to remove low-frequency signals and retain the resonant frequency components of the rolling bearing; multiplying the peak factor and the effective value by preset coefficients respectively and then adding them together to obtain the dimensionless index value.

[0016] Optionally, the linear regression analysis uses the least squares method for parameter estimation.

[0017] Optionally, the sensor time-series data is vibration signal data from large rotating machinery; when generating fault warning information, a visual diagnostic report containing trend graphs is generated simultaneously.

[0018] Optionally, the significance level threshold and the correlation strength threshold are default parameters empirically determined based on historical equipment operation data from the steel, wind power, or port industries, and can be configured and adjusted in the system according to equipment type or operating conditions.

[0019] Secondly, an industrial equipment fault diagnosis system based on multi-parameter statistical significance provided by embodiments of the present invention includes: The data acquisition module is used to acquire time-series sensor data from industrial equipment. The feature engineering module is used to calculate the time series of multiple fault measurement indicators based on the sensor time series data; The trend diagnostic engine performs linear regression analysis on the time series of each fault metric to obtain the slope, The values ​​and correlation coefficients are used to make a joint judgment; An alarm module is used to generate a fault warning message when the trend diagnosis engine identifies a statistically significant deterioration trend.

[0020] Optionally, the trend diagnosis engine is configured to: confirm the existence of a statistically significant deterioration trend in response to a slope greater than zero, a p-value less than a significance level threshold, and an absolute value of the correlation coefficient greater than a correlation strength threshold; wherein the significance level threshold and the correlation strength threshold are preset as empirical verification values ​​in the system and can be modified by user configuration.

[0021] The present invention has achieved the following beneficial effects: 1. Improved diagnostic accuracy and reduced false alarm rate: Through the innovative three-dimensional joint judgment criterion of "slope-P-value-correlation coefficient", this invention can effectively filter out false trends caused by operating condition fluctuations and random noise, accurately capture real and continuous equipment health deterioration signals, thereby significantly reducing the false alarm rate and improving the reliability of the alarm system.

[0022] 2. Early warning capability: The core of this method lies in its sensitivity to "trends" rather than "absolute values". It can identify weak but significant deterioration trends in the early stages of a fault, that is, when the absolute values ​​of various indicators are far from reaching the traditional alarm thresholds, through statistical methods. This provides a time window for planned maintenance and spare parts preparation, demonstrating the value of predictive maintenance.

[0023] 3. Fully automated and objective diagnostic process: The entire diagnostic process, from data acquisition, feature calculation, trend judgment to alarm output, is driven by algorithms, eliminating the subjectivity and experience dependence that may exist in manual diagnosis, and ensuring the consistency, objectivity and repeatability of diagnostic results. It is particularly suitable for unattended monitoring of large-scale equipment clusters.

[0024] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0025] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0026] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of an industrial equipment fault diagnosis method based on multi-parameter statistical significance in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the detailed working principle of the trend diagnosis engine in an embodiment of the present invention; Figure 3 This is a schematic diagram of the software interface of the device health status monitoring platform in an embodiment of the present invention; Figure 4 This is an example diagram showing the output of a specific fault diagnosis conclusion and suggestion in an embodiment of the present invention; Figure 5 This is a schematic diagram of the system cockpit interface applied to the wind power industry in an embodiment of the present invention; Figure 6 This is a schematic diagram of the 3D visualization monitoring interface of the wind turbine drive train in an embodiment of the present invention; Figure 7 This is a schematic diagram of the real-time trend analysis interface for a specific indicator in an embodiment of the present invention; Figure 8 This is a schematic diagram of the fault warning and diagnosis conclusion list interface generated in an embodiment of the present invention. Detailed Implementation

[0027] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0028] Figure 1 This application provides a flowchart of an industrial equipment fault diagnosis method based on multi-parameter statistical significance, as shown in the following embodiment. Figure 1 As shown, the method includes: S1. Acquire sensor timing data from industrial equipment; S2. Based on the time-series data from the sensors, multiple fault measurement indicators reflecting the health status of the equipment are calculated, and time series of each indicator are formed. S3. For each fault measurement index time series, perform linear regression analysis to obtain the slope representing the trend direction and the statistical significance representing the trend. Values, and correlation coefficients that characterize trend stability; S4, the slope, The values ​​and correlation coefficients are combined with preset conditions to identify statistically significant deterioration trends; S5. When the deterioration trend is identified, generate and output fault warning information.

[0029] The working principle and beneficial effects of this embodiment are as follows: The fault diagnosis method proposed in this invention has an overall architecture and data processing flow that constitutes an automated closed loop from data acquisition to diagnostic output: 1. Data Acquisition: Raw time-domain signal data is acquired in real time or periodically through various sensors (such as vibration acceleration sensors, vibration velocity sensors, eddy current displacement sensors, etc.) deployed on the monitored industrial equipment (such as the main drive structure of the equipment).

[0030] 2. Feature Engineering: A series of signal processing and calculations are performed on the acquired raw signals to extract statistical feature indicators that can quantify the specific health status of the equipment, forming a multi-dimensional "fault measurement index matrix" that changes over time. Each row of this matrix represents a point in time, and each column represents a specific health indicator (such as root mean square value, high-frequency bearing failure, kurtosis, gear sideband fault characteristics, dynamic balance failure, skewness, impulse, misalignment, kurtosis, peak value, and margin factor, etc.).

[0031] 3. Trend Diagnosis Engine: The time series data of each indicator in the above indicator matrix is ​​input into the core module of this invention, namely the multi-parameter trend judgment module, for calculation and logical judgment. This module is the technical core of this invention. It executes a strict, three-in-one judgment logic on the time series data of each indicator, namely the "three-dimensional joint judgment criterion" described later.

[0032] 4. Alarms and Reports: Once the system confirms that one or more indicators show a clear fault trend, it will generate specific diagnostic conclusions and push them to maintenance personnel through various means such as the software platform interface, SMS, and email, along with visual evidence such as trend charts for verification.

[0033] This embodiment achieves continuous monitoring and early warning of equipment health status through an automated process. Its advantages lie in overcoming all the shortcomings of static threshold alarms and simple slope analysis in the background technology through a subsequent statistical trend diagnostic engine, achieving high accuracy and low false alarm rate in early fault diagnosis. Furthermore, the entire process requires no manual intervention, ensuring objectivity and consistency.

[0034] In one embodiment, the joint judgment includes determining whether the following conditions are met simultaneously: The slope is greater than zero; The The value is less than the preset significance level threshold; The absolute value of the correlation coefficient is greater than the preset correlation strength threshold.

[0035] The working principle and beneficial effects of this embodiment are as follows: This "three-dimensional joint judgment criterion" is the core technology of the "trend diagnosis engine" of this invention. It provides a rigorous quantitative verification for judging whether a trend is real and effective.

[0036] For any selected fault indicator (as the dependent variable) ), with time (as the independent variable) A linear regression model is established for it. The basic form of the model is: ; in, The intercept is... The slope This represents the random error term. The model parameters are estimated using the least squares method.

[0037] After obtaining the model parameters, the system does not immediately issue an alarm, but instead performs a rigorous logical check on the results of the linear regression analysis. Here, AND logic is used, meaning the system only confirms a "statistically significant deterioration trend" when the following three conditions are simultaneously met: Directional verification (slope b>0): Ensures that the index shows an upward trend over time, which conforms to the physical laws of equipment performance degradation; Significance check (P-value < preset threshold, such as 0.05): This is the core "filter" of this invention to avoid false alarms. The P-value is used to quantify the probability that the current trend is caused by random noise (luck). Only when the P-value is small enough will the system consider the trend to be statistically significant. If the slope is large but the P-value does not meet the standard (e.g., very few data points or extremely large fluctuations), the system will judge it as random fluctuation and will not confirm it as a fault trend; Stability check (correlation coefficient |r|> preset threshold, such as 0.5): ensures that the trend has a certain strength and stability, and excludes weak or extremely unstable linear relationships.

[0038] Alarm triggering logic: Once (and only if) all the joint judgments in step S4 pass, the system confirms the deterioration trend, and the alarm module (S5) is immediately triggered, automatically generating fault warning information containing trend graphs and mechanism conclusions. This mechanism ensures that the alarm is based on statistical facts rather than random fluctuations, thus answering the questions of maintenance personnel about "why it looks like it's rising but doesn't alarm" (because the P-value hasn't passed) or "why it alarms before reaching the threshold" (because the trend is significant).

[0039] In one embodiment, the joint judgment includes determining whether the following conditions are met simultaneously: The slope is greater than zero; The The value is less than the preset significance level threshold.

[0040] This embodiment is similar to the previous embodiment, but it limits the conditions to only the slope condition and the P-value condition. When the slope is greater than zero, When the value is less than the preset significance level threshold (small enough), it is also judged as "upward trend".

[0041] In one embodiment, the significance level threshold is 0.05, and the correlation strength threshold is 0.5.

[0042] The working principle and beneficial effects of this embodiment are as follows: The key judgment threshold proposed in this embodiment, namely the significance level, is and correlation coefficient It was not set arbitrarily.

[0043] These values ​​are based on a large amount of real equipment operation data accumulated over a long period of time in specific industrial fields (such as metallurgy and mining equipment). After repeated empirical studies and algorithm effect verification, a set of parameters that perform optimally in this application scenario has been summarized.

[0044] The beneficial effect of this embodiment is that the discovery and application of this set of specific parameter values ​​constitutes an important technical contribution of the present invention. It closely combines general statistical theory with specific industry knowledge and data characteristics, solving the practical problem of how to extract real fault signals from noisy industrial data, providing a specific and verified quantitative standard for the "three-dimensional joint judgment criterion", and ensuring the realization of high-precision diagnosis.

[0045] In one embodiment, the fault measurement indicators include indicators such as root mean square value, high-frequency bearing fault, kurtosis, gear sideband fault characteristics, dynamic balance fault, skewness, impulse, misalignment, kurtosis, peak value, and margin factor.

[0046] The working principle and beneficial effects of this embodiment are as follows: This embodiment details the specific metrics calculated in the "feature engineering" step. The effectiveness of this method relies on obtaining quantitative metrics that accurately reflect the health status of the equipment. The goal of this step is to calculate and construct a "fault measurement metric matrix" from raw, complex sensor signals.

[0047] These metrics are typically designed and calculated based on well-established mechanical failure mechanism models. For example, for large rotating machinery, this matrix may include, but is not limited to: Gear sideband fault characteristic indicators: used to capture sideband changes on both sides of the gearbox meshing frequency, effectively identifying faults such as broken teeth and wear.

[0048] Dynamic balance index: used to measure the degree of unevenness in rotor mass distribution.

[0049] Misalignment indicator: Used to measure the alignment of the centerlines of two axes.

[0050] Loosening index: Used to measure the loosening of structural components.

[0051] Fault characteristic indicators, such as peak factor, skewness, and specific frequency amplitude of the envelope spectrum, are sensitive to damage to the rolling elements or raceways of bearings, pitting or wear of gears, etc.

[0052] High-Frequency Defender Index (HFDI): A metric for measuring early failures in high-frequency bearings.

[0053] Kurtosis index: It is highly sensitive to the impact components in the signal and is a classic indicator for diagnosing faults such as early pitting corrosion in bearings and gearboxes.

[0054] Root mean square (RMS) index.

[0055] ClearFactor index.

[0056] Peak (PEAK) metric.

[0057] Distortion index.

[0058] Pulse index.

[0059] Table 2 below provides an example of this indicator matrix, detailing the physical meaning of some key indicators and typical associated fault types: Table 2: Example of Fault Measurement Index Matrix The beneficial effect of this embodiment is that by defining an index matrix containing multiple fault mechanism-based indicators, the present invention can comprehensively monitor the health status of equipment from multiple dimensions. Different indicators are specific to different fault types (such as imbalance, misalignment, and bearing damage). When the subsequent trend diagnosis engine detects a statistically significant upward trend in a specific indicator (such as kurtosis), it can not only provide early warning but also provide important evidence for fault location and characterization (e.g., diagnosing early bearing damage). In one embodiment, the fault measurement index includes kurtosis, which is sensitive to impact components and is a dimensionless index for measuring rolling bearing faults; the high-frequency bearing fault index (HFDI) is used to calculate the defect coefficient of the rolling bearing based on high-frequency signals. Its calculation process includes: high-pass filtering the signal to remove low-frequency signals and retain the resonant frequency components of the rolling bearing; multiplying the peak factor and the effective value by preset coefficients and then adding them together to obtain the dimensionless index value.

[0060] The working principle and beneficial effects of this embodiment are as follows: This embodiment is a further detailed explanation of specific indicators in the previous embodiment.

[0061] 1. Kurtosis: As shown in Table 2, kurtosis is a dimensionless index that measures the impact of a vibration signal. In the early stages of rolling bearing or gearbox failure (such as early pitting), the impact will cause the kurtosis value to increase significantly, while the overall impact energy of the signal (such as RMS) may not have changed much at this time.

[0062] 2. High-Frequency Bearing Failure Index (HFDI): Addressing the difficulty in effectively quantifying high-frequency resonance components in the early stages of rolling bearing failure using existing technologies, this embodiment constructs a new calculation index. The calculation process is as follows: Step A (High-pass filtering): Pass the original vibration signal through a high-pass filter to remove all low-frequency signals (such as rotational frequency and its low harmonics) and retain the high rolling bearing resonant frequency components.

[0063] Step B (Weighted Calculation): Based on the filtered high-frequency signal, calculate the peak factor (CrestFactor) and the effective value (RMS) respectively.

[0064] Step C (Indicator Generation): Multiply the peak factor and the effective value by different preset coefficients and then add them together.

[0065] Technical effect: The final HFDI is a dimensionless index value. This index can effectively determine whether the bearing is in the early, development, or expansion stage of a fault, making up for the shortcomings of traditional spectrum analysis in early fault identification.

[0066] In one embodiment, the linear regression analysis uses the least squares method for parameter estimation.

[0067] The working principle and beneficial effects of this embodiment are as follows: This embodiment specifies the specific mathematical method for performing linear regression analysis. In the "Three-Dimensional Joint Judgment Criterion," it is necessary to first establish... The model was constructed, and the key parameters, especially the slope, were solved. .

[0068] This embodiment uses Ordinary Least Squares (OLS) for parameter estimation. OLS is a mature, stable, and computationally efficient mathematical method. Its principle is to minimize the actual observed values ​​(…). ) and model predictions ( The sum of squared residuals between ().

[0069] The beneficial effect of this embodiment is that, by explicitly adopting the classic OLS algorithm, it provides a solid, reliable, and easily implemented mathematical foundation for the core diagnostic logic of this invention, ensuring the slope... P-value and correlation coefficient The accuracy and consistency of the calculation results are prerequisites for achieving automated diagnosis.

[0070] In one embodiment, the sensor time-series data is vibration signal data from large rotating machinery; when generating fault warning information, a visual diagnostic report containing trend graphs is generated simultaneously.

[0071] The working principle and beneficial effects of this embodiment are as follows: This embodiment defines the typical application scenarios and output format of the present invention.

[0072] 1. Application Scenarios (Data Source): This invention is particularly applicable to "large rotating machinery" in industries such as metallurgy and mining. In such scenarios, the most crucial monitoring data is "vibration signal data," which is typically collected by "vibration acceleration sensors." Therefore, this embodiment specifies that the input data is "vibration signal data from large rotating machinery."

[0073] 2. Output Format (Report): When the diagnostic engine identifies a fault trend, simply outputting an "alarm" signal is insufficient. To facilitate verification and decision-making by operations and maintenance personnel, the system will simultaneously generate a "visualized diagnostic report containing trend graphs" when generating early warning information (such as SMS or email).

[0074] The report will clearly show which “failure metric” (such as kurtosis) showed a trend, the time series plot of the trend, the fitted regression line, and the key statistical evidence (b>0, P<0.05, r>0.5).

[0075] The beneficial effect of this embodiment lies in its combination of abstract diagnostic methods with specific industrial practices. By generating visualized diagnostic reports, this invention not only provides "alarms" but also "evidence," greatly enhancing the credibility of diagnostic conclusions (overcoming "alarm fatigue"), helping maintenance personnel intuitively understand the process and severity of equipment condition deterioration, and providing data support for subsequent maintenance decisions.

[0076] In one embodiment, the significance level threshold and the correlation strength threshold are default parameters empirically determined based on historical equipment operation data from the steel, wind power, or port industries, and can be configured and adjusted in the system according to equipment type or operating conditions.

[0077] The working principle and beneficial effects of this embodiment are as follows: Regarding the setting of thresholds: The significance level threshold (recommended 0.05) and correlation strength threshold (recommended 0.5) used in this embodiment are not set arbitrarily.

[0078] These values ​​are derived from extensive historical equipment operation data in industries such as steel, wind power, and ports. Experiments have confirmed that this parameter combination demonstrates extremely high accuracy in trend diagnosis within these industries, effectively balancing false alarms and false negatives.

[0079] While the default values ​​described above apply to most scenarios, the system provided by this invention supports flexible parameter configuration. Users can fine-tune the thresholds in the system configuration interface according to specific equipment types (such as low-speed heavy-load equipment, high-speed light-load equipment) or special operating conditions to achieve the best monitoring results.

[0080] Figure 2 This diagram illustrates the detailed working principle of the trend diagnosis engine in this embodiment of the invention. The process clearly demonstrates the implementation path of the "three-dimensional joint judgment criterion." First, in the "monitoring phase," the system calculates various fault measurement indicators (such as HFDI, kurtosis index, etc.) based on sensor data. Then, the "time series" of these indicators are input into the diagnosis engine. The engine uses the "least squares method" to perform a "linear regression model" analysis on the time series, thereby obtaining the "slope," "P-value," and "correlation coefficient." Next, the system performs a "conditional judgment," that is, it determines whether the slope is greater than zero, the P-value is less than a threshold, and the absolute value of the correlation coefficient is greater than a threshold simultaneously. Only when all conditions are met will the system determine that the trend is "rising" and ultimately "output based on the mechanism conclusion," that is, generate fault warning information.

[0081] Figure 3 This is a schematic diagram of the software interface of the device health status monitoring platform in an embodiment of the present invention. The interface intuitively illustrates the specific implementation of the method of the present invention. The system platform provides a visual monitoring interface.

[0082] Figure 4 This is a specific example of the output of fault diagnosis conclusions and suggestions in an embodiment of the present invention. The figure is... Figure 3 The system interface shows the report output format for diagnostic results. In this example, because the system (trend diagnostic engine) identified, for example... Figure 3 The statistics showed a significant deterioration trend, and the alarm module automatically generated and output a "fault warning message." This report ( Figure 4 The "Operating Status" of the "SU06 Lifting Trolley" is marked as "Alarm," and a clear "Diagnostic Conclusion and Recommendation" is given: "The bearing damage trend of the lifting gearbox is increasing; attention is recommended, and the equipment operation status should be checked as appropriate." This embodiment demonstrates that the present invention can achieve early and accurate early warning, and provide data support and specific suggestions for the subsequent decision-making of maintenance personnel.

[0083] To further illustrate the versatility and effectiveness of this invention in different industrial scenarios Figures 5 to 8 This demonstrates another specific application embodiment of the method of the present invention in the field of wind power generation.

[0084] like Figure 5 The image shows the system's main control panel interface. This interface integrates a GIS map and macro-statistical data, displaying the real-time online status, health score (e.g., "55% health"), and alarm status distribution of all units within the wind farm. This provides maintenance personnel with a global perspective; once the underlying algorithm identifies a deteriorating trend, the status of the relevant units will be immediately indicated on this interface through color changes (e.g., turning red).

[0085] like Figure 6 The image shows the 3D visualization monitoring interface after selecting a wind turbine (e.g., "Wind Turbine #2 Drivetrain"). The system uses digital twin technology to construct an internal structural model of the nacelle and directly maps real-time monitoring values ​​(such as vibration acceleration and velocity, etc.) of key components like the generator drive end and the high-speed shaft of the gearbox to the corresponding locations in the 3D model. This helps maintenance personnel intuitively understand the physical location of the fault and verify the correspondence between the indicators extracted by the "Feature Engineering" module and the physical components in this invention.

[0086] like Figure 7The image shows the core trend analysis and real-time monitoring interface. The list at the top of the interface displays multi-dimensional fault measurement index values ​​(including kurtosis, RMS value, vibration acceleration, and RMS vibration velocity) for different measuring points (such as generators and gearboxes). The bottom of the interface displays a time series curve of the selected index (such as "generator drive end - vibration acceleration"). This is the key interface where the "trend diagnosis engine" of this invention functions: the system background performs linear regression analysis on the historical data corresponding to this curve, calculating the slope, p-value, and correlation coefficient. The system will automatically identify the deterioration characteristics of the curve, rather than simply relying on whether the current value exceeds the limit.

[0087] like Figure 8 The image shows a list of alarm events and diagnostic conclusions generated by the system. When the trend diagnosis engine confirms a statistically significant deterioration trend, the alarm module generates specific diagnostic entries. As shown in the list, the system not only provides the "alarm level" and "status," but also a detailed mechanistic description in the "diagnostic conclusion" column (e.g., "pump bearing wear, upward trend, high vibration...it is recommended to check the equipment operation status"). This detailed text output verifies that the present invention can transform statistical calculation results into actionable operation and maintenance suggestions, realizing a closed loop from data to decision-making.

[0088] This application embodiment provides a schematic diagram of an industrial equipment fault diagnosis system based on multi-parameter statistical significance. The system includes: The data acquisition module is used to acquire time-series sensor data from industrial equipment. The feature engineering module is used to calculate the time series of multiple fault measurement indicators based on the sensor time series data; The trend diagnostic engine performs linear regression analysis on the time series of each fault metric to obtain the slope, The values ​​and correlation coefficients are used to make a joint judgment; An alarm module is used to generate a fault warning message when the trend diagnosis engine identifies a statistically significant deterioration trend.

[0089] In one embodiment, the trend diagnosis engine is configured to: confirm the existence of a statistically significant deterioration trend in response to a slope greater than zero, a p-value less than a significance level threshold, and an absolute value of the correlation coefficient greater than a correlation strength threshold; wherein the significance level threshold and the correlation strength threshold are preset as empirical verification values ​​in the system and can be modified by user configuration.

[0090] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for industrial equipment fault diagnosis based on multi-parameter statistical significance, characterized in that, The method comprises: obtaining sensor time series data of an industrial equipment; based on the sensor time series data, calculating a plurality of failure measurement indicators reflecting the health status of the equipment, and forming time series of each indicator; For each failure measurement index time series, linear regression analysis is performed to obtain the slope representing the trend direction, the value representing the trend statistical significance, and the correlation coefficient representing the trend stability. values, and the correlation coefficient representing the trend stability. combining the slope, the value and the correlation coefficient with preset conditions to identify a statistically significant deterioration trend; when the deterioration trend is identified, generating and outputting a failure warning information.

2. The industrial equipment fault diagnosis method based on multi-parameter statistical saliency according to claim 1, characterized in that, The joint judgment includes: judging whether the following conditions are met at the same time: The slope is greater than zero; The a value less than a preset significance level threshold; The absolute value of the correlation coefficient is greater than a preset correlation strength threshold.

3. The industrial equipment fault diagnosis method based on multi-parameter statistical saliency according to claim 2, characterized in that, The significance level threshold is 0.05, and the correlation strength threshold is 0.

5.

4. The industrial equipment fault diagnosis method based on multi-parameter statistical saliency according to claim 1, characterized in that, The failure measurement indicators include root mean square value, high-frequency bearing failure indicator, kurtosis indicator, gear sideband failure feature indicator or dynamic balance failure indicator.

5. The industrial equipment fault diagnosis method based on multi-parameter statistical saliency according to claim 4, characterized in that, The failure measurement indicators include kurtosis sensitive to impact components, and the kurtosis is a dimensionless indicator for measuring rolling bearing failure and gear failure; the high-frequency bearing failure indicator is used to calculate the defect coefficient of the rolling bearing based on high-frequency signals, and the calculation process includes: high-pass filtering the signal to remove low-frequency signals and retain rolling bearing resonance frequency components, adding the dimensionless indicator value obtained by multiplying the peak factor and the effective value by a preset coefficient respectively.

6. The multi-parameter statistical saliency based industrial equipment fault diagnosis method of claim 1, wherein, The linear regression analysis uses least squares method for parameter estimation.

7. The multi-parameter statistical saliency based industrial equipment fault diagnosis method of claim 1, wherein, The sensor time series data is vibration signal data from a large rotating machine; when the failure warning information is generated, a visual diagnostic report containing a trend chart is also generated.

8. The industrial equipment fault diagnosis method based on multi-parameter statistical saliency according to claim 2, characterized in that, The significance level threshold and the correlation strength threshold are default parameters determined based on historical equipment operation data of steel, wind power or port industry, and can be configured and adjusted in the system according to the type of equipment or working condition.

9. An industrial equipment fault diagnosis system based on multi-parameter statistical significance, characterized in that, The method comprises: a data acquisition module for obtaining sensor time series data of an industrial equipment; a feature engineering module for calculating time series of a plurality of failure measurement indicators based on the sensor time series data; The trend diagnostic engine performs linear regression analysis on the time series of each fault metric to obtain the slope, The values ​​and correlation coefficients are used to make a joint judgment; an alarm module for generating a failure warning information when the trend diagnosis engine identifies a statistically significant deterioration trend.

10. The multi-parameter statistical saliency based industrial equipment fault diagnosis system as claimed in claim 9, wherein, The trend diagnosis engine is configured to confirm the existence of a statistically significant deterioration trend in response to the slope being greater than zero, the P value being less than the significance level threshold, and the absolute value of the correlation coefficient being greater than the correlation strength threshold; wherein the significance level threshold and the correlation strength threshold are preset to empirical verification values in the system and can be configured and modified by the user.

Images