A sensor-based fan operation monitoring method and system

Abstract

This application provides a sensor-based fan operation monitoring method and system, relating to the field of fan operation monitoring technology. It acquires various sensor data reflecting the fan's operating status in an industrial environment, preprocesses and extracts features from the data, identifies potential anomalies exceeding preset fluctuation ranges or instantaneous anomaly thresholds, initiates a multi-dimensional analysis process, quantitatively assesses the fault probability score, and further determines whether the potential anomaly is caused by external interference. If external interference is confirmed, misjudgment of the fan itself is avoided; if it is not external interference and the fault probability score reaches a preset confidence level, a real fault in the fan is confirmed. This effectively overcomes the problems of false alarms caused by environmental interference and missed or delayed reporting of intermittent and transient faults in traditional monitoring systems in complex industrial environments.

Description

Technical Field

[0001] This application relates to the field of fan operation monitoring technology, and more specifically, to a sensor-based fan operation monitoring method and system. Background Technology

[0002] In modern industrial production, the stable operation of key equipment directly affects production efficiency and safety. High-value equipment such as large CNC machine tools, high-power laser cutting machines, and data server clusters generate a lot of heat during long-term operation. They are usually maintained at a suitable operating temperature by a cooling system consisting of cooling fans, thereby ensuring stable operation and extending the service life of the equipment.

[0003] To effectively manage fan operation and prevent overheating shutdowns due to sudden malfunctions, engineers designed and deployed a sensor-based fan operation monitoring system. The system installs multiple sensors on each critical cooling fan, such as vibration sensors near the motor bearings and temperature and current sensors on the motor windings or power supply lines, to collect fan operation data in real time. All data is transmitted to a central processing unit via industrial Ethernet or wireless communication modules, where an analysis program compares the real-time data with preset normal operating parameter ranges. When the vibration amplitude continuously exceeds a safety threshold, or the motor temperature rises abnormally, the system quickly identifies potential problems such as bearing wear, poor dynamic balance, or winding overheating, and generates detailed fault warning information. This information is then sent to maintenance personnel via audible and visual alarms, SMS, or email, enabling them to promptly perform targeted inspections, maintenance, or replacement based on the fan number and fault type. This minimizes the risk of unexpected equipment downtime and ensures the continuous and stable operation of the production line.

[0004] However, actual industrial environments are complex and contain numerous interfering factors. The starting of large motors and the operation of frequency converters generate strong electromagnetic interference; welding and heavy machinery operations cause severe localized vibrations; and dust and oil mist from the production process easily adhere to the surfaces of equipment and sensors. These factors may cause vibration sensors to falsely report abnormally high values, or temperature sensors to read inflated readings due to a sudden rise in ambient temperature, thus triggering erroneous warnings not caused by actual fan malfunctions.

[0005] Even more challenging is that some fan failures are intermittent or transient. For example, minor ball defects in bearings may only produce brief high-frequency vibrations at specific speeds or loads, while tiny cracks in the motor winding insulation may only occasionally cause leakage current after prolonged operation and increased temperature. Traditional monitoring systems based on fixed thresholds struggle to accurately detect these anomalies, often failing to trigger alarms due to untimely sampling or short duration of the anomaly. Even if an alarm is triggered, the anomaly may have disappeared by the time maintenance personnel arrive on site, creating a false impression of "fault disappearance" and making maintenance impossible.

[0006] False alarms caused by environmental interference, as well as missed or delayed reporting of intermittent and transient faults, severely undermine the reliability and maintenance efficiency of the monitoring system. Maintenance personnel need to spend a lot of time verifying false alarms or repeatedly checking intermittent abnormal fans, which increases maintenance costs, prolongs the fault diagnosis cycle, and may even cause real potential faults to be overlooked, ultimately leading to unexpected equipment shutdown. Summary of the Invention

[0007] This application provides a sensor-based fan operation monitoring method and system, which aims to solve the problems of false alarms caused by environmental interference and missed or delayed reporting of intermittent and transient faults in existing industrial fan operation monitoring systems in complex industrial environments, thereby improving the reliability and maintenance efficiency of the monitoring system.

[0008] On the one hand, this application provides a sensor-based fan operation monitoring method, including: Acquire multiple types of sensor data reflecting the operating status of a fan in an industrial environment, including fan bearing vibration data, motor temperature data, and motor current data; The sensor data is preprocessed and features are extracted to identify potential anomalies that exceed a preset fluctuation range or instantaneous anomaly threshold. These potential anomalies include continuous abnormal signals and early weak fault signals. A multi-dimensional analysis process is initiated for the potential anomalies, and a fault probability score is obtained through quantitative evaluation. Determine whether the potential anomaly is caused by external interference. If it is caused by external interference, confirm that there is no real fault in the fan body. If it is not caused by external interference and the fault probability score reaches the preset confidence level, confirm that there is a real fault in the fan body.

[0009] Optionally, the step of determining whether a potential anomaly is caused by external interference includes: Acquire environmental reference data collected by environmental vibration sensors, environmental electromagnetic field sensors, and environmental temperature sensors; If, within a preset time difference, the data from the environmental vibration sensor shows an impact vibration exceeding a preset first threshold, the data from the environmental electromagnetic field sensor shows an electromagnetic pulse exceeding a preset second threshold, or the data from the environmental temperature sensor shows an instantaneous temperature surge exceeding a preset third threshold, then the potential anomaly is determined to be caused by external interference; otherwise, the potential anomaly is determined to be caused by non-external interference.

[0010] Optionally, the step of initiating a multi-dimensional analysis process for the potential anomalies and quantitatively evaluating the failure probability score includes: Check whether there are coordinated changes in different types of sensors on the fan within the same time window, and generate cross-validation scores based on the degree of coordinated change; Analyze the frequency of occurrence of potential outliers at different time scales and their degree of deviation from short-term trend data to generate instantaneous repetition scores and short-term trend deviation scores. Based on the cumulative operating time and maintenance records of the fan, and according to the degree of closeness to the preset lifespan and maintenance cycle, a lifespan closeness score and a maintenance cycle exceeding score are generated. The failure probability score is obtained by weighting and summing the cross-validation score, the instantaneous repetition score, the short-term trend deviation score, the near-life score, and the maintenance cycle exceedance score based on preset weighting coefficients.

[0011] Alternatively, the failure probability score can be obtained using the following formula: Failure probability score = W1 * Cross-validation score + W2 * Instantaneous repetition score + W3 * Short-term trend deviation score + W4 * Lifespan approaching score + W5 * Maintenance cycle exceeding score Among them, W1, W2, W3, W4, and W5 are preset weight coefficients.

[0012] Optionally, the step of preprocessing and feature extraction of the sensor data to identify potential anomalies exceeding a preset fluctuation range or instantaneous anomaly threshold includes: The current operating mode and environmental parameters of the fan are obtained. Based on the current operating mode and environmental parameters, the filtering parameters and noise reduction intensity for removing industrial environmental interference are dynamically adjusted. The sensor data is preprocessed to obtain the first sensor data after noise reduction and purification. Based on the first sensor data, short-time Fourier transform or wavelet transform are used to extract the spectral features and harmonic features of the first sensor data, with a focus on capturing early weak fault signals. The extracted features are normalized, the feature distribution characteristics are evaluated, and the normalization parameters are dynamically adjusted. The normalized features are compared with the baseline features of the fan under the current operating mode and environmental parameters to identify potential anomalies that exceed the preset fluctuation range or instantaneous anomaly threshold.

[0013] Optionally, the steps of normalizing the extracted features, evaluating the feature distribution characteristics, and dynamically adjusting the normalization parameters include: Based on the current operating mode of the fan and environmental parameters, select the normalized parameters; The extracted features are normalized using the normalization parameters to obtain normalized features. Evaluate the distribution characteristics of the normalized features, and adjust the normalization parameters based on the evaluation results.

[0014] Optionally, after the step of comparing the normalized features with the baseline features of the fan under the current operating mode and environmental parameters to identify potential anomalies exceeding a preset fluctuation range or instantaneous anomaly threshold, the method further includes: Based on the current operating mode of the fan and environmental parameters, a set of benchmark features matching the current operating conditions is selected from a preset benchmark feature library; Based on the benchmark feature set, calculate the deviation between the normalized features and the benchmark feature set; The system monitors the fan's running time and the update cycle of the baseline feature set. When the running time reaches a preset threshold or the update cycle reaches a preset threshold, the system triggers the baseline feature update process.

[0015] Optionally, the steps of triggering the update process of the baseline feature include: Under the current operating mode and environmental parameters of the fan, sensor data is collected and processed to extract features that reflect the operating characteristics of the fan; By comparing the characteristics reflecting the fan's operating performance with historical baseline characteristics, it can be determined whether there is a long-term slow drift. If there is a long-term slow drift, a new set of benchmark features is generated based on the extracted features and the historical benchmark features, and the benchmark feature library is updated.

[0016] Optionally, after the step of focusing on capturing early, weak fault signals, the method further includes: If a potential weak fault indication that matches the preset fault characteristic frequency is identified, an operating state incentive strategy within a safe range is generated. The operating state incentive strategy includes speed adjustment, load transient incentive, or start-stop transient incentive. During the execution of the excitation strategy, sensor data is collected at a higher sampling rate, and the signal changes before and after excitation are compared. If the signal amplitude in the target frequency band increases by more than a preset ratio or a new fault resonance peak appears, the fault probability score is improved.

[0017] On the other hand, this application provides a sensor-based fan operation monitoring system, which includes: The data acquisition module is used to acquire multiple different types of sensor data reflecting the operating status of the fan in the industrial environment. The sensor data includes fan bearing vibration data, motor temperature data, and motor current data. Anomaly identification module is used to preprocess and extract features from the sensor data to identify potential anomalies that exceed a preset fluctuation range or instantaneous anomaly threshold. The potential anomalies include continuous abnormal signals and early weak fault signals. The multi-dimensional analysis module is used to initiate a multi-dimensional analysis process for the potential anomalies and to quantitatively evaluate and obtain a failure probability score. The evaluation and decision-making module is used to determine whether potential anomalies are caused by external interference. If they are caused by external interference, it is confirmed that there is no real fault in the fan body. If they are not caused by external interference and the fault probability score reaches the preset confidence level, it is confirmed that there is a real fault in the fan body.

[0018] The sensor-based fan operation monitoring method and system disclosed in this application acquires various sensor data reflecting the fan's operating status in an industrial environment. After preprocessing and feature extraction of the data, it effectively identifies potential anomalies exceeding preset fluctuation ranges or instantaneous anomaly thresholds, including continuous abnormal signals and early weak fault signals. Based on this, the method initiates a multi-dimensional analysis process, quantitatively assesses the fault probability score, and further determines whether the potential anomaly is caused by external interference. If external interference is confirmed, misjudgment of the fan itself is avoided; if it is not external interference and the fault probability score reaches a preset confidence level, a genuine fault in the fan itself is confirmed.

[0019] The method presented in this application overcomes the problems of false alarms caused by environmental interference and missed or delayed reporting of intermittent and transient faults in traditional monitoring systems in complex industrial environments. By introducing multi-dimensional analysis and external interference judgment mechanisms, this application can more accurately distinguish between fan body faults and environmental factors, significantly reducing the false alarm rate. Simultaneously, the focused capture and multi-dimensional evaluation of early, weak fault signals enhances the ability to identify intermittent and transient faults, avoiding the false impression of "fault disappearance." Therefore, the method presented in this application can effectively improve the reliability and diagnostic accuracy of fan operation monitoring systems, reduce the ineffective workload of maintenance personnel, lower maintenance costs, and ultimately ensure the continuous and stable operation of industrial production lines, demonstrating significant practical value and technological advancement. Attached Figure Description

[0020] To illustrate this application more clearly, the accompanying drawings used in the embodiments will be briefly described below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0021] Figure 1 The diagram above illustrates a flow chart of a sensor-based fan operation monitoring method. Figure 2The diagram above illustrates a schematic of a sensor-based fan operation monitoring system.

[0022] Figure reference numerals: 100, sensor-based fan operation monitoring system; 10, data acquisition module; 20, anomaly identification module; 30, multi-dimensional analysis module; 40, evaluation and decision-making module. Detailed Implementation

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application 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 this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] 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 application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] In modern industrial production, critical equipment relies on cooling fans to maintain stable operation. However, environmental factors such as strong electromagnetic interference, severe vibration, dust, and oil mist in the field can easily cause vibration or temperature sensors to generate abnormal readings, leading to false alarms. Meanwhile, faults such as bearing defects and winding insulation cracks often exhibit intermittent and transient characteristics, making them difficult for traditional fixed threshold systems to effectively detect, easily resulting in missed or delayed reports. These problems seriously affect monitoring reliability, causing maintenance personnel to spend excessive time verifying false alarms and struggling to locate intermittent faults, increasing operation and maintenance costs, and potentially delaying the handling of real hidden dangers, leading to unexpected downtime.

[0026] like Figure 1 The diagram illustrates a schematic flowchart of a sensor-based fan operation monitoring method. This application proposes a sensor-based fan operation monitoring method, comprising: S10, acquire multiple different types of sensor data reflecting the operating status of the fan in the industrial environment, including fan bearing vibration data, motor temperature data, and motor current data; Sensor data refers to raw data reflecting the fan's operating status collected by various physical sensors. This data forms the basis for assessing the fan's health. Specifically, fan bearing vibration data is typically collected by accelerometers or vibration sensors to reflect mechanical faults such as bearing wear, imbalance, or misalignment; motor temperature data is typically collected by thermocouples or thermistors to monitor the temperature of the motor windings or housing to determine if there are problems such as overload, poor heat dissipation, or insulation aging; and motor current data is typically collected by current sensors to analyze the motor's load, power quality, and whether there are electrical faults such as short circuits in the windings. These different types of data provide comprehensive information about the fan's operating status from multiple dimensions, including mechanical, thermal, and electrical aspects.

[0027] S20, preprocess and extract features from the sensor data to identify potential anomalies that exceed a preset fluctuation range or instantaneous anomaly threshold. The potential anomalies include continuous abnormal signals and early weak fault signals. Potential anomalies refer to abnormal signals detected in sensor data that may indicate a fan malfunction. These anomalies may manifest as persistent abnormalities, i.e., stable anomalies exceeding the normal fluctuation range for an extended period, such as continuous severe vibration of the bearing or continuous high temperature of the motor; or as early, weak fault signals, i.e., transient anomalies with small amplitude but specific patterns that appear at the initial stage of a fault, such as the brief impact signal generated by a defect in the bearing balls at a specific speed, or the weak current pulse generated by partial discharge of the motor windings. Identifying these potential anomalies is a crucial first step in fault diagnosis.

[0028] S30, initiate a multi-dimensional analysis process for the potential anomalies and quantitatively evaluate and obtain a failure probability score; The multi-dimensional analysis process refers to a series of analytical steps to comprehensively evaluate identified potential anomalies. This process does not rely solely on a single data source or threshold, but rather combines multiple data types, time scales, and historical information to comprehensively consider the nature, severity, and development trend of anomalies. Through this multi-dimensional analysis, it is possible to more accurately assess whether anomalies truly indicate fan failure and quantify their probability of failure.

[0029] A failure probability score is a quantitative indicator used to represent the likelihood of a fan actually failing. This score is typically a numerical value; the higher the value, the greater the probability of a genuine fan failure. Quantitative scoring provides a more objective and refined basis for maintenance decisions, avoiding potential misjudgments that might result from simple "yes / no" responses.

[0030] S40, determine whether the potential anomaly is caused by external interference. If it is caused by external interference, confirm that there is no real fault in the fan body. If it is not caused by external interference and the fault probability score reaches the preset confidence level, confirm that there is a real fault in the fan body.

[0031] External interference refers to environmental factors that are not caused by a fault in the fan itself, but may still lead to abnormal sensor data. For example, the start-up of nearby heavy machinery may cause abnormal data from environmental vibration sensors, thus affecting the accuracy of fan bearing vibration data; electromagnetic interference may cause instantaneous fluctuations in motor current data; and a sudden increase in ambient temperature may cause an increase in motor temperature data. Distinguishing between external interference and actual faults is crucial for improving the accuracy of the monitoring system.

[0032] The core of the sensor-based fan operation monitoring method in this application lies in achieving accurate assessment of the fan's operating status through refined data processing, multi-dimensional analysis, and external interference judgment.

[0033] Several approaches can be used to acquire multiple types of sensor data reflecting the operating status of fans in an industrial environment. One approach involves acquiring fan bearing vibration data by installing a piezoelectric accelerometer near the fan bearing, acquiring motor temperature data by installing a thermistor or thermocouple inside the motor windings or on the surface of the housing, and acquiring motor current data by installing a Hall effect current sensor on the motor power supply line. These sensors are configured to continuously acquire data at a preset sampling frequency (e.g., 10 kHz for vibration data, and 100 Hz for temperature and current data) and transmit the data to a central data processing unit via wired or wireless communication modules. Another approach is to use an integrated multi-functional sensor that integrates vibration, temperature, and current measurement units, allowing for the simultaneous acquisition of multiple types of sensor data with a single device, simplifying installation and wiring.

[0034] When preprocessing and extracting features from sensor data to identify potential anomalies exceeding preset fluctuation ranges or instantaneous anomaly thresholds, the raw sensor data can first be filtered and denoised. For example, a Butterworth filter can be used to remove high-frequency noise, or wavelet denoising methods can be used to eliminate random interference. Then, key features reflecting the fan's operating status are extracted from the purified data. For example, for vibration data, time-domain features such as root mean square (RMS), peak factor, and kurtosis can be extracted, as well as frequency-domain features such as characteristic frequency amplitude and harmonic components obtained from spectral analysis. For temperature data, average temperature and temperature change rate can be extracted. For current data, current harmonic distortion rate and current fluctuation amplitude can be extracted. After feature extraction, these features are compared with baseline features under preset normal operating conditions. If a feature value consistently exceeds its normal fluctuation range (e.g., the RMS vibration value exceeds the normal upper limit for 10 consecutive seconds), or if an instantaneous but extremely high-amplitude anomaly occurs (e.g., the current suddenly spikes to 5 times the normal value within 1 millisecond), it is identified as a potential anomaly. These potential anomalies may include persistent abnormal signals, such as increased vibration due to prolonged bearing wear, or early, subtle fault signals, such as momentary impacts caused by bearing ball defects.

[0035] When initiating a multi-dimensional analysis process for potential anomalies and quantifying the failure probability score, more in-depth analysis can be conducted on the identified potential anomalies. For example, it's possible to examine whether there are coordinated changes in data from different types of sensors within the same time window. If vibration, temperature, and current data show anomalies simultaneously, it indicates a higher probability of fan failure. Furthermore, the frequency and duration of potential anomalies at different time scales, as well as their deviation from the fan's historical operating trends, can be analyzed. For instance, if a minor vibration anomaly repeatedly occurs within a short period and significantly deviates from the fan's normal operating vibration trend, its failure probability score will increase accordingly. Simultaneously, information such as the fan's cumulative operating time, maintenance records, and design life is combined to comprehensively assess the failure probability. For example, a fan nearing its design life will have a higher failure probability score than a new fan, even with minor anomalies. Finally, a weighted summation model is used to synthesize the results of each analysis to obtain a quantified failure probability score.

[0036] When determining whether a potential anomaly is caused by external interference, environmental sensor data can be used to assist in the judgment. For example, environmental vibration sensors, environmental electromagnetic field sensors, and environmental temperature sensors can be installed near the fan. When a potential anomaly is detected in the fan sensor data, the data from these environmental sensors will be checked simultaneously. If, at the same time as the potential anomaly occurs, the environmental vibration sensor detects impact vibration exceeding a preset threshold (e.g., the start-up of nearby heavy equipment), the environmental electromagnetic field sensor detects electromagnetic pulses exceeding a preset threshold (e.g., the operation of a frequency converter), or the environmental temperature sensor detects a sudden temperature rise (e.g., welding work nearby), then the potential anomaly can be determined to be caused by external interference. In this case, even if the fan sensor data is abnormal, it will be confirmed that there is no actual fault in the fan itself, thus avoiding false alarms. If the environmental sensor data does not show obvious external interference, and the fault probability score reaches a preset confidence level (e.g., a score exceeding 80 points), then a real fault in the fan itself will be confirmed, triggering the corresponding warning and maintenance procedures.

[0037] The sensor-based fan operation monitoring method of this application, through the synergistic effect of the above steps, can effectively solve the challenges faced by traditional monitoring systems in industrial environments.

[0038] Specifically, in industrial environments, the operating status of fans is affected by various factors, including equipment aging, environmental interference, and changes in operating conditions. Traditional monitoring methods often rely on single thresholds or simple rules, making it difficult to accurately distinguish between genuine faults and external interference, and also difficult to capture early, weak fault signals. This application provides a foundation for comprehensively assessing the operating status of fans by acquiring various types of sensor data, such as fan bearing vibration data, motor temperature data, and motor current data. These data reflect the health status of the fan from multiple dimensions, including mechanical, thermal, and electrical aspects.

[0039] Subsequently, these sensor data undergo preprocessing and feature extraction to extract key fault-related information from the raw data and identify potential anomalies exceeding preset fluctuation ranges or instantaneous anomaly thresholds. This step not only focuses on persistent abnormal signals but, more importantly, on capturing early, weak fault signals, which are often crucial early warning signs of faults. Refined feature extraction enhances sensitivity to potential faults.

[0040] After identifying potential anomalies, this application initiates a multi-dimensional analysis process and quantifies the failure probability score. This process comprehensively considers various factors, including the coordinated changes between different types of sensor data, the frequency and trend deviation of anomalies at different time scales, and the cumulative operating time and maintenance records of the fan. Through this multi-dimensional and comprehensive analysis, the nature and severity of potential anomalies can be assessed more comprehensively and accurately, resulting in a quantified failure probability score, providing an objective basis for subsequent decision-making.

[0041] Finally, this application introduces a crucial step in determining whether a potential anomaly is caused by external interference. By acquiring environmental reference data from environmental vibration sensors, environmental electromagnetic field sensors, and environmental temperature sensors, and comparing this data with fan sensor data, anomalies caused by fan malfunctions can be effectively distinguished from those caused by external environmental interference. If external interference is confirmed, it confirms that the fan itself does not have a real fault, thus avoiding false alarms. If the anomaly is not caused by external interference and the fault probability score reaches a preset confidence level, it confirms that the fan itself has a real fault and promptly triggers an early warning, guiding maintenance personnel to intervene.

[0042] Through the above technical solution, this application can significantly improve the accuracy and reliability of fan operation monitoring. Compared with existing technologies, the advantages of this application are: First, by fusing data from multiple types of sensors, it provides more comprehensive fan operating status information; second, through refined preprocessing and feature extraction, it can effectively identify early weak fault signals; third, through multi-dimensional analysis and quantitative evaluation, it improves the accuracy and precision of fault diagnosis; most importantly, by introducing an external interference judgment mechanism, it effectively distinguishes between real faults and environmental interference, significantly reducing the false alarm rate and avoiding the waste of maintenance resources. This comprehensive monitoring method makes the maintenance of industrial fans more efficient and precise, thereby ensuring the continuity and safety of industrial production.

[0043] In some embodiments, the step of determining whether a potential anomaly is caused by external interference includes: Acquire environmental reference data collected by environmental vibration sensors, environmental electromagnetic field sensors, and environmental temperature sensors; If, within a preset time difference, the data from the environmental vibration sensor shows an impact vibration exceeding a preset first threshold, the data from the environmental electromagnetic field sensor shows an electromagnetic pulse exceeding a preset second threshold, or the data from the environmental temperature sensor shows an instantaneous temperature surge exceeding a preset third threshold, then the potential anomaly is determined to be caused by external interference; otherwise, the potential anomaly is determined to be caused by non-external interference.

[0044] Specifically, environmental vibration sensors are used to monitor mechanical vibrations in the fan's operating environment, such as external vibration sources like nearby equipment operation, ground vibrations, or structural resonance. Environmental electromagnetic field sensors are used to detect electromagnetic interference in the environment, such as electromagnetic pulses generated by large motor startups, welding operations, or high-frequency equipment. Environmental temperature sensors are used to monitor temperature changes in the environment surrounding the fan to identify sudden temperature rises that may affect the fan's operating status, such as local heat sources or ventilation system anomalies. This environmental reference data, once collected, can be cross-validated with the fan's own sensor data. The preset time difference range refers to a specific time window before and after the point in time when a potential anomaly appears in the fan's own sensor data. Within this time window, if the environmental reference data detects impact vibrations exceeding a preset first threshold, electromagnetic pulses exceeding a preset second threshold, or sudden temperature rises exceeding a preset third threshold, it indicates that the potential anomaly is likely associated with an external environmental event. The preset first, second, and third thresholds are pre-set based on the normal fluctuation range of the industrial environment and the characteristics of potential interference sources, used to distinguish between normal environmental fluctuations and significant external interference events. For example, the impact vibration threshold can be set to an instantaneous peak value that is much higher than the background vibration, the electromagnetic pulse threshold can be set to an instantaneous high amplitude value that exceeds the normal electromagnetic noise level, and the instantaneous temperature rise threshold can be set to the magnitude of a rapid temperature rise in a short period of time.

[0045] Through the above technical solution, this application can significantly improve the accuracy and reliability of fan operation monitoring. By introducing environmental reference data and setting clear judgment criteria, the impact of external interference on fan operation status monitoring can be effectively identified and eliminated, avoiding misjudging anomalies caused by external interference as fan malfunctions, thereby reducing unnecessary downtime for inspection and maintenance costs. At the same time, it also avoids mistakenly attributing early signs of fan malfunction to external interference, thus preventing delays in maintenance and ensuring the long-term stable operation of the fan equipment and the continuity of the production process.

[0046] In some embodiments, the step of initiating a multi-dimensional analysis process for the potential anomalies and quantitatively assessing the failure probability score includes: Check whether there are coordinated changes in different types of sensors on the fan within the same time window, and generate cross-validation scores based on the degree of coordinated change; Analyze the frequency of occurrence of potential outliers at different time scales and their degree of deviation from short-term trend data to generate instantaneous repetition scores and short-term trend deviation scores. Based on the cumulative operating time and maintenance records of the fan, and according to the degree of closeness to the preset lifespan and maintenance cycle, a lifespan closeness score and a maintenance cycle exceeding score are generated. The failure probability score is obtained by weighting and summing the cross-validation score, the instantaneous repetition score, the short-term trend deviation score, the near-life score, and the maintenance cycle exceedance score based on preset weighting coefficients.

[0047] Specifically, checking for coordinated changes among different types of sensors on the fan within the same time window involves synchronously analyzing data from various sensors, such as fan bearing vibration data, motor temperature data, and motor current data. Coordinated changes are considered to exist when multiple sensors simultaneously show abnormal trends or fluctuations outside the normal range at similar time points or over a sustained period. For example, if vibration data shows an abnormal increase while motor temperature data also shows an upward trend, it indicates that these anomalies may originate from the same fault source. Based on the degree of this coordinated change, a cross-validation score can be generated. This score reflects the internal consistency and reliability of the abnormal signal; the more significant the coordinated change, the higher the cross-validation score, thus enhancing the confidence in the potential fault.

[0048] The analysis of potential anomalies across different time scales and their deviation from short-term trend data aims to assess the persistence and evolution patterns of anomalous signals. The instantaneous repetition score quantifies the frequency of repeated occurrences of potential anomalies within a short period; high-frequency repetition typically indicates some form of instability or cyclical fault. The short-term trend deviation score measures the degree of anomaly in the current state by comparing current data with short-term historical trend data of the fan under normal operating conditions; a large deviation suggests that the fan's operating condition may be deteriorating. These scores help distinguish between occasional transient disturbances and persistent or developing faults.

[0049] In practical applications, referring to the fan's cumulative operating time and maintenance records, and assessing its proximity to the preset lifespan and maintenance cycle, provides crucial background information for fault assessment. The lifespan proximity score reflects the degree of wear and aging of fan components due to long-term operation; this score increases as the cumulative operating time approaches or exceeds the design lifespan. The maintenance cycle exceeding the limit score assesses whether the fan has exceeded the recommended maintenance cycle; fans that are not maintained in a timely manner are more prone to failure. These scores provide prior knowledge for assessing the probability of failure from the perspective of the equipment's lifecycle.

[0050] Furthermore, by weighting and summing the scores of the above items based on preset weighting coefficients, the contribution of different dimensions to the probability of failure can be comprehensively considered. The preset weighting coefficients can be set based on experience, historical data analysis, or expert knowledge to reflect the importance of different factors in a specific fan type or operating environment. Through weighted summation, a quantitative failure probability score is finally obtained, which can more comprehensively and objectively reflect the probability of a real fan failure.

[0051] Through the above technical solutions, this application can significantly improve the accuracy and reliability of fan operation status monitoring. Specifically, through multi-sensor collaborative verification, the false alarm rate can be effectively reduced, ensuring that the identified abnormal signals have higher authenticity. By conducting in-depth analysis of the temporal characteristics and trend deviations of abnormal signals, potential minor faults can be detected earlier, and transient interference can be distinguished from persistent faults, thus gaining valuable time for timely intervention. In addition, by incorporating the fan's operating life and maintenance status into the evaluation system, the failure probability score not only reflects the current state but also incorporates the equipment's historical health information, providing stronger data support for predictive maintenance. Therefore, the technical solution of this application can provide a more comprehensive, accurate, and forward-looking failure probability assessment, thereby effectively guiding the maintenance and management of industrial fans, reducing the risk of unplanned downtime, and extending equipment life.

[0052] In some embodiments, the failure probability score is obtained using the following formula: Failure probability score = W1 * Cross-validation score + W2 * Instantaneous repetition score + W3 * Short-term trend deviation score + W4 * Lifespan approaching score + W5 * Maintenance cycle exceeding score Among them, W1, W2, W3, W4, and W5 are preset weight coefficients.

[0053] Specifically, the failure probability score is a comprehensive indicator obtained by weighted summation of scores from multiple dimensions. The cross-validation score reflects the degree of coordinated change among different types of sensors on the fan within the same time window; the instantaneous repetition score and short-term trend deviation score quantify the frequency of potential anomalies at different time scales and their deviation from short-term trend data, respectively; the near-life score and maintenance cycle exceedance score assess the fan's proximity to preset lifespan and maintenance cycle based on its cumulative operating time and maintenance records. W1, W2, W3, W4, and W5 are preset weighting coefficients used to adjust the relative importance of each score item in the final failure probability score. These weighting coefficients can be flexibly configured according to actual application scenarios, fan type, historical failure data, and expert experience to ensure that the score accurately reflects the fan's actual operating condition and potential failure risks.

[0054] Through the above technical solution, this application provides a structured and configurable method for calculating fault probability scores. This method not only enables precise quantification of multi-dimensional analysis results but also significantly enhances adaptability to different fault modes and operating environments by introducing adjustable weight coefficients. Engineers can flexibly adjust the weights of various influencing factors based on actual needs and experience, thereby optimizing the accuracy and sensitivity of fault diagnosis. This allows for more effective identification of potential fan faults and provides a reliable basis for timely maintenance measures, significantly improving the intelligence and precision of fan operation monitoring.

[0055] In some embodiments, the step of preprocessing and feature extraction of the sensor data to identify potential anomalies exceeding a preset fluctuation range or instantaneous anomaly threshold includes: The current operating mode and environmental parameters of the fan are obtained. Based on the current operating mode and environmental parameters, the filtering parameters and noise reduction intensity for removing industrial environmental interference are dynamically adjusted. The sensor data is preprocessed to obtain the first sensor data after noise reduction and purification. Based on the first sensor data, short-time Fourier transform or wavelet transform are used to extract the spectral features and harmonic features of the first sensor data, with a focus on capturing early weak fault signals. The extracted features are normalized, the feature distribution characteristics are evaluated, and the normalization parameters are dynamically adjusted. The normalized features are compared with the baseline features of the fan under the current operating mode and environmental parameters to identify potential anomalies that exceed the preset fluctuation range or instantaneous anomaly threshold.

[0056] Specifically, acquiring the fan's current operating mode and environmental parameters involves real-time monitoring of the fan's operating status, such as speed, load, and start / stop status, as well as its environmental conditions, such as ambient temperature, humidity, air pressure, and background noise. These parameters are used to dynamically adjust the filtering parameters and noise reduction intensity in the preprocessing stage. For example, when the fan is operating at high speed or the ambient noise is high, the filter's cutoff frequency can be increased or the noise reduction algorithm's intensity can be enhanced to more effectively remove interference signals in the industrial environment, thereby obtaining cleaner first-sensor data.

[0057] Based on the data from the first sensor, short-time Fourier transform (STFT) or wavelet transform is used to extract spectral and harmonic features, focusing on capturing early, weak fault signals. STFT and wavelet transform are two commonly used time-frequency analysis methods that can convert time-domain signals to the frequency domain, revealing the energy distribution of the signal at different frequency components and its changes over time. Spectral features can include energy, peak frequency, and amplitude within a specific frequency range; harmonic features focus on the amplitude and phase relationship of the fundamental frequency and its harmonics. These features are particularly effective in identifying weak vibration or current fluctuation signals caused by early faults such as fan bearing wear, imbalance, and misalignment, because these faults often exhibit anomalies at specific frequencies or their harmonics.

[0058] In practical applications, extracted features are normalized, their distribution characteristics are evaluated, and normalization parameters are dynamically adjusted. Normalization aims to eliminate differences in units and numerical ranges between different features, bringing them to a uniform scale for easier subsequent comparison and analysis. Normalization parameters can include mean, standard deviation, maximum, and minimum values. By evaluating the distribution characteristics of the normalized features—such as whether they conform to a normal distribution and whether there is skewness—normalization parameters can be dynamically adjusted to ensure that features maintain good comparability under different operating conditions and to avoid affecting the accuracy of outlier identification due to changes in data distribution.

[0059] Furthermore, the normalized features are compared with the fan's baseline features under the current operating mode and environmental parameters to identify potential anomalies exceeding preset fluctuation ranges or instantaneous anomaly thresholds. The baseline features are the set of features of the fan under specific operating modes and environmental parameters during normal operation. By comparing the real-time extracted and normalized features with these baseline features, the degree of deviation between the current operating state and the normal state can be quantified. The preset fluctuation range and instantaneous anomaly threshold are set based on historical data and expert experience to determine whether the deviation reaches an abnormal level. Feature values ​​exceeding these ranges or thresholds are identified as potential anomalies.

[0060] Through the above technical solutions, this application overcomes the problems of false alarms or missed alarms caused by static preprocessing and feature extraction strategies in complex and ever-changing industrial environments, which are often resulting from traditional methods. Specifically, dynamically adjusting filtering parameters and noise reduction intensity allows for adaptation to different operating modes and environmental noise levels, effectively improving data quality. Employing short-time Fourier transform or wavelet transform enhances the ability to capture early, weak fault signals, significantly improving the timeliness of fault warnings. Simultaneously, adaptive feature normalization and the comparison mechanism with benchmark features ensure the accuracy and robustness of anomaly identification, maintaining high detection performance even when fan operating conditions change, thus providing a more accurate and reliable monitoring basis for the reliable operation of fans.

[0061] For example, suppose a large industrial fan is operating in a chemical plant, with three operating modes: low, medium, and high speed, and the ambient temperature and humidity vary with the seasons. When the fan switches from low-speed to high-speed mode, the fundamental frequency and harmonic frequency of its bearing vibration signal will increase accordingly, and the motor current and temperature will also change. If static preprocessing parameters are used, normal vibrations under high-speed operation may be misjudged as abnormal, or early fault signals may be missed due to insufficient filtering during low-speed operation.

[0062] According to the technical solution of this application, parameters such as the fan's current high-speed operation mode and the ambient temperature of 30℃ are first obtained. Based on these parameters, the bandpass filtering range of the vibration sensor data is dynamically adjusted. For example, the filtering range is adjusted to a wider frequency range to adapt to the characteristic frequencies under high-speed operation, and the noise reduction algorithm is enhanced to cope with the greater background noise that may be brought about by high-speed operation. The first sensor data obtained after preprocessing is then subjected to time-frequency analysis through wavelet transform to extract specific high-frequency harmonic features that may occur in bearing failure under high-speed mode. After normalization processing by dynamically adjusted parameters, these features are compared with the normal reference features of the fan under "high-speed operation mode, ambient temperature 30℃". For example, if the amplitude of a certain high-frequency harmonic is detected to continuously exceed the preset fluctuation range under this operating condition, even if its amplitude change is relatively weak, it can be identified as a potential anomaly, and a multi-dimensional analysis process can be further initiated. This ability of dynamic adaptation and refined analysis enables this technical solution to effectively distinguish between changes in normal operating conditions and early signs of failure, avoid false alarms, and improve the sensitivity to potential failures.

[0063] In some embodiments, the steps of normalizing the extracted features, evaluating the feature distribution characteristics, and dynamically adjusting the normalization parameters include: Based on the current operating mode of the fan and environmental parameters, select the normalized parameters; The extracted features are normalized using the normalization parameters to obtain normalized features. Evaluate the distribution characteristics of the normalized features, and adjust the normalization parameters based on the evaluation results.

[0064] The current operating mode and environmental parameters of the fan refer to data reflecting the fan's working status and external environmental conditions under specific operating conditions, such as fan speed, load, ambient temperature, and humidity. Normalization parameters are coefficients or functions used to transform feature data with different dimensions or numerical ranges to a uniform scale. For example, Min-Max normalization, Z-score normalization, or other nonlinear normalization methods can be selected based on the current operating mode and environmental parameters. Choosing appropriate normalization parameters aims to ensure that feature data can be effectively standardized under different operating conditions, thereby eliminating the influence of dimensional differences on subsequent analysis. The purpose of normalization is to eliminate differences in dimensions and numerical ranges between different features, ensuring that all features have the same weight in subsequent comparisons and analyses, and preventing certain features with larger values ​​from dominating the analysis results.

[0065] Furthermore, evaluating the distribution characteristics of the normalized features can include calculating statistics such as mean, variance, skewness, and kurtosis, as well as analyzing the overall distribution shape of the data. By evaluating these characteristics, the effectiveness of the normalization process can be understood, such as whether the data is uniformly distributed, whether outliers exist, and whether it conforms to the expected statistical model. Based on the evaluation results of the distribution characteristics of the normalized features, the initially selected normalization parameters can be iteratively adjusted. For example, if severe skewness or outliers are still found in the normalized data, it may be necessary to reselect the normalization method or adjust its internal parameters to better adapt to the characteristics of the current data and ensure optimal normalization results.

[0066] Through the above technical solution, the normalization parameters can be dynamically selected and adjusted according to the actual operating mode and environmental parameters, making the feature normalization process more adaptable and robust. This effectively solves the problem of feature distortion or information loss that may occur when the fan operating conditions change, which is a problem with traditional fixed normalization methods. Therefore, this application can more accurately compare the normalized features with the baseline features, significantly improving the identification accuracy and sensitivity of potential anomalies, especially early weak fault signals, reducing the risk of false alarms and missed alarms, and thus providing a more reliable basis for fan operation monitoring.

[0067] In some embodiments, after the step of comparing the normalized features with the baseline features of the fan under the current operating mode and environmental parameters to identify potential anomalies exceeding a preset fluctuation range or instantaneous anomaly threshold, the method further includes: Based on the current operating mode of the fan and environmental parameters, a set of benchmark features matching the current operating conditions is selected from a preset benchmark feature library; Based on the benchmark feature set, calculate the deviation between the normalized features and the benchmark feature set; The system monitors the fan's running time and the update cycle of the baseline feature set. When the running time reaches a preset threshold or the update cycle reaches a preset threshold, the system triggers the baseline feature update process.

[0068] Specifically, selecting a set of benchmark features matching the current operating condition from a pre-set benchmark feature library involves searching and loading the benchmark feature data that best matches the current actual operating condition from the pre-stored library based on the fan's current operating mode (e.g., speed, load) and environmental parameters (e.g., ambient temperature, humidity). This benchmark feature library can contain normal operating features of the fan under different operating modes and combinations of environmental parameters. Based on the benchmark feature set, the deviation between the normalized features and the benchmark feature set is calculated. This can be understood as quantifying the degree of difference between the currently collected and processed normalized features and the selected benchmark feature set using a distance metric method (e.g., Euclidean distance, Mahalanobis distance, or cosine similarity). The greater the deviation, the greater the difference between the current state and the normal benchmark state. In practical applications, the fan's operating time and the update cycle of the benchmark feature set are monitored. When the operating time reaches a preset threshold or the update cycle reaches a preset threshold, the benchmark feature update process is triggered to ensure the timeliness and accuracy of the benchmark features. When the running time reaches a preset threshold, such as a certain cumulative running hour, it may mean that the fan has entered a wear or aging stage and needs to be re-established as a baseline. When the update cycle reaches a preset threshold, such as when the baseline characteristics have exceeded a certain number of days since the last update, an update will be forced to adapt to long-term changes in the environment or equipment status. Triggering the baseline characteristic update process means that a dedicated program will be started to re-collect and process data and generate new baseline characteristics.

[0069] Through the above technical solution, this application significantly improves the accuracy and robustness of sensor-based fan operation monitoring methods. Dynamically selecting a set of baseline features matching the current operating conditions effectively avoids false alarms caused by changes in operating conditions, making anomaly identification more accurate. Simultaneously, by introducing an automatic update mechanism for the baseline features, it can adapt to long-term changes in fan operating status and environmental drift, ensuring the timeliness and effectiveness of the baseline features, thereby reducing the risk of missed alarms. This not only improves the reliability of early fault warnings but also reduces the frequency of manual intervention and maintenance, extends the effective service life of the monitoring system, and provides more reliable technical support for predictive maintenance of industrial fans.

[0070] For example, suppose an industrial fan operates in summer and winter modes, and its normal vibration spectrum and motor temperature characteristics differ significantly. Traditional monitoring methods may use only a fixed reference feature, leading to numerous false alarms during seasonal transitions. The technical solution of this application automatically selects a set of reference features from a preset reference feature library that matches the current season and operating mode based on ambient temperature parameters (e.g., high temperature in summer, low temperature in winter) and the fan's operating mode (e.g., high speed operation, low speed operation). For example, in summer high-speed operation mode, the reference vibration spectrum and motor temperature range for summer high-speed operation mode are compared. Further, suppose the fan has been running continuously for a year, and its bearings may have slight wear, causing a slight increase in the baseline of the normal vibration signal. If the reference features are not updated for a long time, this normal slight drift may be misjudged as an early fault. The technical solution of this application monitors the cumulative operating time of the fan, and when a preset operating time threshold (e.g., 8000 hours) is reached, a reference feature update process is automatically triggered. In this update process, when the fan is in a stable and normal operating state, sensor data is re-acquired, new features are extracted, and compared with historical reference features. If a long-term slow drift is detected, a baseline feature library will be generated and updated so that the new baseline features can accurately reflect the current healthy operating status of the fan, thereby avoiding misjudging normal equipment aging trends as faults and ensuring the accuracy of monitoring.

[0071] In some embodiments, the step of triggering the update process of the baseline feature includes: Under the current operating mode and environmental parameters of the fan, sensor data is collected and processed to extract features that reflect the operating characteristics of the fan; By comparing the characteristics reflecting the fan's operating performance with historical baseline characteristics, it can be determined whether there is a long-term slow drift. If there is a long-term slow drift, a new set of benchmark features is generated based on the extracted features and the historical benchmark features, and the benchmark feature library is updated.

[0072] Specifically, when triggering the baseline feature update process, sensor data needs to be collected and processed under the current fan operating mode and environmental parameters to obtain features reflecting the current operating characteristics of the fan. These features can be spectral features, harmonic features, etc., aiming to comprehensively characterize the fan's operating state. The current fan operating mode can refer to operating parameters such as fan speed and load, while environmental parameters can refer to ambient temperature, humidity, etc.

[0073] Furthermore, the currently collected and extracted features reflecting fan operating characteristics are compared with historical baseline features. These historical baseline features refer to representative feature data accumulated at different time points or within different operating cycles under healthy fan operating conditions. This comparison allows for the determination of whether the fan's operating characteristics exhibit long-term, slow drift. Long-term, slow drift refers to the fan's feature values ​​showing a non-sudden, gradual trend of change over a longer timescale. This change is usually due to normal wear, aging, or environmental adaptation adjustments, rather than instantaneous failure.

[0074] If a long-term, slow drift is identified, a new set of benchmark features is generated based on the currently extracted features and the historical benchmark features. The generation of this new set of benchmark features can employ various methods, such as weighted averaging, trend fitting, or adaptive learning, to ensure that the new set accurately reflects the fan's operating characteristics in its current healthy state while also considering its historical evolution trend. Subsequently, the generated new set of benchmark features is updated in the benchmark feature library for use in subsequent anomaly identification and fault diagnosis.

[0075] Through the above technical solution, this application enables long-term, stable, and high-precision monitoring of fan operating status. Specifically, by determining whether there is a long-term, slow drift and updating the baseline characteristics accordingly, it ensures that the baseline characteristics always remain consistent with the actual healthy operating state of the fan, even if the fan undergoes normal aging or performance fine-tuning during long-term operation. This significantly improves the accuracy of anomaly identification, effectively reduces false alarms caused by changes in the normal operating characteristics of the fan, and enhances the ability to capture early, weak fault signals, as the dynamic updating of the baseline characteristics allows for more accurate identification of subtle deviations from the current healthy state. Therefore, the monitoring method of this application has stronger robustness and adaptability, and can better cope with the complexity and dynamism of fan operating characteristics in industrial environments.

[0076] For example, suppose that when an industrial fan is initially put into operation, the spectral characteristics of its bearing vibration signal exhibit a stable amplitude distribution in a specific frequency range, and these data are recorded as initial baseline characteristics. As the fan runs continuously for several months, due to normal bearing wear, the amplitude of certain frequency components of its vibration signal may show a slow and continuous small increase, but has not yet reached the fault threshold.

[0077] When a baseline feature is detected as needing an update, the current fan bearing vibration data is collected, and its spectral characteristics are extracted. These currently extracted spectral characteristics are then compared with historical baseline features. Through trend analysis or statistical testing, a continuous, slowly increasing trend is found between the current features and the initial baseline features; this is determined to be a long-term slow drift.

[0078] In this scenario, instead of simply using the current features as the new baseline, a new set of baseline features is generated based on the current features and historical baseline features, using algorithms such as Exponentially Weighted Moving Average (EWMA) to reflect this normal aging trend. For example, the new set of baseline features might slightly increase the amplitude of specific frequency components to adapt to the fan's current health status. Subsequently, this new set of baseline features is updated in the baseline feature library.

[0079] In this way, subsequent anomaly identification will be based on comparisons with this updated baseline characteristic, which better reflects the fan's current health status. If the fan subsequently experiences a genuine malfunction, such as a bearing failure causing a sudden and significant increase in vibration amplitude, this anomaly will be accurately identified, without misinterpreting normal long-term drift as a malfunction or missing genuine fault signals due to outdated baseline characteristics.

[0080] In some embodiments, after the step of focusing on capturing early, weak fault signals, the method further includes: If a potential weak fault indication that matches the preset fault characteristic frequency is identified, an operating state incentive strategy within a safe range is generated. The operating state incentive strategy includes speed adjustment, load transient incentive, or start-stop transient incentive. During the execution of the excitation strategy, sensor data is collected at a higher sampling rate, and the signal changes before and after excitation are compared. If the signal amplitude in the target frequency band increases by more than a preset ratio or a new fault resonance peak appears, the fault probability score is improved.

[0081] Specifically, when potential weak fault signs matching preset fault characteristic frequencies are identified through preprocessing and feature extraction of sensor data—for example, a slight increase in energy at a specific frequency detected in spectrum analysis, but not yet reaching the threshold for diagnosing a fault—a safe-range operating state incentive strategy is generated. This "safe-range operating state incentive strategy" refers to a strategy that, without compromising the normal operation and safety of the equipment, briefly alters the fan's operating conditions to amplify potential fault characteristics. For example, the operating state incentive strategy may include "speed adjustment," which involves slightly increasing or decreasing the fan speed for a short period to observe whether the fault characteristic frequency changes or strengthens accordingly; "load transient incentive," which involves briefly increasing or decreasing the fan load to observe the response of the fault signal under different load conditions; or "start-stop transient incentive," which, if safe to do so, performs a brief start-stop operation to capture fault characteristics that may appear during startup or shutdown. These incentive strategies aim to make originally weak fault signals more significant under specific operating conditions by altering the fan's dynamic response.

[0082] During the execution of the aforementioned operational state excitation strategy, sensor data is collected at a "higher sampling rate." This "higher sampling rate" refers to acquiring data at more frequent time intervals compared to the sampling frequency used in conventional monitoring. The purpose is to more precisely capture transient signal changes and high-frequency fault characteristics that may occur during the excitation process, avoiding the omission of crucial information due to insufficient sampling rate. Subsequently, a comparative analysis of the "signal changes before and after excitation" is performed. This comparative analysis aims to isolate the effects of the excitation, thereby more clearly identifying signal enhancements or pattern changes related to the fault. For example, the signal amplitude changes in the target frequency range before and after excitation can be compared, or whether new fault resonance peaks appear after excitation can be observed. The "target frequency range" typically refers to the characteristic frequency range associated with a specific fault type (such as bearing failure, blade imbalance, etc.). If, within this target frequency range, the signal amplitude enhancement exceeds a preset proportion, or a new fault resonance peak appears that was not previously present, it indicates that potential weak fault signs have been effectively verified. In this case, the "fault probability score" is "increased" to reflect a higher confidence level in the existence of the fault.

[0083] Through the above technical solution, this application can significantly improve the detection sensitivity and accuracy of early-stage weak fault signals. Traditional passive monitoring methods often struggle to distinguish extremely weak fault signals from normal noise or transient interference, easily leading to missed or false alarms. This application, by introducing an active excitation strategy, can specifically amplify potential fault characteristics, making these weak signals detectable and verifiable under controlled conditions. This not only helps to detect problems in their early stages, thus gaining valuable repair time and preventing fault escalation, but also provides a more solid basis for subsequent maintenance decisions by improving the fault probability score, effectively reducing unplanned equipment downtime and maintenance costs, and extending equipment lifespan.

[0084] This application also proposes a sensor-based fan operation monitoring system, such as... Figure 2 As shown, a sensor-based fan operation monitoring system 100 includes: The data acquisition module 10 is used to acquire multiple different types of sensor data reflecting the operating status of the fan in the industrial environment. The sensor data includes fan bearing vibration data, motor temperature data, and motor current data. Anomaly identification module 20 is used to preprocess and extract features from the sensor data to identify potential anomalies that exceed a preset fluctuation range or instantaneous anomaly threshold. The potential anomalies include continuous abnormal signals and early weak fault signals. The multi-dimensional analysis module 30 is used to initiate a multi-dimensional analysis process for the potential anomalies and to quantitatively evaluate and obtain a failure probability score. The evaluation and decision module 40 is used to determine whether potential anomalies are caused by external interference. If they are caused by external interference, it is confirmed that there is no real fault in the fan body. If they are not caused by external interference and the fault probability score reaches the preset confidence level, it is confirmed that there is a real fault in the fan body.

[0085] This sensor-based fan operation monitoring system, through its modular architecture and the collaborative work of its modules, effectively solves the problems of false alarms and missed alarms in existing technologies. Compared with traditional single-threshold or simple rule monitoring systems, this system can more comprehensively acquire and process multi-type sensor data. Through refined anomaly identification and multi-dimensional analysis, it improves the ability to capture early, weak fault signals. More importantly, by introducing an external interference judgment mechanism in the evaluation and decision module, this system can accurately distinguish between anomalies caused by fan body failures and those caused by external environmental interference, thereby significantly reducing the false alarm rate and avoiding unnecessary maintenance work and resource waste. This systematic solution not only improves the accuracy and reliability of fan operation monitoring but also provides stronger technical support for predictive maintenance of industrial equipment, thus ensuring the continuity and safety of industrial production.

[0086] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A sensor-based fan operation monitoring method, characterized in that, include: Acquire multiple types of sensor data reflecting the operating status of a fan in an industrial environment, including fan bearing vibration data, motor temperature data, and motor current data; The sensor data is preprocessed and features are extracted to identify potential anomalies that exceed a preset fluctuation range or instantaneous anomaly threshold. These potential anomalies include continuous abnormal signals and early weak fault signals. A multi-dimensional analysis process is initiated for the potential anomalies, and a fault probability score is obtained through quantitative evaluation. Determine whether the potential anomaly is caused by external interference. If it is caused by external interference, confirm that there is no real fault in the fan body. If it is not caused by external interference and the fault probability score reaches the preset confidence level, confirm that there is a real fault in the fan body.

2. The sensor-based fan operation monitoring method according to claim 1, characterized in that, The step of determining whether a potential anomaly is caused by external interference includes: Acquire environmental reference data collected by environmental vibration sensors, environmental electromagnetic field sensors, and environmental temperature sensors; If, within a preset time difference, the data from the environmental vibration sensor shows an impact vibration exceeding a preset first threshold, the data from the environmental electromagnetic field sensor shows an electromagnetic pulse exceeding a preset second threshold, or the data from the environmental temperature sensor shows an instantaneous temperature surge exceeding a preset third threshold, then the potential anomaly is determined to be caused by external interference; otherwise, the potential anomaly is determined to be caused by non-external interference.

3. The sensor-based fan operation monitoring method according to claim 1, characterized in that, The steps of initiating a multi-dimensional analysis process for the potential anomalies and quantitatively evaluating and obtaining a failure probability score include: Check whether there are coordinated changes in different types of sensors on the fan within the same time window, and generate cross-validation scores based on the degree of coordinated change; Analyze the frequency of occurrence of potential outliers at different time scales and their degree of deviation from short-term trend data to generate instantaneous repetition scores and short-term trend deviation scores. Based on the cumulative operating time and maintenance records of the fan, and according to the degree of closeness to the preset lifespan and maintenance cycle, a lifespan closeness score and a maintenance cycle exceeding score are generated. The failure probability score is obtained by weighting and summing the cross-validation score, the instantaneous repetition score, the short-term trend deviation score, the near-life score, and the maintenance cycle exceedance score based on preset weighting coefficients.

4. The sensor-based fan operation monitoring method according to claim 3, characterized in that, The probability score of failure is obtained using the following formula: Failure probability score = W1 * Cross-validation score + W2 * Instantaneous repetition score + W3 * Short-term trend deviation score + W4 * Lifespan approaching score + W5 * Maintenance cycle exceeding score Among them, W1, W2, W3, W4, and W5 are preset weight coefficients.

5. The sensor-based fan operation monitoring method according to claim 1, characterized in that, The step of preprocessing and feature extraction of the sensor data to identify potential anomalies exceeding a preset fluctuation range or instantaneous anomaly threshold includes: The current operating mode and environmental parameters of the fan are obtained. Based on the current operating mode and environmental parameters, the filtering parameters and noise reduction intensity for removing industrial environmental interference are dynamically adjusted. The sensor data is preprocessed to obtain the first sensor data after noise reduction and purification. Based on the first sensor data, short-time Fourier transform or wavelet transform are used to extract the spectral features and harmonic features of the first sensor data, with a focus on capturing early weak fault signals. The extracted features are normalized, the feature distribution characteristics are evaluated, and the normalization parameters are dynamically adjusted. The normalized features are compared with the baseline features of the fan under the current operating mode and environmental parameters to identify potential anomalies that exceed the preset fluctuation range or instantaneous anomaly threshold.

6. The sensor-based fan operation monitoring method according to claim 5, characterized in that, The steps of normalizing the extracted features, evaluating the feature distribution characteristics, and dynamically adjusting the normalization parameters include: Based on the current operating mode of the fan and environmental parameters, select the normalized parameters; The extracted features are normalized using the normalization parameters to obtain normalized features. Evaluate the distribution characteristics of the normalized features, and adjust the normalization parameters based on the evaluation results.

7. The sensor-based fan operation monitoring method according to claim 5, characterized in that, After the step of comparing the normalized features with the baseline features of the fan under the current operating mode and environmental parameters to identify potential anomalies exceeding the preset fluctuation range or instantaneous anomaly threshold, the method further includes: Based on the current operating mode of the fan and environmental parameters, a set of benchmark features matching the current operating conditions is selected from a preset benchmark feature library; Based on the benchmark feature set, calculate the deviation between the normalized features and the benchmark feature set; The system monitors the fan's running time and the update cycle of the baseline feature set. When the running time reaches a preset threshold or the update cycle reaches a preset threshold, the system triggers the baseline feature update process.

8. The sensor-based fan operation monitoring method according to claim 7, characterized in that, The steps of the triggering baseline feature update process include: Under the current operating mode and environmental parameters of the fan, sensor data is collected and processed to extract features that reflect the operating characteristics of the fan; By comparing the characteristics reflecting the fan's operating performance with historical baseline characteristics, it can be determined whether there is a long-term slow drift. If there is a long-term slow drift, a new set of benchmark features is generated based on the extracted features and the historical benchmark features, and the benchmark feature library is updated.

9. The sensor-based fan operation monitoring method according to claim 5, characterized in that, Following the step of focusing on capturing early, weak fault signals, the method further includes: If a potential weak fault indication that matches the preset fault characteristic frequency is identified, an operating state incentive strategy within a safe range is generated. The operating state incentive strategy includes speed adjustment, load transient incentive, or start-stop transient incentive. During the execution of the excitation strategy, sensor data is collected at a higher sampling rate, and the signal changes before and after excitation are compared. If the signal amplitude in the target frequency band increases by more than a preset ratio or a new fault resonance peak appears, the fault probability score is improved.

10. A sensor-based fan operation monitoring system, characterized in that, The system includes: The data acquisition module is used to acquire multiple different types of sensor data reflecting the operating status of the fan in the industrial environment. The sensor data includes fan bearing vibration data, motor temperature data, and motor current data. Anomaly identification module is used to preprocess and extract features from the sensor data to identify potential anomalies that exceed a preset fluctuation range or instantaneous anomaly threshold. The potential anomalies include continuous abnormal signals and early weak fault signals. The multi-dimensional analysis module is used to initiate a multi-dimensional analysis process for the potential anomalies and to quantitatively evaluate and obtain a failure probability score. The evaluation and decision-making module is used to determine whether potential anomalies are caused by external interference. If they are caused by external interference, it is confirmed that there is no real fault in the fan body. If they are not caused by external interference and the fault probability score reaches the preset confidence level, it is confirmed that there is a real fault in the fan body.

Images