Intelligent early warning method and system for bearing vibration of primary air fan

Abstract

The invention discloses a primary fan bearing vibration intelligent early warning method and system. A single vibration signal is easily interfered by working condition fluctuation (such as load change and inlet pressure sudden change) of equipment and environmental noise, so that the false alarm rate is high. The method comprises the following steps: multi-source data synchronous acquisition; working condition adaptive parameter adjustment; dynamic threshold generation; performing multi-source information fusion early warning, wherein the auxiliary parameters comprise at least one of bearing temperature, motor three-phase current unbalance degree and fan inlet pressure fluctuation amplitude; and when the main early warning criterion and the at least one auxiliary early warning criterion are met at the same time, triggering a graded early warning signal. The method is used for intelligent early warning of the primary fan bearing vibration.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment condition monitoring technology, specifically a vibration intelligent early warning method and system for primary air fans, which achieves early fault warning through multi-source information fusion and dynamic threshold adjustment. Background Technology

[0002] Wind turbines are critical equipment in process industries such as power generation, metallurgy, and chemicals, and the operating condition of their bearings directly affects the safety and stability of the unit. Currently, wind turbine vibration monitoring commonly employs online monitoring systems based on a single acceleration sensor, with fixed alarm thresholds. However, this method has the following significant drawbacks: 1. Poor anti-interference ability: Single vibration signals are easily affected by fluctuations in the equipment's own operating conditions (such as load changes and sudden changes in inlet pressure) and environmental noise, resulting in a high false alarm rate.

[0003] 2. Insufficient adaptability: Fixed thresholds cannot reflect the inherent variation of equipment vibration characteristics under different operating loads. Under low loads, this may lead to insufficient sensitivity and delayed fault warnings; under high loads, it may trigger false alarms frequently due to increased normal vibration levels.

[0004] 3. Weak fault identification capability: It is difficult to distinguish between vibration caused by real mechanical faults (such as bearing damage or rotor imbalance) and transient vibration caused by abnormal process parameters (such as drastic fluctuations in system pressure) based solely on vibration signals, which is not conducive to maintenance personnel making accurate judgments.

[0005] Therefore, there is an urgent need for a wind turbine bearing condition monitoring method that can adapt to changing operating conditions, integrate multi-dimensional information, and achieve accurate early warning. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for intelligent early warning of primary wind turbine bearing vibration, so as to solve the above-mentioned technical problems.

[0007] The above objectives are achieved through the following technical solutions: A method for intelligent early warning of primary air turbine bearing vibration, the method comprising the following steps: S1. Multi-source data synchronous acquisition: Real-time synchronous acquisition of bearing vibration acceleration signal, bearing temperature signal, motor current signal, fan inlet pressure signal and unit real-time load signal of primary air fan; S2. Adaptive Parameter Adjustment: The load change rate is calculated based on the real-time load signal of the unit. When the load change rate exceeds the preset first threshold, a dynamic filtering algorithm is triggered to suppress the vibration transient interference caused by sudden load changes. S3. Dynamic threshold generation: Based on the real-time load signal, a preset load-vibration reference curve library is matched to obtain the reference vibration level and statistical characteristics corresponding to the current load interval, and a dynamic alarm threshold is generated accordingly; the load-vibration reference curve library is established based on historical data according to the statistical vibration amplitude distribution of the load interval. S4. Multi-source information fusion early warning: Determine whether the real-time vibration value exceeds the dynamic alarm threshold and continues for a first preset duration. If so, trigger the main early warning criterion. At the same time, monitor whether one or more auxiliary parameters meet the corresponding auxiliary early warning criteria. The auxiliary parameters include at least one of the following: bearing temperature, motor three-phase current imbalance, and fan inlet pressure fluctuation amplitude. When the main early warning criterion and at least one of the auxiliary early warning criteria are met simultaneously, trigger a graded early warning signal.

[0008] In the aforementioned intelligent early warning method for primary wind turbine bearing vibration, step S1 includes the bearing vibration acceleration signal comprising acceleration values ​​measured in the X, Y, and Z directions of the bearing housing.

[0009] In the aforementioned intelligent early warning method for primary wind turbine bearing vibration, step S2 involves calculating the load change rate using a sliding time window.

[0010] In the aforementioned intelligent early warning method for primary wind turbine bearing vibration, step S3 involves constructing the load-vibration reference curve library as follows: during the offline phase, historical load data is divided into multiple continuous load intervals, and the historical vibration amplitude data within each interval is statistically analyzed to determine its confidence interval as the vibration reference range for that interval; the dynamic alarm threshold is calculated based on the reference range and the standard deviation in the statistical characteristics.

[0011] In the aforementioned intelligent early warning method for primary wind turbine bearing vibration, step S4 includes the following auxiliary early warning criteria: (4) The bearing temperature exceeds the temperature threshold and the rate of temperature rise exceeds the rate of temperature rise threshold; (4) The three-phase current imbalance of the motor exceeds the current imbalance threshold; (5) The pressure fluctuation at the inlet of the blower exceeds the pressure fluctuation threshold.

[0012] In the aforementioned intelligent early warning method for primary wind turbine bearing vibration, step S4 involves a graded early warning system comprising multiple early warning levels. The final early warning level is determined based on the degree of exceedance of the primary early warning criterion and the type and number of the triggered auxiliary early warning criteria.

[0013] A system for intelligent early warning of primary air turbine bearing vibration comprises: The data acquisition layer includes a sensor network deployed on the primary wind turbine for acquiring the multi-source data; The data analysis layer includes: The operating condition identification module is used to identify the load range to which the current real-time load belongs; The dynamic threshold engine, which incorporates the load-vibration reference curve library, is used to generate the dynamic alarm threshold based on the output of the working condition identification module. The fusion decision module is used to execute the multi-source information fusion early warning logic and output the early warning level; The application layer is used to receive and display the warning level. Beneficial effects

[0014] 1. This invention integrates information from multiple sources such as vibration, temperature, current, and pressure to make comprehensive decisions, effectively distinguishing between real mechanical faults and process interference, and significantly reducing false alarm rate and false negative rate.

[0015] 2. By constructing a load-vibration reference curve library, this invention achieves dynamic adaptive adjustment of the alarm threshold, enabling the early warning system to adapt to the normal operating conditions of equipment under various load conditions, thereby improving the sensitivity and rationality of monitoring.

[0016] 3. This invention adopts a hierarchical early warning mechanism that combines primary and secondary criteria, which can more precisely characterize the severity and development stage of the fault, and provide maintenance personnel with tiered decision support from "prompt inspection" to "emergency handling", thereby improving the pertinence and efficiency of maintenance.

[0017] 4. The system architecture of this invention is clear, combining edge computing with cloud / server analysis, balancing real-time performance and computational depth. Specific implementation cases demonstrate its effectiveness and reliability in real-world industrial scenarios. Attached Figure Description

[0018] Appendix Figure 1 This is an architectural diagram of the present invention; Detailed Implementation

[0019] Reference Figure 1 A method for intelligent early warning of primary wind turbine bearing vibration includes the following steps: S1. Multi-source data synchronous acquisition: Deploy a sensor network on the primary air turbine to synchronously acquire multi-source monitoring data, including bearing vibration acceleration, bearing temperature, motor current, and turbine inlet pressure, as well as the unit's real-time load signal. S2. Adaptive Filtering under Working Conditions: Real-time load change rate of the computer group; when the load change rate exceeds a preset first threshold (e.g., 5% / min), a dynamic filtering algorithm is activated to filter the collected vibration acceleration signal in order to suppress vibration transient interference caused by sudden load changes; S3. Construct a load-vibration baseline curve library: Offline phase: Based on historical operating data, the vibration amplitude is statistically analyzed according to the preset load intervals (e.g., each interval is 5% load), a statistical distribution model of the vibration amplitude in each load interval is established, and its confidence interval (e.g., 95% confidence interval) is calculated, thereby forming the vibration reference curve and the corresponding reference vibration level for each load interval. Online phase: Based on the real-time collected unit load, match the load range to which it belongs, and call the corresponding benchmark vibration level and statistical characteristics (such as standard deviation σ) to generate dynamic alarm thresholds (such as benchmark value ±3σ). S4. Multi-source information fusion early warning decision-making: Set the main warning criterion: the real-time vibration value exceeds the dynamic alarm threshold and the duration exceeds the preset duration (e.g., 10 seconds). One or more auxiliary early warning criteria are set, and the auxiliary early warning criteria include: The bearing temperature exceeds the temperature threshold (e.g., 80℃) and the rate of temperature rise exceeds the rate threshold (e.g., 3℃ / min). The three-phase current imbalance of the motor exceeds the current imbalance threshold (e.g., 10%). The pressure fluctuation at the fan inlet exceeds the pressure fluctuation threshold (e.g., 15% of the rated pressure). When the main warning criterion and at least one of the auxiliary warning criterions are met simultaneously, a graded warning is triggered. Based on the combination of the severity of the main and auxiliary criterions, the warning is divided into multiple levels, including but not limited to: Level 1 warning (maintenance reminder), Level 2 warning (reduced load operation recommendation) and Level 3 warning (emergency shutdown recommendation).

[0020] In step S1, the sensor network includes vibration acceleration sensors deployed in the X, Y, and Z directions of the wind turbine bearing housing, temperature probes deployed in the drive end and non-drive end bearings, current transformers for measuring the three-phase current of the motor, and pressure transmitters installed at the wind turbine inlet.

[0021] In step S2, the load change rate is calculated using the sliding time window method.

[0022] The specific rules for the graded early warning system are as follows: a general alarm is issued when only the main early warning criterion is met; a level one early warning is triggered when both the main early warning criterion and a minor abnormal auxiliary criterion are met; a level two early warning is triggered when both the main early warning criterion and a significant abnormal auxiliary criterion or multiple auxiliary criteria are met; and a level three early warning is triggered when the vibration severely exceeds the limit and is accompanied by an extreme abnormal auxiliary criterion (such as a rapid rise in temperature approaching a dangerous value).

[0023] A primary air turbine bearing vibration intelligent early warning system based on multi-source information fusion, used to implement the above method, includes: Data acquisition layer: Includes a sensor network and edge computing nodes deployed on the primary wind turbine. The sensor network is used to collect multi-source monitoring data, and the edge computing nodes are used to realize synchronous data acquisition, preliminary processing and uploading. The synchronous sampling rate is not less than 10kHz. Data analysis layer: Communicatively connected to the data acquisition layer, including: The operating condition identification module is used to identify and classify the current operating load range of the unit in real time; The dynamic threshold engine, with the built-in load-vibration reference curve library, is used to match and generate dynamic alarm thresholds in real time based on the output of the working condition identification module. The fusion decision module integrates a rule engine and a lightweight anomaly detection model, and is used to perform fusion logic judgment based on the main warning criteria and auxiliary warning criteria, and output the warning level. Application layer: includes an early warning platform, which receives early warning information output by the data analysis layer, provides a visual human-machine interface for display, and outputs early warning signals to the power plant distributed control system (DCS) or other operation and maintenance management systems through a standard industrial interface (such as OPC UA).

[0024] The lightweight anomaly detection model is a Long Short-Term Memory (LSTM) network model, which is used to learn the time-series features of vibration signals to help identify complex anomaly patterns that are difficult for rule engines to cover. Example

[0025] The early warning system of this invention is deployed on the primary air turbine of the 300MW unit of Laicheng Power Plant.

[0026] Data collection: The vibration sensor is a PCB 352C33 type accelerometer (range ±50g), which is installed in three orthogonal directions on the fan bearing housing.

[0027] The temperature probe is a PT100 platinum resistance thermometer (accuracy ±0.1℃), which is installed on the outer ring of the bearing.

[0028] The effective value of the three-phase current of the motor is collected using existing current transformers.

[0029] The inlet static pressure is collected using the existing pressure transmitter at the fan inlet.

[0030] The unit load signal is obtained from the power plant's DCS system through the OPC interface.

[0031] All signals are acquired synchronously via high-speed edge computing nodes, with a sampling frequency set to 12kHz.

[0032] Baseline curve library construction: Before the system is put into operation, it first enters an offline learning phase. Historical normal operation data of the wind turbine over the past 3 months are collected.

[0033] The load range is divided into multiple intervals at 5% increments. For example, statistical analysis is performed on all vibration RMS data within the 50%-55% load range to calculate the mean and standard deviation, and the upper limit of its 95% confidence interval is determined as the benchmark threshold for that range (e.g., threshold A = 4.2 mm / s). Similarly, a benchmark threshold is established for the 80%-85% load range (e.g., threshold B = 6.8 mm / s). This process continues to form a complete benchmark curve library.

[0034] Online early warning operation: When the load of Unit 2 was 75%, the vibration value suddenly increased to 7.2 mm / s. The system triggered a level 2 warning within 1.5 seconds, and the investigation revealed a crack in the bearing cage.

Claims

1. A method for intelligent early warning of primary air turbine bearing vibration, characterized by: The method includes the following steps: S1. Multi-source data synchronous acquisition: Real-time synchronous acquisition of bearing vibration acceleration signal, bearing temperature signal, motor current signal, fan inlet pressure signal and unit real-time load signal of primary air fan; S2. Adaptive Parameter Adjustment: The load change rate is calculated based on the real-time load signal of the unit. When the load change rate exceeds the preset first threshold, a dynamic filtering algorithm is triggered to suppress the vibration transient interference caused by sudden load changes. S3. Dynamic threshold generation: Based on the real-time load signal, a preset load-vibration reference curve library is matched to obtain the reference vibration level and statistical characteristics corresponding to the current load interval, and a dynamic alarm threshold is generated accordingly; the load-vibration reference curve library is established based on historical data according to the statistical vibration amplitude distribution of the load interval. S4. Multi-source information fusion early warning: Determine whether the real-time vibration value exceeds the dynamic alarm threshold and continues for a first preset duration. If so, trigger the main early warning criterion. At the same time, monitor whether one or more auxiliary parameters meet the corresponding auxiliary early warning criteria. The auxiliary parameters include at least one of the following: bearing temperature, motor three-phase current imbalance, and fan inlet pressure fluctuation amplitude. When the main warning criterion and at least one of the auxiliary warning criteria are simultaneously satisfied, a graded warning signal is triggered.

2. The intelligent early warning method for primary air turbine bearing vibration according to claim 1, characterized in that: In step S1, the bearing vibration acceleration signal includes acceleration values ​​measured in the X, Y, and Z directions of the bearing housing.

3. The intelligent early warning method for primary air turbine bearing vibration according to claim 1, characterized in that: In step S2, the load change rate is calculated using a sliding time window.

4. The intelligent early warning method for primary air turbine bearing vibration according to claim 1, characterized in that: In step S3, the method for constructing the load-vibration reference curve library is as follows: In the offline stage, the historical load data is divided into multiple continuous load intervals, and the historical vibration amplitude data in each interval is statistically analyzed to determine its confidence interval as the vibration reference range of that interval. The dynamic alarm threshold is calculated based on the baseline range and the standard deviation of the statistical characteristics.

5. The intelligent early warning method for primary air turbine bearing vibration according to claim 1, characterized in that: Step S4 In this context, the auxiliary early warning criteria include: The bearing temperature exceeds the temperature threshold and the rate of temperature rise exceeds the temperature rise rate threshold. The three-phase current imbalance of the motor exceeds the current imbalance threshold. The pressure fluctuation at the fan inlet exceeds the pressure fluctuation threshold.

6. The intelligent early warning method for primary air turbine bearing vibration according to claim 1, characterized in that: In step S4, the graded early warning includes multiple early warning levels. The final early warning level is determined based on the degree of exceeding the limit of the main early warning criterion and the type and number of the auxiliary early warning criteria that are triggered.

7. A system for implementing the intelligent early warning method for primary air turbine bearing vibration as described in any one of claims 1-6, characterized in that: Its components include: The data acquisition layer includes a sensor network deployed on the primary wind turbine for acquiring the multi-source data; The data analysis layer includes: The operating condition identification module is used to identify the load range to which the current real-time load belongs; The dynamic threshold engine, which incorporates the load-vibration reference curve library, is used to generate the dynamic alarm threshold based on the output of the working condition identification module. The fusion decision module is used to execute the multi-source information fusion early warning logic and output the early warning level; The application layer is used to receive and display the warning level.

Images