Blower bearing temperature anomaly detection and early warning method

Abstract

The invention discloses an air feeder bearing temperature anomaly detection and early warning method, and belongs to the technical field of equipment state monitoring. Comprising the following steps: setting a blower bearing temperature data sampling period and parameters of a dynamic sliding window, and performing first-order difference processing on blower bearing temperature time sequence data observed by the dynamic sliding window to obtain a blower bearing temperature difference time sequence; in each dynamic sliding window, adopting a GARCH (1, 1) model to fit a temperature difference time sequence of the air feeder bearing; constructing a parameter matrix according to core parameters of the GARCH (1, 1) model; performing anomaly detection on detection points contained in the parameter matrix by using an LOF anomaly detection algorithm; if the LOF value of the detection point is larger than the threshold value, it is judged that the detection point is an abnormal point, and temperature abnormal fluctuation occurs in the dynamic sliding window associated with the detection point in an early warning mode. According to the method, the anomaly detection sensitivity is remarkably improved, the false alarm rate is reduced, and safe and stable operation of the power production system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment status monitoring, and in particular to a method for detecting and warning abnormal temperature of a blower bearing. Background Art

[0002] The blower in a large coal-fired power plant is a core component of boiler operation. It provides sufficient oxygen for combustion, precisely controls air volume to optimize combustion efficiency, stabilize furnace temperature, and reduce pollutant emissions. It also ensures stable steam parameters, directly impacting the safety and economics of power generation and a key component in efficient energy conversion. The blower's bearings are a key component for efficient rotor operation, supporting high-speed rotation and reducing friction losses. Their stability directly impacts blower reliability. Failure can lead to air supply interruption, combustion disruption, and even unit shutdown. Furthermore, the high-load, high-temperature environment places extremely high demands on bearing material precision and lubrication performance, making them crucial for the long-term, safe operation of the equipment.

[0003] Abnormal bearing temperatures in blower fans at large coal-fired power plants can trigger chain reactions, including: high temperatures exacerbating bearing wear, leading to raceway spalling, cage deformation, and even shaft seizure, causing equipment damage; abnormal vibrations may spread to the fan blades and motor, threatening the safe operation of the unit; in severe cases, protective shutdowns may be triggered, causing boiler combustion interruptions and affecting power generation stability. The main causes of abnormal bearing temperatures in blowers include lubrication failure (grease deterioration, insufficient oil), misaligned or poorly aligned bearings causing frictional heat, blockage of cooling ducts or failure of the heat dissipation system, coal dust infiltration damaging the lubricating film, and long-term overload operation accelerating metal fatigue. If such faults are not handled promptly, not only will maintenance costs increase, but they may also cause unplanned shutdowns, seriously affecting the economic viability and power supply reliability of the power plant. These faults require early prevention and control through intelligent detection and regular maintenance.

[0004] Detecting abnormal bearing temperatures in large coal-fired power plants' fan fans is a key step in preventing equipment failure. Currently, these systems rely on single-threshold alarms, which suffer from poor dynamic adaptability, insufficient multi-source data fusion, and weak ability to capture nonlinear features. These systems are susceptible to fluctuations in operating conditions, resulting in a high rate of misjudgments. Furthermore, they lack a deep integration mechanism with the full lifecycle management of the equipment, increasing operational complexity.

[0005] Therefore, a method for detecting and warning abnormal temperature of blower bearings is needed to achieve real-time monitoring, provide early warning of lubrication failure, bearing wear or cooling abnormalities, avoid serious accidents such as shaft seizure and blade breakage caused by high temperature, ensure continuous and stable operation of the unit, reduce unplanned downtime losses, and reduce maintenance costs, thereby supporting the safe production, economic benefits and environmental protection standards of power plants. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for detecting and warning abnormal temperature of a blower bearing, comprising the following steps:

[0007] Step 1: Set the sampling period of the fan bearing temperature data and the parameters of the dynamic sliding window, perform first-order difference processing on the fan bearing temperature time series data observed by the dynamic sliding window, and obtain the fan bearing temperature difference time series;

[0008] Step 2: In each dynamic sliding window, the GARCH (1,1) model is used to fit the fan bearing temperature difference time series and extract the core parameters of the GARCH (1,1) model;

[0009] Step 3: Construct a parameter matrix based on the core parameters of the GARCH(1,1) model;

[0010] Step 4: Use the LOF anomaly detection algorithm to perform anomaly detection on the detection points included in the parameter matrix; if the LOF value of the detection point is greater than the threshold, the detection point is determined to be an anomaly point, and an alarm is issued for abnormal temperature fluctuations in the dynamic sliding window associated with the detection point.

[0011] Furthermore, the parameters of the dynamic sliding window in step 1 include the number, length, and sliding step of the sliding windows.

[0012] Furthermore, the number of sliding windows is 1000.

[0013] Furthermore, the core parameters include long-term average variance, short-term impact coefficient, and long-term attenuation factor.

[0014] Furthermore, the parameter matrix in step 3

[0015]

[0016] Among them, ω i is the long-term average variance of the i-th detection point, α i is the short-term impact coefficient of the i-th detection point; β i is the long-term attenuation factor of the i-th detection point, i = 1, 2, …, n, and n is the number of detection points.

[0017] The beneficial effects of the present invention are:

[0018] 1. This paper captures the temperature fluctuation aggregation effect and time-varying characteristics through the GARCH (1,1) model, dynamically quantifies the conditional heteroskedasticity of the temperature series, and improves the flexibility of anomaly detection;

[0019] 2. The LOF algorithm, based on local density differences in data, can effectively identify complex and sparsely distributed outliers that are difficult to capture with traditional methods. Working in conjunction with the GARCH(1,1) model, it significantly improves anomaly detection sensitivity, reduces false positive rates, and effectively addresses the limitations of traditional static thresholding methods in nonlinear, time-varying scenarios.

[0020] 3. The method of the present invention can achieve efficient early warning before equipment failure occurs, and can ensure the safe and stable operation of the power production system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figure is a flow chart of the method for detecting and warning abnormal temperature of a blower bearing according to the present invention.

[0022] Figure 2 This is a data flow diagram of the blower bearing temperature anomaly detection and early warning method of the present invention.

[0023] Figure 3 This is the time series data of the bearing temperature of the fan in a power plant.

[0024] Figure 4 This is the first-order difference time series diagram of the bearing temperature time series data of the fan in a power plant.

[0025] Figure 5 is the core parameter point of the GARCH(1,1) model.

[0026] Figure 6 This is the distribution diagram of LOF value and threshold.

[0027] Figure 7 It is the abnormal point of the core parameter of the GARCH (1,1) model. DETAILED DESCRIPTION

[0028] The present invention provides a method for detecting and warning abnormal temperature of a blower bearing, which will be further described below with reference to the accompanying drawings and specific embodiments.

[0029] Figure 1 The figure is a flow chart of the method for detecting and warning abnormal temperature of a blower bearing according to the present invention.

[0030] Figure 2 This is the data flow diagram of the blower bearing temperature anomaly detection and early warning method of the present invention. Figure 2 It can be seen that the blower bearing temperature data is collected, stored in the SIS database of the power plant production equipment, and finally sent to the server for processing.

[0031] The method specifically includes:

[0032] Step 1: Set the fan bearing temperature data sampling period and dynamic sliding window parameters, perform first-order difference processing on the fan bearing temperature time series data observed by the dynamic sliding window, and obtain the fan bearing temperature difference time series, specifically:

[0033] Set the sampling period for the fan bearing temperature data to Δt = 60 seconds. Set the parameters for the dynamic sliding window: the number of sliding windows is n, set to n = 1000; m is set to 300, and the sliding window length is m + 1, meaning that the sliding window contains m + 1 fan bearing temperature time series data; the sliding window step is Δt × s (s ≥ 1), set to s = 60. The initial time for collecting fan bearing temperature data is t0, and the start and end time periods of the i-th sliding window are [t0 + (i - 1) × Δt × s, t0 + (i - 1) × Δt × s + m × Δt].

[0034] The dynamic sliding window is used to observe the time series of the fan bearing temperature, and the Figure 3 The figure shows the time series data of the bearing temperature of the fan in a power plant.

[0035] The first-order difference processing is performed on the m+1 blower bearing temperature time series data in the sliding window to obtain the blower bearing temperature difference time series. The length of the temperature difference time series is m, where m=300. Figure 4 This is the first-order difference time series diagram of the bearing temperature time series data of the fan in a power plant.

[0036] Step 2: In each dynamic sliding window, the GARCH (1,1) model is used to fit the fan bearing temperature difference time series, and the three core parameters of the GARCH (1,1) model variance equation are extracted.

[0037] Among them, the three core parameters include, ω: long-term average variance (constant term), α: short-term shock impact coefficient (reflecting the impact of residual square), β: long-term attenuation factor (reflecting the impact of historical volatility).

[0038] Step 3: Construct a parameter matrix based on the core parameters of the GARCH(1,1) model;

[0039] The three core parameters of the variance equation of the GARCH(1,1) model for the temperature difference time series within each sliding window and the three core parameters of the variance equation of the GARCH(1,1) model for the difference time series within n sliding windows constitute the n×3 parameter matrix Θ.

[0040]

[0041] It should be noted that the i-th row of the parameter matrix Θ corresponds to the core coefficient of the i-th GARCH(1,1) model; the first column: ω i (long-term variance parameter); second column: α i (short-term impact coefficient); third column: β i (long-term attenuation factor); where i = 1, 2, …, n.

[0042] Step 4: Use the LOF anomaly detection algorithm to perform anomaly detection on the detection points included in the parameter matrix; if the LOF value of the detection point is greater than the threshold, the detection point is determined to be an anomaly point, and an alarm is issued for abnormal temperature fluctuations in the dynamic sliding window associated with the detection point.

[0043] In three-dimensional space, the three core parameters of the variance equation of the GARCH (1,1) model form a point; the n×3 parameter matrix Θ represents n points. Using the LOF anomaly detection algorithm, anomaly detection is performed on n points, the same as the number of sliding windows, which is 1000, as shown in Figure 5 shown.

[0044] If a point (ω k , α k , β k ) If the LOF value is greater than the threshold, the point is determined to be an abnormal point, and an alert is issued for abnormal temperature fluctuations within the start and end time periods of the sliding window associated with the point [t0+(k-1)×Δt×s, t0+(k-1)×Δt×s+m×Δt].

[0045] Figure 6 is a distribution diagram of LOF value and threshold value. Figure 6 Dynamically set the Threshold. In the figure, the Threshold = 1.5178 is set for the LOF value at a certain point, where k∈[1,2,…,n].

[0046] The LOF value of a point is equal to the inverse of the ratio of its local density to the average density of its nearest neighboring points. The local outlier factor (LOF) is a density-based anomaly detection algorithm that identifies outliers by comparing the density difference between a data point and its neighbors, and finds data points whose density in a local area is significantly lower than that of the surrounding points as anomalies, such as Figure 7 It is the abnormal point of the core parameter of the GARCH (1,1) model.

[0047] In summary, the method of the present invention can capture the time-varying characteristics and memory effect of temperature fluctuations by identifying abnormal changes in the GARCH (1,1) model parameters of the blower bearing temperature difference series, and effectively distinguish between short-term anomalies of parameter mutations and long-term trend anomalies.

[0048] Compared with traditional technical methods, this method can effectively and significantly improve the efficiency of detecting abnormal operating conditions of blower bearings, reduce the false alarm rate of bearing warnings, and provide quantitative decision-making support for preventive maintenance of thermal power production equipment.

Claims

1. A method for detecting and warning abnormal temperature of a blower bearing, characterized in that: The following steps are involved: Step 1: Set the sampling period of the fan bearing temperature data and the parameters of the dynamic sliding window, perform first-order difference processing on the fan bearing temperature time series data observed by the dynamic sliding window, and obtain the fan bearing temperature difference time series; Step 2: In each dynamic sliding window, the GARCH (1,1) model is used to fit the fan bearing temperature difference time series and extract the core parameters of the GARCH (1,1) model; Step 3: Construct a parameter matrix based on the core parameters of the GARCH(1,1) model; Step 4: Use the LOF anomaly detection algorithm to perform anomaly detection on the detection points included in the parameter matrix; if the LOF value of the detection point is greater than the threshold, the detection point is determined to be an anomaly point, and an alarm is issued for abnormal temperature fluctuations in the dynamic sliding window associated with the detection point.

2. The method for detecting and warning abnormal temperature of a blower bearing according to claim 1, characterized in that: The parameters of the dynamic sliding window in step 1 include the number, length, and sliding step of the sliding windows.

3. The method for detecting and warning abnormal temperature of a blower bearing according to claim 2, characterized in that: The number of sliding windows is 1000.

4. The method for detecting and warning abnormal temperature of a blower bearing according to claim 1, characterized in that: The core parameters include long-term average variance, short-term impact coefficient, and long-term attenuation factor.

5. The method for detecting and warning abnormal temperature of a blower bearing according to claim 4, characterized in that: The parameter matrix in step 3 Among them, ω i is the long-term average variance of the i-th detection point, α i is the short-term impact coefficient of the i-th detection point; β i is the long-term attenuation factor of the i-th detection point, i = 1, 2, …, n, and n is the number of detection points.

Images