A method for predicting faults in hot rolling direct-connection roller conveyors based on motor current values.
By collecting and analyzing the current values of the direct-connection roller conveyor motors and using the quantile calculation method to predict faults, the problem of fault prediction for hot rolling direct-connection roller conveyors has been solved, enabling early warning and stable production.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI BAOSTEEL IND TECHNOLOGICAL SERVICE
- Filing Date
- 2021-08-30
- Publication Date
- 2026-05-26
AI Technical Summary
Existing technologies cannot effectively predict hot rolling direct-connection roller conveyor failures, leading to frequent unplanned downtime and affecting production efficiency and product quality.
By collecting the current value of the direct-connected roller conveyor motor, normal and abnormal data are filtered out using the quantile calculation method. The L quantile, the U quantile difference C, and the maximum D value F when there is no fault are calculated. Combined with the current value B, roller conveyor faults are predicted and early warnings are given.
It enables early warning of hot rolling direct roller conveyor failures, reduces unplanned downtime, improves production line stability and efficiency, and lowers production costs.
Smart Images
Figure CN115722540B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault diagnosis technology, and in particular to a method for predicting faults in hot rolling direct-connection roller conveyors based on motor current values. Background Technology
[0002] The cold roll conveyor on the hot rolling line is a crucial process before coiling. The conveyor is directly driven by a motor; if it jams, stalls, or trips, it can cause scratches on the strip surface. Because the cold roll conveyor operates in a harsh environment and is prone to jamming, an online motor current monitoring system is installed to provide early warnings and reduce unexpected failures.
[0003] However, in practical applications, when the threshold alarm rule is based on the motor current value, the alarm will only be triggered when an abnormality occurs (such as a slight or severe jamming) and the motor current deviates significantly, requiring the machine to be stopped as soon as possible. Instead, it cannot predict the fault.
[0004] Taking the laminar cooling roller conveyor of a hot rolling mill as an example: there are 300 roller conveyors in total, and there are scheduled downtime for inspection and maintenance. Currently, the roller conveyor inspection methods rely on manual visual inspection, listening, and using a clamp meter to test the motor current. These methods are inefficient, make it difficult to detect potential faults, and on average, a jamming fault occurs every 2 to 3 months, causing unplanned downtime. The downtime for replacement is about 0.5 to 3 hours, which seriously affects product quality, reduces production efficiency, and increases production costs. Summary of the Invention
[0005] The technical problem to be solved by the present invention is to provide a method for predicting hot rolling direct-connection roller conveyor faults based on motor current values. This method overcomes the defects of traditional roller conveyor fault detection, provides early warning of unplanned downtime caused by roller conveyor jamming, provides a basis for production line maintenance, reduces unplanned downtime, enhances production line stability, and ensures product quality.
[0006] To solve the above-mentioned technical problems, the present invention provides a method for predicting hot rolling direct-connection roller conveyor faults based on motor current values, comprising the following steps:
[0007] Step 1: Collect the current value of the direct-connected roller conveyor motor. Use the Q quantile of the historical sample current value set as the threshold, denoted as G. Change all records in the current data within the target time window and the historical sample current value set that are less than or equal to G to null values.
[0008] Step 2: Using the target evaluation time point as the endpoint, extract the L quantile of the historical sample current value set within the time window W1, denoted as A. The L quantile is higher than the proportion of records in the historical sample current value set where the current value drops due to shutdown within all fixed time windows W1.
[0009] Step 3: Using the target evaluation time point as the endpoint, extract the U quantile of the historical sample current value set within the time window W1, denoted as B. The U quantile is lower than the proportion of outlier current records within all fixed time windows W1 in the historical sample current value set, and higher than the L quantile.
[0010] Step 4: Calculate the difference between the L quantile and the U quantile, C=BA. With the target evaluation time point as the endpoint, extract a time window W2. Take D=the maximum value of C within the W2 window minus the C value at the target evaluation time point. If the minimum value of C within the W2 window is not the target evaluation time point, then set D=0.
[0011] Step 5: Using time window W1, calculate the U quantile of the current value within time window W1 corresponding to all time points in the historical sample current data of the target roller conveyor, and obtain the historical sample U quantile set. Take the M quantile of the historical sample U quantile set, denoted as E.
[0012] Step 6: Using time window W1, calculate the current value D for all time points in the historical sample dataset of the target roller conveyor within time window W1, as in Step 4. Take the maximum D value when there is no fault and record it as F.
[0013] Step 7: If the B value at the calculated time point is less than the E value and the D value is greater than the F value, then predict the hot rolling direct roller conveyor failure and give an early warning.
[0014] Furthermore, the current value of the direct-connected roller conveyor motor is collected and stored using a current transformer.
[0015] Furthermore, the time window W1 and time window W2 are both 3 days.
[0016] The present invention, based on motor current values, employs the aforementioned technical solution for predicting hot rolling direct-connection roller conveyor faults. Specifically, this method collects the motor current values of the direct-connection roller conveyor, calculates the quantiles of various parameters in a historical sample current value set, calculates and stores the E and F values based on the historical sample data, and extracts the current data B and D within the current time window W1+W2 at a certain frequency. After calculation, the B and E values, and the D and F values are compared. If the results meet the early warning conditions, a corresponding alarm is output to the intelligent alarm platform. This method overcomes the shortcomings of traditional roller conveyor fault detection, provides early warning of unplanned downtime caused by roller conveyor jamming, provides a basis for production line maintenance, reduces unplanned downtime, enhances production line stability, and ensures product quality. Attached Figure Description
[0017] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments:
[0018] Figure 1 This is a flowchart of the method for predicting hot rolling direct-connection roller conveyor faults based on motor current values according to the present invention.
[0019] Figure 2 This is a schematic diagram of the raw data collected from the current value of the direct-drive roller conveyor motor in this method;
[0020] Figure 3 This is a schematic diagram showing the preprocessed raw data in this method;
[0021] Figure 4 This is a schematic diagram illustrating the method of calculating the feature values of historical sample data at each time point using a sliding window. Detailed Implementation
[0022] Implementation, for example Figure 1 As shown, the method for predicting hot rolling direct-connection roller conveyor faults based on motor current values according to the present invention includes the following steps:
[0023] Step 1, such as Figure 2 As shown, the current values of the direct-drive roller conveyor motor are collected. The Q quantile of the historical sample current value set is used as the threshold, denoted as G. All current data within the target time window and records in the historical sample current value set less than or equal to G are converted to null values. This is to clear all current data from shutdown maintenance to avoid affecting subsequent calculations. During operation, the motor current value of the direct-drive roller conveyor motor experiences fluctuations due to processes such as start-up oil feeding, steel biting, steel throwing, and shutdown. The current value is unstable and irregular, making it difficult to distinguish processes and reflect the state when performing steel feeding work. Therefore, the collected direct-drive roller conveyor motor current values need to be preprocessed. The preprocessed motor current values are shown below. Figure 3 As shown;
[0024] In this context, Q represents a numerical value. Q quantiles can be lower quantiles such as the 10th quantile and the 15th quantile, used to completely filter out the effects of low current during shutdown. The Q quantiles may be different for roller conveyor motors with different working functions.
[0025] Step 2: Using the target evaluation time point as the endpoint, extract the L quantile of the historical sample current value set within the time window W1, denoted as A. The L quantile is higher than the proportion of records in the historical sample current value set where the current value drops due to shutdown within all fixed time windows W1.
[0026] In the L quantile, L represents a numerical value, such as the 20th quantile, 30th quantile, etc., and L is higher than Q;
[0027] Step 3: Using the target evaluation time point as the endpoint, extract the U quantile of the historical sample current value set within the time window W1, denoted as B. The U quantile is lower than the proportion of outlier current records within all fixed time windows W1 in the historical sample current value set, and higher than the L quantile.
[0028] In the U quantile, U represents a numerical value, such as the 70th quantile, 80th quantile, etc. U is higher than L and higher than Q. The U quantile is used to filter out non-fault points that occasionally jump.
[0029] Step 4: Calculate the difference between the L quantile and the U quantile, C=BA. With the target evaluation time point as the endpoint, extract a time window W2. Take D=the maximum value of C within the W2 window minus the C value at the target evaluation time point. If the minimum value of C within the W2 window is not the target evaluation time point, then set D=0. For all time points within the W2 time window, calculate A, B, and C using the W1 window.
[0030] Step 5: Using time window W1, calculate the U quantile of the current value within time window W1 corresponding to all time points in the historical sample current data of the target roller conveyor, and obtain the historical sample U quantile set. Take the M quantile of the historical sample U quantile set, denoted as E.
[0031] Step 6: Using time window W1, calculate the current value D for all time points in the historical sample dataset of the target roller conveyor within time window W1, as in Step 4. Take the maximum D value when there is no fault and record it as F.
[0032] The feature values of historical sample data at each time point are calculated using a sliding window, as follows: Figure 4 As shown;
[0033] Step 7: If the B value at the calculated time point is less than the E value and the D value is greater than the F value, then predict the hot rolling direct roller conveyor failure and give an early warning.
[0034] Preferably, the current value of the direct-connected roller conveyor motor is collected and stored using a current transformer.
[0035] Preferably, the time window W1 and the time window W2 are 3 days.
[0036] The input to this method is the motor current value of the direct-drive roller conveyor motor during operation, and the output is whether there is a risk of unplanned downtime for the roller conveyor at the calculated time point. If the output indicates a risk, the operator can choose to replace the roller conveyor during the most recent routine maintenance downtime to avoid unplanned downtime, thereby improving production line stability and increasing production capacity.
[0037] This method can provide early warning of unplanned downtime caused by roller conveyor jamming, providing a reliable basis for production line maintenance. It facilitates the replacement of roller conveyors that are about to fail during the most recent routine maintenance, reducing unplanned downtime and playing a significant role in enhancing production line stability and increasing production capacity.
[0038] This method is simple and inexpensive to acquire input data, requires low computing resources during operation, can be used for real-time monitoring, and is adaptable to different individual roller conveyors. It does not require parameter adjustment for each roller conveyor, is easy to deploy, improves product quality and production efficiency, and reduces production costs.
Claims
1. A method for predicting hot rolling direct-connection roller conveyor faults based on motor current values, characterized in that... This method includes the following steps: Step 1: Collect the current value of the direct-connected roller conveyor motor. Use the Q quantile of the historical sample current value set as the threshold, denoted as G. Change all records in the current data within the target time window and the historical sample current value set that are less than or equal to G to null values. Step 2: Using the target evaluation time point as the endpoint, extract the L quantile of the historical sample current value set within the time window W1, denoted as A. The L quantile is higher than the proportion of records in the historical sample current value set where the current value drops due to shutdown within all fixed time windows W1. Step 3: Using the target evaluation time point as the endpoint, extract the U quantile of the historical sample current value set within the time window W1, denoted as B. The U quantile is lower than the proportion of outlier current records within all fixed time windows W1 in the historical sample current value set, and higher than the L quantile. Step 4: Calculate the difference between the L quantile and the U quantile, C=BA. With the target evaluation time point as the endpoint, extract a time window W2. Take D=the maximum value of C within the W2 window minus the C value at the target evaluation time point. If the minimum value of C within the W2 window is not the target evaluation time point, then set D=0. Step 5: Using time window W1, calculate the U quantile of the current value within time window W1 corresponding to all time points in the historical sample current data of the target roller conveyor, and obtain the historical sample U quantile set. Take the M quantile of the historical sample U quantile set, denoted as E. Step 6: Using time window W1, calculate the current value D for all time points in the historical sample dataset of the target roller conveyor within time window W1, as in Step 4. Take the maximum D value when there is no fault and record it as F. Step 7: If the B value at the calculated time point is less than the E value and the D value is greater than the F value, then predict the hot rolling direct roller conveyor failure and give an early warning.
2. The method for predicting hot rolling direct-connection roller conveyor faults based on motor current values according to claim 1, characterized in that: The current value of the direct-connected roller conveyor motor is collected and stored using a current transformer.
3. The method for predicting hot rolling direct-connection roller conveyor faults based on motor current values according to claim 1 or 2, characterized in that: The time windows W1 and W2 are both 3 days.