Water gate structure fault prediction system based on deep learning and multi-feature fusion
Through deep learning and multi-feature fusion, the sluice structure fault prediction system uses acoustic wave detection and image comparison technology to accurately predict sluice faults, solving the problem of difficult prediction of gate leakage, vibration and start-stop abnormalities in existing technologies, achieving early prediction of faults and reducing the amount of detection work.
Patent Information
- Application Number
- CN202510767689.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-10
- Publication Date
- 2025-09-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing technologies make it difficult to accurately predict sluice gate failures, especially gate leakage, vibration, and abnormal start-stop, which leads to the expansion of the impact of the failure.
A sluice structure fault prediction system based on deep learning and multi-feature fusion is used. The standard and actual reflection images of the gate are obtained through acoustic wave detection, and image comparison is performed to screen and associate degraded areas. The vibration triggering probability is calculated, the possibility of gate vibration and the estimated vibration amplitude are formed, and the type and degree of fault are determined.
It achieves accurate prediction of sluice faults, reduces the workload of inspections, and enables targeted repairs before defects expand, thus avoiding the expansion of the impact of faults.
Smart Images
Figure CN120673146A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of water conservancy projects, and in particular to a sluice structure fault prediction system based on deep learning and multi-feature fusion. Background Art
[0002] A low-head hydraulic structure built on a river or canal that uses gates to control flow and regulate water levels. In water conservancy projects, sluice gates are widely used as structures for retaining, discharging, or taking in water. Sluice gates need to be repaired in the early stages of a fault to prevent the fault from becoming too serious. The initial faults of sluice gates are mainly gate leakage, gate vibration, and abnormal gate start and stop, which are mainly caused by wear of the bearings that control the gates or gaps in the connection between the gates and the bottom of the water and on both sides. The prediction of the initial faults of the gates can allow for repairs to be made in advance to avoid the impact caused by the faults. However, the existing technology does not have a sufficient understanding of the extent of the impact of defects on the gates, resulting in an unclear understanding of whether the defects of the gates are sufficient to cause a fault, making it difficult to accurately predict the fault. Summary of the Invention
[0003] In order to solve the above technical problems, a sluice structure fault prediction system based on deep learning and multi-feature fusion is provided. This technical solution solves the problems raised in the above background technology.
[0004] In order to achieve the above objects, the technical solution adopted by the present invention is:
[0005] The sluice structure fault prediction system based on deep learning and multi-feature fusion includes:
[0006] a fault classification module, wherein the fault classification module obtains at least one gate structure fault, wherein the gate structure fault includes gate leakage, gate vibration, and gate start-stop abnormality;
[0007] A triggering module, wherein the triggering module forms an early warning triggering condition for gate vibration, wherein the triggering condition is caused by gate corrosion or bearing wear;
[0008] An image acquisition module, which uses acoustic wave detection to acquire a standard reflection image of the gate when it is operating normally, and uses acoustic wave detection to acquire an actual reflection image of the gate when it is actually operating;
[0009] an image comparison module, which compares the standard reflection image with the actual reflection image to obtain at least one degraded area in the actual reflection image;
[0010] A region association module, wherein the region association module screens and associates the degraded regions to form at least one degraded region set, wherein the degraded region set is composed of a shaft degraded region set, a side degraded region set, and a bottom degraded region set;
[0011] A vibration prediction module, wherein the vibration prediction module calculates the vibration trigger probability of the set of degraded areas and forms an overall probability of gate vibration and an estimated vibration amplitude based on the vibration trigger probability;
[0012] Fault judgment module, the vibration judgment module forms a probability critical value of gate vibration based on the early warning trigger condition of gate vibration. When the overall possibility exceeds the probability critical value, the fault is predicted to be gate vibration. Otherwise, no processing is performed to form an impact model of the defect on gate leakage and gate start-stop abnormalities. When the fault is predicted to be gate vibration, the degree of gate leakage and the degree of gate start-stop abnormalities are predicted according to the impact model, and the prediction result is output. The prediction is a prediction of the situation within a preset time, and the preset time is set based on the prediction requirements.
[0013] Preferably, the triggering condition for the early warning of gate vibration comprises the following steps:
[0014] Acquire at least one historical monitoring data of the gate, and acquire the vibration condition of the gate in the historical monitoring data;
[0015] The vibration condition with a vibration amplitude equal to the allowable vibration amplitude is taken as the baseline vibration condition. In the baseline vibration condition, the defect areas on both sides of the gate and at the bearing position are counted to obtain the baseline defect area.
[0016] During the actual monitoring process, when the defect area on both sides of the gate and at the bearing position exceeds the benchmark defect area, a vibration warning is triggered.
[0017] Preferably, the comparing the standard reflection image with the actual reflection image to obtain at least one degraded region in the actual reflection image comprises the following steps:
[0018] The standard reflection image is evenly divided into at least one standard block, and in the same manner, the actual reflection image is evenly divided into at least one actual block;
[0019] Pairing the standard blocks and actual blocks with the same relative positions in the standard reflection image and the actual reflection image;
[0020] Calculate the average value of the pixel values of the pixels in the standard block to obtain the standard value, and calculate the average value of the pixel values of the pixels in the actual block to obtain the actual value;
[0021] Compare the actual value of the actual block with the standard value of the standard block to obtain a comparison value;
[0022] The actual blocks with contrast values not equal to 1 are taken as target areas, and the target areas connected by edges are aggregated to form at least one degraded area.
[0023] Preferably, the screening and associating of the degraded regions to form at least one degraded region set comprises the following steps:
[0024] The degraded areas on both sides of the gate are referred to as side degraded areas, the degraded areas at the gate bearings are referred to as shaft degraded areas, and the degraded areas at the gate bottom are referred to as bottom degraded areas.
[0025] The shaft degradation areas on the same bearing are aggregated into a shaft degradation area set;
[0026] In a historical corrosion image of a gate undergoing corrosion, at least one defective region is obtained, and a relationship between two defective regions that are fused into one region by corrosion is set as an association;
[0027] The maximum distance between the two associated defect areas is taken as the critical distance;
[0028] The side degenerate regions whose distances between each other are less than the critical distance are aggregated into a side degenerate region set;
[0029] The bottom degradation areas whose distances to each other are less than the critical distance are aggregated into a bottom degradation area set;
[0030] The shaft degenerate region set, the side degenerate region set, and the bottom degenerate region set are collectively referred to as the degenerate region set.
[0031] Preferably, the calculating and obtaining the vibration triggering probability of the degraded area set includes the following steps:
[0032] The probability of vibration triggering in the bottom degenerate area set is 0;
[0033] Obtaining the formation time of the side degradation region, dividing the total area of the side degradation region by the formation time of the side degradation region to obtain the water corrosion rate, and multiplying the water corrosion rate by the preset time to obtain the estimated corrosion area;
[0034] Obtaining the formation time of the shaft degradation region, dividing the total area of the shaft degradation region by the formation time of the shaft degradation region to obtain the wear rate, and multiplying the wear rate by the preset time to obtain the estimated wear area;
[0035] The total area of the side degradation area in the side degradation area set is superimposed with the estimated corrosion area to obtain the side defect area;
[0036] The total area of the shaft degradation area in the shaft degradation area set is superimposed on the estimated wear area to obtain the shaft defect area;
[0037] Comparing the side defect area with the reference defect area, the vibration triggering probability of the side degradation area set is obtained;
[0038] The shaft defect area is compared with the reference defect area to obtain the vibration triggering probability of the shaft degradation area set.
[0039] Preferably, forming the overall possibility of gate vibration and estimating the vibration amplitude based on the vibration trigger probability includes the following steps:
[0040] Using the trigger probability formula, the overall probability of gate vibration is calculated;
[0041] Accumulating the vibration trigger probabilities of at least one side degradation area set and the shaft degradation area set to obtain a target multiple, and multiplying the target multiple by the allowable vibration amplitude to obtain an estimated vibration amplitude;
[0042] The trigger probability formula is as follows:
[0043]
[0044] Where P is the overall probability, i and j are subscripts, n is the number of side degenerate regions, a i is the vibration triggering probability of the i-th side degradation area set, m is the number of shaft degradation area sets, b j is the vibration triggering probability of the jth shaft degradation area set.
[0045] Preferably, the early warning triggering condition based on gate vibration to form a critical probability value of gate vibration includes the following steps:
[0046] Superimposing at least one side defect area and at least one shaft defect area to obtain an overall defect area;
[0047] The baseline defect area is divided by the total defect area to obtain the probability critical value.
[0048] Preferably, the formation of the defect impact model on gate leakage and gate start-stop abnormality includes the following steps:
[0049] Obtaining a first area value range of the defective regions on both sides and the bottom of the gate, dividing the first area value range into equal intervals to obtain at least one first identification point;
[0050] Under the condition that the area of the defective areas on both sides and the bottom of the gate is equal to the first identification point, the water leakage of the gate is obtained;
[0051] Pairing and fitting the first identified point with the water leakage amount to obtain a water leakage fitting function;
[0052] Obtaining a second area value range of the defective area at the gate bearing position, dividing the second area value range into equal intervals to obtain at least one second identification point;
[0053] Under the condition that the area of the defective region at the gate bearing position is equal to the second identification point, the start and stop delay time of the gate is obtained;
[0054] Pairing and fitting the second identification point with the start-stop delay time to obtain a start-stop fitting function;
[0055] The water leakage fitting function and the start-stop fitting function are used as the impact models.
[0056] Preferably, the step of predicting the extent of gate leakage based on the impact model comprises the following steps:
[0057] Superimpose the defect area of at least one side surface to obtain the total area of the side surface;
[0058] Compare the gate bottom length with twice the gate height to obtain the conversion coefficient, and multiply the total side area by the conversion coefficient to obtain the total bottom area;
[0059] Substitute the sum of the bottom total area and the side total area into the leakage fitting function to obtain the leakage prediction value.
[0060] Preferably, the predicting of the degree of gate start-stop abnormality includes the following steps:
[0061] The areas of at least one shaft defect are superimposed to obtain the total shaft area, which is then substituted into the start-stop fitting function to obtain the start-stop prediction delay.
[0062] Compared with the prior art, the present invention has the following beneficial effects:
[0063] By setting up a trigger formation module, a regional association module, a vibration prediction module and a fault judgment module, starting from the prediction of the vibration situation, first determine whether the vibration situation of the gate within the future preset time is acceptable based on the impact of the defect situation on the vibration, because the vibration situation is positively correlated with the gate defect, and thus decide whether to predict the start and stop and water leakage of the gate based on the prediction results of the vibration situation. Therefore, when there is no abnormality, there is no need to analyze all the conditions of the gate, which can reduce the amount of analysis work. At the same time, the impact of the defect on the gate failure is analyzed, and the expansion of the defect within the future preset time is predicted, so that the possible fault degree of the gate structure failure can be comprehensively predicted. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 This is a flow chart of the sluice structure fault prediction system based on deep learning and multi-feature fusion of the present invention;
[0065] Figure 2 A flow chart of the triggering conditions for the early warning of gate vibration according to the present invention;
[0066] Figure 3A schematic diagram of a process of comparing a standard reflection image with an actual reflection image to obtain at least one degraded region in the actual reflection image according to the present invention;
[0067] Figure 4 A schematic diagram of a process for screening and associating degraded regions to form at least one degraded region set according to the present invention;
[0068] Figure 5 A schematic diagram of a flow chart of calculating the vibration triggering probability of a set of degraded regions according to the present invention;
[0069] Figure 6 A schematic diagram of a process for forming the overall possibility of gate vibration and estimating the vibration amplitude based on the vibration trigger probability of the present invention;
[0070] Figure 7 This is a flow chart of forming a critical value of gate vibration probability based on the early warning triggering condition of the gate vibration of the present invention;
[0071] Figure 8 A flow chart of a model for the influence of defects on gate leakage and gate start-stop abnormalities according to the present invention;
[0072] Figure 9 The figure is a flow chart of predicting the degree of gate leakage based on the impact model of the present invention. DETAILED DESCRIPTION
[0073] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.
[0074] Reference Figure 1 As shown in the figure, the sluice structure fault prediction system based on deep learning and multi-feature fusion includes:
[0075] a fault classification module, wherein the fault classification module obtains at least one gate structure fault, wherein the gate structure fault includes gate leakage, gate vibration, and gate start-stop abnormality;
[0076] A triggering module, wherein the triggering module forms an early warning triggering condition for gate vibration, wherein the triggering condition is caused by gate corrosion or bearing wear;
[0077] An image acquisition module, which uses acoustic wave detection to acquire a standard reflection image of the gate when it is operating normally, and uses acoustic wave detection to acquire an actual reflection image of the gate when it is actually operating;
[0078] an image comparison module, which compares the standard reflection image with the actual reflection image to obtain at least one degraded area in the actual reflection image;
[0079] A region association module, wherein the region association module screens and associates the degraded regions to form at least one degraded region set, wherein the degraded region set is composed of a shaft degraded region set, a side degraded region set, and a bottom degraded region set;
[0080] A vibration prediction module, wherein the vibration prediction module calculates the vibration trigger probability of the set of degraded areas and forms an overall probability of gate vibration and an estimated vibration amplitude based on the vibration trigger probability;
[0081] Fault judgment module, the vibration judgment module forms a probability critical value of gate vibration based on the early warning trigger condition of gate vibration. When the overall possibility exceeds the probability critical value, the fault is predicted to be gate vibration. Otherwise, no processing is performed to form an impact model of the defect on gate leakage and gate start-stop abnormalities. When the fault is predicted to be gate vibration, the degree of gate leakage and the degree of gate start-stop abnormalities are predicted according to the impact model, and the prediction result is output. The prediction is a prediction of the situation within a preset time, and the preset time is set based on the prediction requirements.
[0082] In this solution, the main idea is to use gate vibration as the starting point for prediction. Gate vibration is mainly caused by defects on both sides of the gate and wear of the bearings that control the gate. Due to the defects, the fit is not tight, which will cause vibration. These defects are also related to gate leakage and abnormal gate start and stop. Because when the defects on both sides of the gate are large, the corrosion rate of water is the same for the bottom and both sides of the gate. As long as the length of the bottom and both sides is different, the defect situation of the bottom can be obtained. Therefore, when analyzing, it is only necessary to predict the defects on both sides of the gate within the preset time. When the impact of the defects on various faults is determined, targeted predictions can be made. The initial failures of the gate are mainly water leakage, gate vibration and abnormal gate start-stop. Among them, the cause of gate vibration is that when the gate vibrates beyond the allowable range for a long time, the vibration will cause the gate defect to expand faster, thereby further aggravating the defect. Water leakage and gate start-stop abnormalities are both caused by defects, which are connection gaps and bearing wear respectively. Because defects are related to vibration, it is possible to determine whether the defect is large through vibration. When the defect is very small, the vibration is very small. It can be known that the degree of water leakage and gate start-stop abnormalities caused by it is very small. Therefore, no additional detection is required. In this way, the amount of inspection work during normal use of the gate can be reduced to a certain extent.
[0083] Reference Figure 2 As shown, the triggering conditions for the early warning of gate vibration include the following steps:
[0084] Acquire at least one historical monitoring data of the gate, and acquire the vibration condition of the gate in the historical monitoring data;
[0085] The vibration condition with a vibration amplitude equal to the allowable vibration amplitude is taken as the baseline vibration condition. In the baseline vibration condition, the defect areas on both sides of the gate and at the bearing position are counted to obtain the baseline defect area.
[0086] During the actual monitoring process, when the defect area on both sides of the gate and at the bearing position exceeds the benchmark defect area, a vibration warning is triggered.
[0087] The reference defect area is the defect corresponding to the allowable amplitude of vibration. It is easy to know that when there are more defect locations, there are more gaps in the connection. Therefore, the fit is not tight, which will lead to greater vibration during movement. Therefore, the defect area can be used to judge the vibration situation, and the reference defect area can be used as a benchmark.
[0088] Reference Figure 3 As shown, comparing the standard reflection image with the actual reflection image to obtain at least one degraded area in the actual reflection image includes the following steps:
[0089] The standard reflection image is evenly divided into at least one standard block, and in the same manner, the actual reflection image is evenly divided into at least one actual block;
[0090] Pairing the standard blocks and actual blocks with the same relative positions in the standard reflection image and the actual reflection image;
[0091] Calculate the average value of the pixel values of the pixels in the standard block to obtain the standard value, and calculate the average value of the pixel values of the pixels in the actual block to obtain the actual value;
[0092] Compare the actual value of the actual block with the standard value of the standard block to obtain a comparison value;
[0093] The actual blocks with contrast values not equal to 1 are taken as target areas, and the target areas connected by edges are aggregated to form at least one degraded area.
[0094] Under different corrosion conditions and wear, the sound wave development images of the sluice are different. Therefore, the location of defects can be analyzed based on this.
[0095] Reference Figure 4 As shown, screening and associating the degraded regions to form at least one degraded region set includes the following steps:
[0096] The degraded areas on both sides of the gate are referred to as side degraded areas, the degraded areas at the gate bearings are referred to as shaft degraded areas, and the degraded areas at the gate bottom are referred to as bottom degraded areas.
[0097] The shaft degradation areas on the same bearing are aggregated into a shaft degradation area set;
[0098] In a historical corrosion image of a gate undergoing corrosion, at least one defective region is obtained, and a relationship between two defective regions that are fused into one region by corrosion is set as an association;
[0099] The maximum distance between the two associated defect areas is taken as the critical distance;
[0100] The side degenerate regions whose distances between each other are less than the critical distance are aggregated into a side degenerate region set;
[0101] The bottom degradation areas whose distances to each other are less than the critical distance are aggregated into a bottom degradation area set;
[0102] The shaft degenerate region set, the side degenerate region set, and the bottom degenerate region set are collectively referred to as the degenerate region set.
[0103] Assuming that the area of the defect is the baseline defect area, but it is evenly dispersed at different locations of the sluice gate, and the area of each location is very small, it will not have an impact. Therefore, when conducting analysis, the defect area cannot be used for single analysis. What needs to be considered is the impact of each continuous defect area. Since the locations of defects are dispersed, there may or may not be a superposition effect between them, and the defects may expand over time. Therefore, it is necessary to classify them into sets, and obtain the predicted defect results through comprehensive analysis of the areas within the set.
[0104] Reference Figure 5 As shown, calculating the vibration triggering probability of the degraded area set includes the following steps:
[0105] The probability of vibration triggering in the bottom degenerate area set is 0;
[0106] Obtaining the formation time of the side degradation region, dividing the total area of the side degradation region by the formation time of the side degradation region to obtain the water corrosion rate, and multiplying the water corrosion rate by the preset time to obtain the estimated corrosion area;
[0107] Obtaining the formation time of the shaft degradation region, dividing the total area of the shaft degradation region by the formation time of the shaft degradation region to obtain the wear rate, and multiplying the wear rate by the preset time to obtain the estimated wear area;
[0108] The total area of the side degradation area in the side degradation area set is superimposed with the estimated corrosion area to obtain the side defect area;
[0109] The total area of the shaft degradation area in the shaft degradation area set is superimposed on the estimated wear area to obtain the shaft defect area;
[0110] Comparing the side defect area with the reference defect area, the vibration triggering probability of the side degradation area set is obtained;
[0111] The shaft defect area is compared with the reference defect area to obtain the vibration triggering probability of the shaft degradation area set.
[0112] The bottom degraded area set only affects water leakage. Each side degraded area set or shaft degraded area set has the possibility of generating vibration, even if its area does not exceed the reference defect area. Because there are multiple side degraded area sets or shaft degraded area sets that have a coordinated impact, it is necessary to obtain the corresponding vibration trigger probability based on their area.
[0113] Reference Figure 6 As shown, based on the vibration trigger probability, the overall possibility of gate vibration and the estimated vibration amplitude include the following steps:
[0114] Using the trigger probability formula, the overall probability of gate vibration is calculated;
[0115] Accumulating the vibration trigger probabilities of at least one side degradation area set and the shaft degradation area set to obtain a target multiple, and multiplying the target multiple by the allowable vibration amplitude to obtain an estimated vibration amplitude;
[0116] The trigger probability formula is as follows:
[0117]
[0118] Where P is the overall probability, i and j are subscripts, n is the number of side degenerate regions, a i is the vibration triggering probability of the i-th side degradation area set, m is the number of shaft degradation area sets, b j is the vibration triggering probability of the jth shaft degradation area set.
[0119] According to the calculation process of the vibration trigger probability, the cumulative vibration trigger probability of the side degraded area set and the shaft degraded area set is actually the ratio of the sum of their corresponding defect areas to the reference defect area. Therefore, the overall vibration amplitude is a corresponding multiple of the allowable vibration amplitude.
[0120] Reference Figure 7 As shown, based on the early warning triggering conditions of gate vibration, forming the probability critical value of gate vibration includes the following steps:
[0121] Superimposing at least one side defect area and at least one shaft defect area to obtain an overall defect area;
[0122] The baseline defect area is divided by the total defect area to obtain the probability critical value.
[0123] When the overall defect area is larger, the probability critical value is smaller, that is, the overall defect area is more likely to trigger a vibration warning, which is consistent with the usual practice. Therefore, it is reasonable to use the above-mentioned probability critical value to make a judgment on fault prediction.
[0124] Reference Figure 8 As shown in Figure 1, the model for the impact of defects on gate leakage and gate start-stop abnormalities includes the following steps:
[0125] Obtaining a first area value range of the defective regions on both sides and the bottom of the gate, dividing the first area value range into equal intervals to obtain at least one first identification point;
[0126] Under the condition that the area of the defective areas on both sides and the bottom of the gate is equal to the first identification point, the water leakage of the gate is obtained;
[0127] Pairing and fitting the first identified point with the water leakage amount to obtain a water leakage fitting function;
[0128] Obtaining a second area value range of the defective area at the gate bearing position, dividing the second area value range into equal intervals to obtain at least one second identification point;
[0129] Under the condition that the area of the defective region at the gate bearing position is equal to the second identification point, the start and stop delay time of the gate is obtained;
[0130] Pairing and fitting the second identification point with the start-stop delay time to obtain a start-stop fitting function;
[0131] The water leakage fitting function and the start-stop fitting function are used as the impact models.
[0132] When predicting the extent of water leakage and abnormal start-stop, it is necessary to analyze the impact of the defect area on water leakage and abnormal start-stop. However, the corresponding defect area needs to be determined. Not all defect areas on the gate will affect water leakage and abnormal start-stop, so they need to be treated differently.
[0133] Among them, under the condition that the area of the defective area on both sides and the bottom of the gate is equal to the first identification point, the water leakage of the gate is obtained. Here, when obtaining the water leakage of the gate, the gate needs to be closed, mainly by evenly mixing the fluorescent agent into the water on the upstream side of the gate, obtaining the density of the mixed fluorescent agent, and obtaining the amount of fluorescent agent passing through the gate on the downstream side of the gate. Thus, the water leakage of the gate is calculated.
[0134] Reference Figure 9 As shown in the figure, based on the impact model, the prediction of the gate leakage degree includes the following steps:
[0135] Superimpose the defect area of at least one side surface to obtain the total area of the side surface;
[0136] Compare the gate bottom length with twice the gate height to obtain the conversion coefficient, and multiply the total side area by the conversion coefficient to obtain the total bottom area;
[0137] Substitute the sum of the bottom total area and the side total area into the leakage fitting function to obtain the leakage prediction value.
[0138] The defects on both sides of the gate exist on both sides of the gate, so the ratio of the defect area at the bottom and on both sides is the ratio of the gate bottom length to twice the gate height. Since the total area of the side is known according to the analysis, the total area of the bottom can be inferred, and then the predicted amount of water leakage can be obtained.
[0139] Predicting the degree of gate start and stop abnormality includes the following steps:
[0140] The areas of at least one shaft defect are superimposed to obtain the total shaft area, which is then substituted into the start-stop fitting function to obtain the start-stop prediction delay.
[0141] The starting and stopping are mainly caused by the defective area of the shaft. The defects on both sides of the gate have little effect on the starting and stopping and can be ignored.
[0142] Furthermore, the present solution also proposes a storage medium on which a computer-readable program is stored. When the computer-readable program is called, it runs the above-mentioned sluice structure fault prediction system based on deep learning and multi-feature fusion.
[0143] It is understandable that the storage medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid state disk (SSD).
[0144] To sum up, the advantages of the present invention are: by setting a trigger formation module, a regional association module, a vibration prediction module and a fault judgment module, starting from the prediction of the vibration situation, first according to the influence of the defect situation on the vibration, determine whether the vibration situation of the gate within the future preset time is acceptable, because the vibration situation is positively correlated with the gate defect, and thus according to the prediction result of the vibration situation, decide whether to predict the start and stop and leakage of the gate, so that when there is no abnormality, there is no need to analyze all the conditions of the gate, which can reduce the amount of analysis work. At the same time, the influence of the defect on the gate failure is analyzed, and the expansion of the defect within the future preset time is predicted, so that the possible fault degree of the gate structure failure can be comprehensively predicted.
[0145] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.
Claims
1. A sluice structure fault prediction system based on deep learning and multi-feature fusion, characterized by: include: a fault classification module, wherein the fault classification module obtains at least one gate structure fault, wherein the gate structure fault includes gate leakage, gate vibration, and gate start-stop abnormality; A triggering module, wherein the triggering module forms an early warning triggering condition for gate vibration, wherein the triggering condition is caused by gate corrosion or bearing wear; An image acquisition module, which uses acoustic wave detection to acquire a standard reflection image of the gate when it is operating normally, and uses acoustic wave detection to acquire an actual reflection image of the gate when it is actually operating; an image comparison module, which compares the standard reflection image with the actual reflection image to obtain at least one degraded area in the actual reflection image; A region association module, wherein the region association module screens and associates the degraded regions to form at least one degraded region set, wherein the degraded region set is composed of a shaft degraded region set, a side degraded region set, and a bottom degraded region set; A vibration prediction module, wherein the vibration prediction module calculates the vibration trigger probability of the set of degraded areas and forms an overall probability of gate vibration and an estimated vibration amplitude based on the vibration trigger probability; Fault judgment module, the vibration judgment module forms a probability critical value of gate vibration based on the early warning trigger condition of gate vibration. When the overall possibility exceeds the probability critical value, the fault is predicted to be gate vibration. Otherwise, no processing is performed to form an impact model of the defect on gate leakage and gate start-stop abnormalities. When the fault is predicted to be gate vibration, the degree of gate leakage and the degree of gate start-stop abnormalities are predicted according to the impact model, and the prediction result is output. The prediction is a prediction of the situation within a preset time, and the preset time is set based on the prediction requirements.
2. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 1 is characterized in that: The triggering condition for the gate vibration warning comprises the following steps: Acquire at least one historical monitoring data of the gate, and acquire the vibration condition of the gate in the historical monitoring data; The vibration condition with a vibration amplitude equal to the allowable vibration amplitude is taken as the baseline vibration condition. In the baseline vibration condition, the defect areas on both sides of the gate and at the bearing position are counted to obtain the baseline defect area. During the actual monitoring process, when the defect area on both sides of the gate and at the bearing position exceeds the benchmark defect area, a vibration warning is triggered.
3. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 2 is characterized in that: The comparing the standard reflection image with the actual reflection image to obtain at least one degradation region in the actual reflection image comprises the following steps: The standard reflection image is evenly divided into at least one standard block, and in the same manner, the actual reflection image is evenly divided into at least one actual block; Pairing the standard blocks and actual blocks with the same relative positions in the standard reflection image and the actual reflection image; Calculate the average value of the pixel values of the pixels in the standard block to obtain the standard value, and calculate the average value of the pixel values of the pixels in the actual block to obtain the actual value; Compare the actual value of the actual block with the standard value of the standard block to obtain a comparison value; The actual blocks with contrast values not equal to 1 are taken as target areas, and the target areas connected by edges are aggregated to form at least one degraded area.
4. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 3 is characterized in that: The screening and associating of the degraded regions to form at least one degraded region set comprises the following steps: The degraded areas on both sides of the gate are referred to as side degraded areas, the degraded areas at the gate bearings are referred to as shaft degraded areas, and the degraded areas at the gate bottom are referred to as bottom degraded areas. The shaft degradation areas on the same bearing are aggregated into a shaft degradation area set; In a historical corrosion image of a gate undergoing corrosion, at least one defective region is obtained, and a relationship between two defective regions that are fused into one region by corrosion is set as an association; The maximum distance between the two associated defect areas is taken as the critical distance; The side degenerate regions whose distances between each other are less than the critical distance are aggregated into a side degenerate region set; The bottom degradation areas whose distances to each other are less than the critical distance are aggregated into a bottom degradation area set; The shaft degenerate region set, the side degenerate region set, and the bottom degenerate region set are collectively referred to as the degenerate region set.
5. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 4 is characterized in that: The calculation of the vibration triggering probability of the degraded area set includes the following steps: The probability of vibration triggering in the bottom degenerate area set is 0; Obtaining the formation time of the side degradation region, dividing the total area of the side degradation region by the formation time of the side degradation region to obtain the water corrosion rate, and multiplying the water corrosion rate by the preset time to obtain the estimated corrosion area; Obtaining the formation time of the shaft degradation region, dividing the total area of the shaft degradation region by the formation time of the shaft degradation region to obtain the wear rate, and multiplying the wear rate by the preset time to obtain the estimated wear area; The total area of the side degradation area in the side degradation area set is superimposed with the estimated corrosion area to obtain the side defect area; The total area of the shaft degradation area in the shaft degradation area set is superimposed on the estimated wear area to obtain the shaft defect area; Comparing the side defect area with the reference defect area, the vibration triggering probability of the side degradation area set is obtained; The shaft defect area is compared with the reference defect area to obtain the vibration triggering probability of the shaft degradation area set.
6. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 5 is characterized in that: The overall possibility of gate vibration and estimated vibration amplitude based on vibration trigger probability include the following steps: Using the trigger probability formula, the overall probability of gate vibration is calculated; Accumulating the vibration trigger probabilities of at least one side degenerate area set and the shaft degenerate area set to obtain a target multiple, and multiplying the target multiple by the allowable vibration amplitude to obtain an estimated vibration amplitude; The trigger probability formula is as follows: Where P is the overall probability, i and j are subscripts, n is the number of side degenerate regions, a i is the vibration triggering probability of the i-th side degradation area set, m is the number of shaft degradation area sets, b j is the vibration triggering probability of the jth shaft degradation area set.
7. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 6 is characterized in that: The gate vibration-based early warning triggering condition to form a gate vibration probability critical value includes the following steps: Superimposing at least one side defect area and at least one shaft defect area to obtain an overall defect area; The baseline defect area is divided by the total defect area to obtain the probability critical value.
8. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 7 is characterized in that: The model for the influence of defects on gate leakage and gate start-stop abnormalities includes the following steps: Obtaining a first area value range of the defective regions on both sides and the bottom of the gate, dividing the first area value range into equal intervals to obtain at least one first identification point; Under the condition that the area of the defective areas on both sides and the bottom of the gate is equal to the first identification point, the water leakage of the gate is obtained; Pairing and fitting the first identified point with the water leakage amount to obtain a water leakage fitting function; Obtaining a second area value range of the defective area at the gate bearing position, dividing the second area value range into equal intervals to obtain at least one second identification point; Under the condition that the area of the defective region at the gate bearing position is equal to the second identification point, the start and stop delay time of the gate is obtained; Pairing and fitting the second identification point with the start-stop delay time to obtain a start-stop fitting function; The water leakage fitting function and the start-stop fitting function are used as the impact models.
9. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 8 is characterized in that: The method of predicting the extent of gate leakage based on the impact model includes the following steps: Superimpose the defect area of at least one side surface to obtain the total area of the side surface; Compare the gate bottom length with twice the gate height to obtain the conversion coefficient, and multiply the total side area by the conversion coefficient to obtain the total bottom area; Substitute the sum of the bottom total area and the side total area into the leakage fitting function to obtain the leakage prediction value.
10. The sluice structure fault prediction system based on deep learning and multi-feature fusion according to claim 9 is characterized in that: The prediction of the degree of gate start and stop abnormality includes the following steps: The areas of at least one shaft defect are superimposed to obtain the total shaft area, which is then substituted into the start-stop fitting function to obtain the start-stop prediction delay.