A metal roof state judgment method and system based on decision tree algorithm

Abstract

The invention discloses a method and system for judging the state of a metal roof based on a decision tree algorithm, wherein the method comprises: obtaining the stress value, longitudinal displacement value and lateral slip value of a metal roof panel to be monitored through a first heterogeneous sensor and a second heterogeneous sensor; respectively obtaining the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value by using a time domain analysis method; calculating the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor; inputting the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient into a trained decision tree model, and outputting the state results of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor; through the method, the deterioration trend of the metal roof during service can be effectively grasped, and an early warning can be issued when a fault sign appears in a certain area of ​​the roof, so as to facilitate the management personnel to carry out maintenance and timely remedy according to the situation, and avoid the occurrence of serious accidents.

Description

Technical Field

[0001] The present invention belongs to the technical field of metal roof health management, and in particular relates to a metal roof state judgment method and system based on a decision tree algorithm. Background Art

[0002] The metal roof enclosure system is an important part of the large-span steel structure building. It uses cold-formed thin-walled alloy materials and integrates insulation, sound absorption, waterproofing, windproofing and other functions. In recent years, it has been widely used in various venues, airport terminals, exhibition centers, workshops, warehouses and other large public buildings and industrial buildings. The metal roof of large public buildings needs to take into account both functionality and aesthetics. The metal roof enclosure system has been rapidly developed in public buildings due to its excellent structural performance, beautiful appearance, and green environmental protection.

[0003] Metal roofs are subject to various weather loads in the natural environment, among which wind loads and temperature loads are the most critical, which may cause deformation, cracks, wind-lifting and other accidents. In the early days, the maintenance of metal roofs generally adopted the method of regular manual inspection, which consumed a lot of manpower and time and failed to understand the situation in a timely manner.

[0004] Therefore, based on the above factors, how to grasp the deterioration trend of metal roofs during their service and issue an early warning when there are signs of failure in a certain area of ​​the roof, so as to facilitate management personnel to carry out repairs and timely remedies as appropriate to avoid serious accidents, has become an urgent problem to be solved. Summary of the invention

[0005] The main purpose of the present invention is to provide a method and system for judging the status of a metal roof based on a decision tree algorithm, which can effectively grasp the deterioration trend of the metal roof during its service and issue an early warning when a sign of failure appears in a certain area of ​​the roof, so as to facilitate management personnel to carry out maintenance and timely remedies according to the situation and avoid the occurrence of serious accidents.

[0006] An embodiment of the present invention provides a method for determining the state of a metal roof based on a decision tree algorithm, comprising:

[0007] S1. Obtaining the stress value, longitudinal displacement value and lateral slip value of the metal roof panel to be monitored through the first heterogeneous sensor and the second heterogeneous sensor;

[0008] S2. Preprocessing the stress value, the longitudinal displacement value and the lateral slip value by using a time domain analysis method to obtain characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value respectively; and calculating the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor at the same time;

[0009] S3, inputting the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor into the trained decision tree model, and outputting the status results of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor;

[0010] The decision tree model is trained by taking stress value, longitudinal displacement value, lateral slip value, and correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor as input, and taking corresponding metal roof panel and status result of the first heterogeneous sensor or the second heterogeneous sensor as output.

[0011] Furthermore, it also includes:

[0012] S4. If, in the status result, the metal roof panel, the first heterogeneous sensor and / or the second heterogeneous sensor are faulty, a corresponding warning is issued through an alarm device.

[0013] Furthermore, the S1 specifically includes:

[0014] S11, dividing the metal roof panel into a plurality of collection areas; each of the collection areas contains a preset number of collection nodes;

[0015] S12. A first heterogeneous sensor and a second heterogeneous sensor are installed at each of the acquisition nodes; the stress value, longitudinal displacement value and lateral slip value of the metal roof panel corresponding to each of the acquisition nodes are obtained through the first heterogeneous sensor and the second heterogeneous sensor.

[0016] Further, the first heterogeneous sensor includes a strain gauge and a laser sensor, which are respectively used to obtain the stress value and the longitudinal displacement value of the metal roof panel corresponding to the acquisition node;

[0017] The second heterogeneous sensor includes an ultrasonic sensor, which is used to obtain a lateral slip value of the metal roof panel corresponding to the collection node.

[0018] Furthermore, the S2 specifically includes:

[0019] Taking a preset number of collection cycles corresponding to each of the collection nodes as a time window;

[0020] Calculating the stress value, longitudinal displacement value and lateral slip value corresponding to each acquisition node in each time window, and obtaining characteristic quantities corresponding to the stress value, longitudinal displacement value and lateral slip value of each acquisition node in each time window;

[0021] At the same time, a correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor corresponding to each of the acquisition nodes is calculated.

[0022] Furthermore, the characteristic quantities of the stress value include the maximum stress value, the standard deviation of stress and the mean stress value;

[0023] The characteristic quantity of the longitudinal displacement value includes the longitudinal displacement mean;

[0024] The characteristic quantity of the lateral slip value includes a lateral slip mean value.

[0025] Further, for each of the collection nodes, the corresponding S3 specifically includes:

[0026] S31, inputting the longitudinal displacement mean, lateral slip mean, stress maximum, stress standard deviation and stress mean within the current time window, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor into the trained decision tree model;

[0027] S32, judging the ratio of the longitudinal displacement mean or the lateral slip mean by the decision tree model;

[0028] If the ratio of the longitudinal displacement mean or the lateral slip mean is greater than the first threshold, it indicates that the metal roof panel at the acquisition node has a fault, and the fault has the possibility of recovery;

[0029] If the ratio of the longitudinal displacement mean or the lateral slip mean is less than the first threshold, proceed to step S33;

[0030] S33, judging the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor by the decision tree model;

[0031] If the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is less than a second threshold, it indicates that the first heterogeneous sensor or the second heterogeneous sensor corresponding to the acquisition node is faulty;

[0032] If the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is greater than a second threshold, proceed to step S34;

[0033] S34, judging the maximum stress value, the stress standard deviation and the stress mean value by the decision tree model;

[0034] If the maximum stress value is greater than the third threshold, or the standard deviation of the stress is greater than the fourth threshold, or the mean stress value is greater than the fifth threshold, it indicates that the metal roof panel corresponding to the acquisition node has an irreversible fault;

[0035] If the maximum stress value is less than the third threshold, the standard deviation of the stress is less than the fourth threshold, and the mean stress value is less than the fifth threshold, it indicates that there is no fault in the metal roof panel corresponding to the acquisition node.

[0036] On the other hand, an embodiment of the present invention further provides a metal roof state judgment system based on a decision tree algorithm, applying the above method, the system comprises: a collection module, a preprocessing module and a judgment module;

[0037] The acquisition module is used to obtain the stress value, longitudinal displacement value and lateral slip value of the metal roof panel to be monitored through the first heterogeneous sensor and the second heterogeneous sensor;

[0038] The preprocessing module is used to preprocess the stress value, the longitudinal displacement value and the lateral slip value by using a time domain analysis method to obtain characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value respectively; and calculate the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor at the same time;

[0039] The judgment module is used to judge the status of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor according to the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor.

[0040] Furthermore, it also includes an early warning module;

[0041] The early warning module is used to issue a corresponding early warning when a fault occurs in the metal roof panel, the first heterogeneous sensor and / or the second heterogeneous sensor in the status result.

[0042] Compared with the prior art, the present invention has the following beneficial effects: Through this method, the deterioration trend of the metal roof during service can be effectively grasped, and an early warning can be issued when signs of failure appear in a certain area of ​​the roof, so as to facilitate management personnel to carry out maintenance and timely remedies according to the situation and avoid the occurrence of serious accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of the basic structure of a decision tree provided by an embodiment of the present invention.

[0044] Figure 2 A decision tree training flow chart provided for an embodiment of the present invention.

[0045] Figure 3 A flow chart of a method for determining the state of a metal roof based on a decision tree algorithm provided in an embodiment of the present invention.

[0046] Figure 4 A decision tree judgment flow chart provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0048] In the description of the present invention, it should be noted that the terms "upper", "lower", "inner", "outer", "front end", "rear end", "two ends", "one end", "the other end" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0049] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "provided with", "connected", etc. should be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0050] First, the existing related technologies are briefly described.

[0051] The decision tree algorithm is a supervised learning algorithm, whose goal is to find the mapping relationship between the feature vector describing the element and the decision category, and to use this mapping relationship to distinguish unknown category samples. This patent intends to use the decision tree algorithm to extract multiple features to distinguish the fault type and realize real-time fault warning.

[0052] Usually, applying the decision tree algorithm for classification can be summarized into two steps:

[0053] (1) A decision tree is trained based on a dataset with known feature vectors and categories. The input is a dataset D = {attribute A, attribute B, decision} containing feature quantities and decision categories. The output is the generated decision tree, i.e., a tree structure with the root at the top and the leaves at the bottom, as shown in the figure below. Figure 1 As shown in the figure, attribute A and attribute B are feature vectors, which can be located at the root node and branch node of the tree; a1, a2, a3, b1, b2 are the conditional criteria for root generation during the training process, and decision 1 and decision 2 are classification categories, which are located at the leaf nodes.

[0054] (2) Use the generated decision tree to predict and classify the feature vector of unknown category. The input is the data set to be classified D = {attribute A, attribute B}. Figure 1 The decision tree performs classification and the output is the decision category.

[0055] The decision tree algorithm training process is as follows:

[0056] (1) Data preparation: Preprocess the original data, select feature quantities, and collect a training set D = {feature vector set X, decision category} containing feature vectors and categories as input.

[0057] (2) Generate a decision tree. Recursively call the decision tree generation function TreeGenerate(D,X), where X is the feature vector set and D is the training sample set, including the original data set and the data subset. After the training is completed, a decision tree is returned. For detailed training process, see the attached Figure 2 , specifically described as follows:

[0058] Create an empty node N;

[0059] If the samples in D belong to the same category C, then mark node N as a leaf node of category C;

[0060] If the attribute set X is empty or the samples in D have the same value on the attribute set X, then the node N is marked as a leaf node, and its category is the class with the largest number of samples in D;

[0061] Otherwise, select the optimal partition attribute A from X, mark node N with A, generate branches for each value of A, and let Dv be the sample subset of each value of A; if Dv is empty, mark the branch of N as a leaf node, and the marking type is the class with the most samples in D; if Dv is not empty, call the recursive function TreeGenerate(Dv,X);

[0062] (3) Verify the classification accuracy. According to the decision tree generated in (2), the test set D = {feature vector set X} is used as input, and the output is the discriminant type. Compare the correct rate of the output discriminant type of the decision tree to verify the classification accuracy of the decision tree.

[0063] As attached Figure 3 As shown, an embodiment of the present invention provides a method for determining the state of a metal roof based on a decision tree algorithm, comprising:

[0064] S1. Obtaining the stress value, longitudinal displacement value and lateral slip value of the metal roof panel to be monitored through the first heterogeneous sensor and the second heterogeneous sensor;

[0065] S2. Preprocessing the stress value, the longitudinal displacement value and the lateral slip value by using a time domain analysis method to obtain characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value respectively; and calculating the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor at the same time;

[0066] S3, inputting the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor into the trained decision tree model, and outputting the status results of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor;

[0067] The decision tree model is trained by taking the stress value, the longitudinal displacement value, the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor as inputs, and taking the corresponding metal roof panel and the state result of the first heterogeneous sensor or the second heterogeneous sensor as outputs;

[0068] S4. If the status result shows that the metal roof panel, the first heterogeneous sensor and / or the second heterogeneous sensor are faulty, a corresponding warning is issued through the alarm device.

[0069] Next, the above steps are described in detail.

[0070] In the above step S1, the most basic work for the health management of metal roofs is to determine the current state of the roof based on the sensor data collected in real time. Since the collected structural monitoring signals are easily affected by factors such as noise, electromagnetics, network transmission, and sensor failure, it is easy to cause misjudgment if the threshold is judged based on a single sensor. Therefore, it is necessary to fuse the heterogeneous sensor signals at the same location to eliminate the interference of external factors on the judgment;

[0071] In the embodiment of the present invention, firstly, the metal roof panel is divided into a plurality of collection areas; each collection area contains a preset number of collection nodes; secondly, a first heterogeneous sensor and a second heterogeneous sensor are installed at each collection node; the stress value, longitudinal displacement value and lateral slip value of the metal roof panel corresponding to each collection node are obtained through the first heterogeneous sensor and the second heterogeneous sensor;

[0072] Among them, the first heterogeneous sensor includes a strain gauge and a laser sensor, which are respectively used to obtain the stress value and longitudinal displacement value of the metal roof panel corresponding to the collection node; the second heterogeneous sensor includes an ultrasonic sensor, which is used to obtain the lateral slip value of the metal roof panel corresponding to the collection node.

[0073] In the above step S2, the change of the state of the metal roof is a process that changes over time. Since the metal panel will produce elastic deformation under the influence of wind, the data exceeding the threshold at only one moment cannot determine whether the roof has been damaged. Therefore, it is not possible to determine whether to alarm based on the data at a single moment. What needs to be paid attention to is the frequency of the metal roof state exceeding the threshold within a period of time; in addition, the change of the metal roof state is a large-scale state change. There will be data associations between adjacent nodes in the same area of ​​the metal roof. When the panel state in a certain area is abnormal, the abnormal data will be reflected in one or more nodes in this area. Therefore, the data analysis and feature extraction of the metal roof need to be considered from the two dimensions of time and space to extract the statistical features of the data.

[0074] In the metal roof monitoring system, the selection of feature quantities is very critical and will seriously affect the complexity and accuracy of the decision algorithm. The purpose of feature selection is to select the feature vector that best reflects the roof status based on the collected sensor data. The present invention intends to use the time domain analysis method to extract feature quantities. Time domain analysis is to perform statistical analysis on the time domain signals collected by the sensor, which can usually extract the mean, variance, standard deviation and correlation coefficient of the time domain signals.

[0075] In an embodiment of the present invention, firstly, a preset number of acquisition cycles corresponding to each acquisition node is taken as a time window; secondly, the stress value, longitudinal displacement value and lateral slip value corresponding to each acquisition node in each time window are calculated, and the characteristic quantities corresponding to the stress value, longitudinal displacement value and lateral slip value of each acquisition node in each time window are respectively obtained; and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor corresponding to each acquisition node is calculated at the same time; wherein the characteristic quantity of the stress value includes the maximum stress value, the stress standard deviation and the stress mean value; the characteristic quantity of the longitudinal displacement value includes the longitudinal displacement mean value; and the characteristic quantity of the lateral slip value includes the lateral slip mean value.

[0076] In actual monitoring, 5 acquisition nodes can be used as a whole to form an area, and an alarm mechanism can be set for each area; when the weather is bad or the data at a certain point is abnormal, enter the real-time monitoring mode; the data acquisition cycle can be 5s, and the data time series can be divided into a time window of 1min, and the state of this node can be judged by statistical analysis of the characteristic quantity within 1min. For regional statistics, the ratio of nodes in a region whose mean reaches the threshold is counted. For single-point statistics, the stress mean within a time window reflects the central change trend; the stress standard deviation reflects the vibration amplitude of the roof; the maximum stress reflects the severity of the deformation of the roof panel within a time window; among them, the calculation formula of the stress standard deviation is expressed as:

[0077]

[0078] In this formula, A(str) represents the stress standard deviation; n represents the total number of data measured by the strain gauge; x i represents the stress value measured by the first heterogeneous sensor at the i-th moment; X(str) represents the stress mean value within a time window;

[0079] The correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor at the same time can reflect the correlation of the change trends of the two sensors, and thus can be used to identify sensor failures; wherein, the calculation formula of the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is expressed as:

[0080]

[0081] In this formula, X represents the measurement value of the first heterogeneous sensor; Y represents the measurement value of the second heterogeneous sensor; ρ XY represents the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor; D(X) represents the variance of X; D(Y) represents the variance of Y; COV(X,Y) represents the covariance of X and Y; X(str) represents the stress mean within a time window; Y(dis) represents the longitudinal displacement mean within a time window; x i represents the stress value measured by the first heterogeneous sensor at the i-th moment; y i represents the longitudinal displacement value measured by the first heterogeneous sensor at the i-th moment.

[0082] In the above step S3, the overall alarm idea is to find the association between the data of the first heterogeneous sensor and the second heterogeneous sensor at the same location to perform fault identification; horizontal comparison of sensor data at various locations at the same time can eliminate abnormal points and reduce misjudgment. The metal roof warning strategy needs to distinguish three situations: normal roof, roof fault warning, and sensor failure; among them, the normal roof is manifested as the failure of the roof is a process of overall regional change; when the weather is bad, all node data in the same area will change significantly. At this time, the data abnormality of a certain node needs to be alarmed. When the weather is good, only one node data abnormality does not need to be alarmed; the roof fault warning is manifested as an alarm in units of regions, and the main alarm forms include roof bulging, roof plastic deformation, lateral slippage of roof panels, etc.; sensor failure is judged by the association of heterogeneous sensor data at the same location.

[0083] The process of establishing a decision tree model and making judgments based on the established model is to conduct experiments based on the three situations of normal roof, roof fault warning, and sensor failure mentioned above, collect sample data corresponding to various categories in a targeted manner, and extract feature quantities to obtain sample sets; divide the sample sets into training sets and test sets, and train the decision tree model based on the training set samples; finally, verify the effect of the decision tree model, input the test set feature quantities, obtain the judgment results through the trained decision tree model, compare the judgment results with the actual categories, and calculate the judgment accuracy.

[0084] In the embodiment of the present invention, the decision tree model is trained by taking the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor as input, and taking the corresponding metal roof panel and the state result of the first heterogeneous sensor or the second heterogeneous sensor as output. The state result of the metal roof panel includes a normal panel and a panel failure; the panel failure includes one or more of panel deformation, cracks and wind lift;

[0085] In the embodiment of the present invention, refer to the attached Figure 4 As shown, for each acquisition node, the process of making judgments through the trained decision tree model specifically includes:

[0086] S31, inputting the longitudinal displacement mean, lateral slip mean, stress maximum, stress standard deviation and stress mean within the current time window, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor into the trained decision tree model;

[0087] S32, judging the ratio of the longitudinal displacement mean or the lateral slip mean by a decision tree model; first, setting a preset value, if within a time window, the ratio of the data of the longitudinal displacement mean or the lateral slip mean that exceeds the preset value to the total data of the longitudinal displacement mean or the lateral slip mean is taken as a ratio, and the ratio is compared with a first threshold;

[0088] If the ratio of the longitudinal displacement mean or the lateral slip mean is greater than the first threshold, it indicates that the metal roof panel at the acquisition node is faulty and the fault may be recovered; if the ratio of the longitudinal displacement mean or the lateral slip mean is less than the first threshold, step S33 is performed; in the embodiment of the present invention, the first threshold is taken as 40%;

[0089] S33, judging the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor by the decision tree model; if the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is less than the second threshold, it indicates that the first heterogeneous sensor or the second heterogeneous sensor corresponding to the acquisition node is faulty; if the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is greater than the second threshold, proceeding to step S34; in the embodiment of the present invention, the second threshold is taken as -0.5;

[0090] S34. The decision tree model is used to judge the maximum stress value, the standard deviation of stress and the mean stress value. If the maximum stress value is greater than the third threshold, or the standard deviation of stress is greater than the fourth threshold, or the mean stress value is greater than the fifth threshold, it indicates that the metal roof panel corresponding to the acquisition node has an irrecoverable fault. If the maximum stress value is less than the third threshold, and the standard deviation of stress is less than the fourth threshold, and the mean stress value is less than the fifth threshold, it indicates that the metal roof panel corresponding to the acquisition node has no fault. In the embodiment of the present invention, the third threshold value is taken as 111; the fourth threshold value is taken as 0.3; and the fifth threshold value is taken as 50.

[0091] The embodiment of the present invention also provides a metal roof state judgment system based on a decision tree algorithm, applying the above method, the system comprises: a collection module, a preprocessing module, a judgment module and an early warning module;

[0092] Among them, the acquisition module is used to obtain the stress value, longitudinal displacement value and lateral slip value of the metal roof panel to be monitored through the first heterogeneous sensor and the second heterogeneous sensor; the preprocessing module is used to preprocess the stress value, longitudinal displacement value and lateral slip value by using the time domain analysis method, and obtain the characteristic quantities corresponding to the stress value, longitudinal displacement value and lateral slip value respectively; at the same time, the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is calculated; the judgment module is used to judge the status of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor according to the characteristic quantities corresponding to the stress value, longitudinal displacement value and lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor; the early warning module is used to issue a corresponding early warning when there is a fault in the metal roof panel, the first heterogeneous sensor and / or the second heterogeneous sensor in the status result.

[0093] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for judging the state of a metal roof based on a decision tree algorithm, characterized in that: include: S1. Obtaining the stress value, longitudinal displacement value and lateral slip value of the metal roof panel to be monitored through the first heterogeneous sensor and the second heterogeneous sensor; S2. Preprocessing the stress value, the longitudinal displacement value and the lateral slip value by using a time domain analysis method to obtain characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value respectively; and calculating the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor at the same time; S3, inputting the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor into the trained decision tree model, and outputting the status results of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor; The decision tree model is trained by taking the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor as inputs, and taking the corresponding metal roof panel and the state result of the first heterogeneous sensor or the second heterogeneous sensor as outputs; The S1 specifically includes: S11, dividing the metal roof panel into a plurality of collection areas; each of the collection areas contains a preset number of collection nodes; S12, a first heterogeneous sensor and a second heterogeneous sensor are installed at each of the acquisition nodes; the stress value, the longitudinal displacement value and the lateral slip value of the metal roof panel corresponding to each of the acquisition nodes are obtained through the first heterogeneous sensor and the second heterogeneous sensor; The S2 specifically includes: Taking a preset number of collection cycles corresponding to each of the collection nodes as a time window; Calculating the stress value, longitudinal displacement value and lateral slip value corresponding to each acquisition node in each time window, and obtaining the characteristic quantities corresponding to the stress value, longitudinal displacement value and lateral slip value of each acquisition node in each time window; At the same time, calculating the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor corresponding to each of the acquisition nodes; For each of the acquisition nodes, the corresponding S3 specifically includes: S31, inputting the longitudinal displacement mean, lateral slip mean, stress maximum, stress standard deviation and stress mean within the current time window, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor into the trained decision tree model; S32, judging the ratio of the longitudinal displacement mean or the lateral slip mean by the decision tree model; If the ratio of the longitudinal displacement mean or the lateral slip mean is greater than the first threshold, it indicates that the metal roof panel at the acquisition node has a fault, and the fault has the possibility of recovery; If the ratio of the longitudinal displacement mean or the lateral slip mean is less than the first threshold, proceed to step S33; S33, judging the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor by the decision tree model; If the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is less than a second threshold, it indicates that the first heterogeneous sensor or the second heterogeneous sensor corresponding to the acquisition node is faulty; If the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor is greater than a second threshold, proceed to step S34; S34, judging the maximum stress value, the stress standard deviation and the stress mean value by the decision tree model; If the maximum stress value is greater than the third threshold, or the standard deviation of the stress is greater than the fourth threshold, or the mean stress value is greater than the fifth threshold, it indicates that the metal roof panel corresponding to the acquisition node has an irreversible fault; If the maximum stress value is less than the third threshold, the standard deviation of the stress is less than the fourth threshold, and the mean stress value is less than the fifth threshold, it indicates that there is no fault in the metal roof panel corresponding to the acquisition node.

2. A method for judging the state of a metal roof based on a decision tree algorithm according to claim 1, characterized in that: Also includes: S4. If, in the status result, the metal roof panel, the first heterogeneous sensor and / or the second heterogeneous sensor are faulty, a corresponding warning is issued through an alarm device.

3. A method for judging the state of a metal roof based on a decision tree algorithm according to claim 1, characterized in that: The first heterogeneous sensor includes a strain gauge and a laser sensor, which are respectively used to obtain the stress value and the longitudinal displacement value of the metal roof panel corresponding to the acquisition node; The second heterogeneous sensor includes an ultrasonic sensor, which is used to obtain a lateral slip value of the metal roof panel corresponding to the collection node.

4. A method for judging the state of a metal roof based on a decision tree algorithm according to claim 1, characterized in that: The characteristic quantities of the stress value include the maximum stress value, the standard deviation of stress and the mean stress value; The characteristic quantity of the longitudinal displacement value includes the longitudinal displacement mean; The characteristic quantity of the lateral slip value includes a lateral slip mean value.

5. A metal roof status judgment system based on a decision tree algorithm, characterized in that: The method described in any one of claims 1 to 4 is applied, the system comprising: a collection module, a preprocessing module and a judgment module; The acquisition module is used to obtain the stress value, longitudinal displacement value and lateral slip value of the metal roof panel to be monitored through the first heterogeneous sensor and the second heterogeneous sensor; The preprocessing module is used to preprocess the stress value, the longitudinal displacement value and the lateral slip value by using a time domain analysis method to obtain characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value respectively; and calculate the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor at the same time; The judgment module is used to judge the status of the metal roof panel and the first heterogeneous sensor or the second heterogeneous sensor according to the characteristic quantities corresponding to the stress value, the longitudinal displacement value and the lateral slip value, and the correlation coefficient between the first heterogeneous sensor and the second heterogeneous sensor.

6. A metal roof status judgment system based on decision tree algorithm according to claim 5, characterized in that: It also includes an early warning module; The early warning module is used to issue a corresponding early warning when a fault occurs in the metal roof panel, the first heterogeneous sensor and / or the second heterogeneous sensor in the status result.

Images