Communication iron tower with fault identification device

Abstract

The invention provides a communication iron tower with a fault recognition device, and belongs to the technical field of fault recognition. The communication iron tower with the fault recognition device comprises a feature acquisition module used for acquiring fault features of the communication iron tower; the fault identification module is used for identifying the fault features based on the decision trees of the target number to obtain a plurality of fault results; the number of the fault results is equal to that of the decision trees; and the fault decision-making module is used for carrying out weighted calculation on the multiple fault results to obtain a target fault result, and the training data sets of the decision-making tree model are different. The accuracy and reliability of iron tower fault identification can be improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of fault identification, and in particular to a communication tower with a fault identification device. Background Art

[0002] In modern communication networks, communication towers are key infrastructure for signal transmission and reception, and are widely distributed in cities, villages, and remote areas. Their stable operation is crucial to ensuring the continuity and reliability of communication services. However, communication towers are exposed to the natural environment for a long time, facing the test of severe weather conditions such as strong winds, heavy rains, and lightning strikes. At the same time, the aging of their own structures, the failure of electrical equipment, and the loss of communication lines may all cause communication towers to fail, thus affecting the quality of communication.

[0003] Traditional communication tower fault detection methods mostly rely on manual inspections. This method not only consumes a lot of manpower, material resources and time, but also makes it difficult to detect and handle sudden faults in a timely manner due to the long detection cycle. At the same time, the accuracy of manual detection depends largely on the experience and professional level of the detection personnel, which has certain subjectivity and limitations. With the rapid development of communication technology and the continuous improvement of communication quality requirements, there is an urgent need for an efficient and intelligent fault identification device that can monitor and identify communication tower faults in real time and accurately. Summary of the invention

[0004] The embodiment of the present disclosure provides a communication tower with a fault identification device to solve the problem of low accuracy and reliability in tower fault identification.

[0005] The embodiment of the present disclosure provides a communication tower with a fault identification device, including: a feature acquisition module, used to acquire the fault features of the communication tower; A fault identification module is used to identify fault features based on a target number of decision trees to obtain multiple fault results; the number of fault results is equal to the number of decision trees; The fault decision module is used to perform weighted calculation on multiple fault results to obtain a target fault result, wherein the training data set of the decision tree model is different.

[0006] In an exemplary embodiment of the present disclosure, a communication tower having a fault identification device further includes: a decision tree training module; A decision tree training module, configured to train a first number of decision trees based on a first fault data set; training a second number of decision trees based on a second fault data set; training a third number of decision trees based on a third fault data set; The data quantities of each fault type in the first fault data set, the second fault data set and the third fault data set are different, and the sum of the first quantity, the second quantity and the third quantity is the target quantity.

[0007] In an exemplary embodiment of the present disclosure, the decision tree training module is specifically used to: determining hyperparameters for a first number of decision trees based on a first tuning strategy; determining hyperparameters for a second number of decision trees based on a second tuning strategy; determining hyperparameters of a third number of decision trees based on a third tuning strategy; Among them, the adjustment directions of the first adjustment strategy, the second adjustment strategy and the third adjustment strategy are different.

[0008] In an exemplary embodiment of the present disclosure, the hyperparameters include: splitting criterion, maximum depth, minimum sample split, minimum sample leaf, and maximum number of features; The number of structural deformation fault data in the first fault data set accounts for the largest proportion; The decision tree training module is also used for: In response to the degree of influence of the external environment on the structural deformation fault being greater than or equal to a first environmental influence threshold, using information entropy as a splitting criterion for a first number of decision trees; In response to the influence of the external environment on the structural deformation failure being less than a first environmental influence threshold, using Gini impurity as a splitting criterion for a first number of decision trees; In response to the structural complexity of the communication tower being greater than the first complexity, increasing the reference value of the maximum depth of the first number of decision trees according to the first depth step to obtain the maximum depth; Using the reference value of the minimum sample split as the minimum sample split of the first number of decision trees, and using the reference value of the minimum sample leaf as the minimum sample leaf of the first number of decision trees; In response to the feature dimension of the structural deformation fault data being smaller than the first dimension, a maximum feature number of the first number of decision trees is determined as a square root of the number of features.

[0009] In an exemplary embodiment of the present disclosure, the fault decision module is specifically configured to: In response to the type of the fault feature being a deformation type, increasing the weight of the fault results obtained by the first number of decision trees based on a first deformation weight step, decreasing the weight of the fault results obtained by the second number of decision trees based on a second deformation weight step, and decreasing the weight of the fault results obtained by the third number of decision trees based on a third deformation weight step; In response to the type of the fault feature being electrical, the weight of the fault results obtained by the first number of decision trees is reduced based on a first electrical weight step, the weight of the fault results obtained by the second number of decision trees is increased based on a second electrical weight step, and the weight of the fault results obtained by the third number of decision trees is reduced based on a third electrical weight step; In response to the type of fault feature being communication type, the weight of the fault results obtained by the first number of decision trees is reduced based on the first communication weight step, the weight of the fault results obtained by the second number of decision trees is reduced based on the second communication weight step, and the weight of the fault results obtained by the third number of decision trees is increased based on the third communication weight step.

[0010] In an exemplary embodiment of the present disclosure, a communication tower having a fault identification device further includes: a step size adjustment module; A step size adjustment module, used for adjusting the first deformation weight step size, the second deformation weight step size and the third deformation weight step size based on the data quantity of each fault type in the first fault data set; adjusting the first electrical weight step, the second electrical weight step, and the third electrical weight step based on the data quantity of each fault type in the second fault data set; The first communication weight step size, the second communication weight step size, and the third communication weight step size are adjusted based on the data quantity of each fault type in the third fault data set.

[0011] In an exemplary embodiment of the present disclosure, the step size adjustment module is further configured to: In response to the proportion of data of electrical fault types in the first fault data set exceeding the first electrical ratio, and / or the proportion of data of communication fault types in the first fault data set exceeding the first communication ratio, the first deformation weight step is reduced based on the first deformation adjustment step, the second deformation weight step is increased based on the second deformation adjustment step, and the third deformation weight step is increased based on the third deformation adjustment step.

[0012] In an exemplary embodiment of the present disclosure, the number of structural deformation fault data accounts for the largest proportion in the first fault data set, the number of electrical fault data accounts for the largest proportion in the second fault data set, and the number of communication line fault data accounts for the largest proportion in the third fault data set.

[0013] In an exemplary embodiment of the present disclosure, a communication tower having a fault identification device further includes: a first signal processing module; The first signal output module is used to send the target fault result to the terminal device.

[0014] In an exemplary embodiment of the present disclosure, a communication tower having a fault identification device further includes: a second signal processing module; The second signal processing module is used for outputting an alarm signal in response to the target fault result being a fault.

[0015] The beneficial effects of a communication tower with a fault identification device provided by the embodiment of the present disclosure are: The present disclosure utilizes multiple decision trees based on different training data sets for fault identification, and can analyze fault features from multiple angles, thereby increasing the accuracy and reliability of identification. Due to the different training data, each decision tree can capture different fault modes or feature combinations, which helps to more comprehensively understand and identify faults. The fault decision module can quickly integrate the outputs of different decision trees by performing weighted calculations on multiple fault results to obtain the final target fault result, thereby reducing the time for manual analysis and judgment. The training data sets of each decision tree model are different, so that the present disclosure can better adapt to different fault scenarios and data distributions. Even if a decision tree gives an incorrect fault result due to data bias or noise, other decision trees may provide correct information, thereby enhancing the robustness of the present disclosure and improving the accuracy and reliability of communication tower fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0017] Figure 1 is a structural schematic diagram of a communication tower with a fault identification device provided by an embodiment of the present disclosure; Figure 2 It is a structural schematic diagram of a second communication tower with a fault identification device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] In order to enable people in the technical field to better understand the present solution, the technical solution in the embodiment of the present solution will be clearly described below in conjunction with the drawings in the embodiment of the present solution. Obviously, the described embodiment is an embodiment of a part of the present solution, not all of the embodiments. Based on the embodiments in the present solution, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present solution.

[0019] The term "including" and any other variations in the specification and claims of this solution and the above drawings mean "including but not limited to", and is intended to cover non-exclusive inclusions and is not limited to the examples listed in the text. In addition, the terms "first" and "second" are used to distinguish different objects, not to describe a specific order.

[0020] The following is a detailed description of the implementation of the present disclosure in conjunction with the specific drawings: Figure 1 A schematic diagram of the structure of a communication tower with a fault identification device provided by an embodiment of the present disclosure. Figure 1 , the communication tower with a fault identification device comprises a feature acquisition module 10, which is used to obtain the fault features of the communication tower; A fault identification module 11 is used to identify fault features based on a target number of decision trees to obtain multiple fault results; the number of fault results is equal to the number of decision trees; The fault decision module 12 is used to perform weighted calculation on multiple fault results to obtain a target fault result, wherein the training data sets of the decision tree model are different.

[0021] In this embodiment, the feature acquisition module 10 collects various fault-related information from the communication tower. This information is the fault characteristics and is the information input end of the fault identification system. The fault characteristics can cover multiple aspects of the communication tower, such as physical structure, electrical performance, operating environment, etc.

[0022] Various sensors are installed on the communication tower, including inclination sensors for measuring the inclination angle of the tower, temperature sensors for detecting temperature, electrical parameter sensors for monitoring current and voltage, etc. The data collected by these sensors are all fault characteristics. The feature acquisition module 10 will integrate the data transmitted by these sensors as the basis for fault identification.

[0023] In this embodiment, various sensors can be installed in the following locations: Tilt sensors: Tilt sensors are installed at different heights and directions of the communication tower. For example, a tilt sensor is installed at a certain height (such as 10-20 meters) on each of the four sides of the tower, starting from the bottom. This allows real-time monitoring of the tower's tilt at different heights, and timely detection of whether the tower is tilted due to foundation settlement, strong winds or other external forces.

[0024] Strain gauges: Pasted on the surface of the main load-bearing components of the tower (such as the main poles, cross bars and diagonal braces of the tower body). The focus is on the connection parts and stress concentration areas of the components, because these places are prone to produce large strains when the structure deforms. By monitoring the changes in strain, we can understand the stress state of the components and determine whether there is local deformation or component damage.

[0025] Displacement sensor: Installed at key nodes of the tower (such as the connection between different components) or at locations prone to relative displacement. For example, a displacement sensor is installed at the connection point between the tower platform and the tower body to measure the displacement of the platform relative to the tower body, thereby monitoring the stability of the structure.

[0026] Settlement sensors: Settlement sensors are installed at the four corners or edges of the tower foundation to monitor the settlement of the foundation. Foundation settlement may cause the entire tower to tilt or structural damage. By monitoring settlement data in real time, possible structural failures can be warned in advance.

[0027] Current sensors and voltage sensors: Both can be installed at the incoming and outgoing terminals of distribution boxes and cabinets. By monitoring the current and voltage of the incoming and outgoing lines, electrical faults such as current overload and abnormal voltage fluctuations can be detected in a timely manner. For example, high-precision current and voltage sensors are installed on the lines that supply power to communication equipment. When the equipment is short-circuited or overloaded, sudden changes in current and fluctuations in voltage can be quickly detected.

[0028] Harmonic analyzer: connected to the main busbar of the electrical system, used to detect the harmonic content in the power grid. Harmonics are generated by nonlinear loads (such as rectifiers and inverters in electronic equipment). Excessive harmonics may cause overheating, damage or communication interference of electrical equipment. Harmonic analyzers can detect and take measures to suppress harmonics in a timely manner.

[0029] Signal strength sensor: Installed near the connection point between the feeder and the antenna, it is used to monitor the strength of the transmitted and received signals. Changes in signal strength may be caused by feeder damage, antenna failure, or external interference. By monitoring the signal strength in real time, communication line failures can be discovered in a timely manner.

[0030] Bit error rate tester: installed at the intermediate node of the communication line (such as optical fiber amplifier, signal repeater, etc.), used to monitor the bit error rate during digital signal transmission. Bit error rate is one of the key indicators to measure the quality of communication lines. The increase in bit error rate may be caused by line aging, electromagnetic interference or equipment failure. The bit error rate tester can locate the fault location in time and take measures to repair it. The fault identification device can be placed in the machine room at the bottom of the tower. The above-mentioned types of sensors are not exhaustive, but only some sensors, and the fault data set contains the monitoring data of the above sensors.

[0031] The fault identification module 11 analyzes and judges the fault features transmitted by the feature acquisition module 10, thereby obtaining multiple fault results. The decision tree here is a classification and regression model based on a tree structure, which determines the fault category by gradually dividing the fault features. Each decision tree runs independently, so the number of fault results is the same as the number of decision trees. Different decision trees have different interpretation angles on fault features because they are based on different training data sets, which increases the comprehensiveness and accuracy of fault identification. For example, assuming that there are 5 decision trees, the fault identification module 11 inputs the collected communication tower fault feature data into these 5 decision trees respectively, and each decision tree judges the fault according to its own learned rules, and may obtain different fault results such as "slight structural deformation fault" and "electrical component overheating fault", and a total of 5 fault results are obtained. The target number is the number of all decision trees, which can be determined through experiments and tests according to the required fault identification performance.

[0032] The fault decision module 12 performs comprehensive processing on the multiple fault results generated by the fault identification module 11, and obtains the final target fault result by weighted calculation. Since different decision trees have different focuses and reliability on fault judgment, different weights are assigned to the fault results of each decision tree, and these results are accumulated or other mathematical operations are performed according to the weights to obtain a more authoritative and accurate final fault judgment. The different training data sets of the decision tree model here mean that there are differences in the fault modes and laws learned by different decision trees. For example, the weights of 100 decision trees are the same, of which the outputs of 80 decision trees are electrical insulation faults, the outputs of 15 decision trees are electrical short circuits, and the outputs of 5 decision trees are no faults. Then, the final target fault result is electrical insulation faults.

[0033] It can be concluded from the above that the present disclosure utilizes multiple decision trees based on different training data sets for fault identification, and can analyze fault features from multiple angles, thereby increasing the accuracy and reliability of identification. Due to the different training data, each decision tree can capture different fault modes or feature combinations, which helps to understand and identify faults more comprehensively. The fault decision module 12 can quickly integrate the outputs of different decision trees by performing weighted calculations on multiple fault results to obtain the final target fault result, thereby reducing the time for manual analysis and judgment. The training data sets of each decision tree model are different, so that the present disclosure can better adapt to different fault scenarios and data distributions. Even if a decision tree gives an incorrect fault result due to data bias or noise, other decision trees may provide correct information, thereby enhancing the robustness of the present disclosure and improving the accuracy and reliability of communication tower fault identification.

[0034] Figure 2is a structural schematic diagram of a second communication tower with a fault identification device provided by an embodiment of the present disclosure. Figure 2 , in one embodiment of the present disclosure, a communication tower with a fault identification device further includes: a decision tree training module 13; A decision tree training module 13, configured to train a first number of decision trees based on a first fault data set; training a second number of decision trees based on a second fault data set; training a third number of decision trees based on a third fault data set; The data quantities of each fault type in the first fault data set, the second fault data set and the third fault data set are different, and the sum of the first quantity, the second quantity and the third quantity is the target quantity.

[0035] In one embodiment of the present disclosure, the number of structural deformation fault data accounts for the largest proportion in the first fault data set, the number of electrical fault data accounts for the largest proportion in the second fault data set, and the number of communication line fault data accounts for the largest proportion in the third fault data set.

[0036] In this embodiment, the purpose is to train different decision trees to identify different fault types, thereby achieving a more accurate identification.

[0037] The first fault data set mainly contains data on structural deformation faults and normal data, including the inclination angle of the tower, the deformation of the components, stress data at different positions, etc., as well as the structural deformation fault type labels corresponding to these features, such as "tower tilting", "component bending", "normal structure", etc.

[0038] The second fault data set mainly contains a large amount of electrical fault data and normal data, such as voltage fluctuation, current anomaly, short circuit and open circuit data, as well as their corresponding electrical fault labels, such as "overvoltage fault", "short circuit fault", "electrical normal", etc.

[0039] The third fault data set mainly contains data on communication line faults, such as signal attenuation, abnormal bit error rate, transmission delay and other fault data as well as normal data, and corresponding communication line fault labels, such as "weak signal fault", "high bit error rate fault", "normal communication", etc.

[0040] The first number, the second number and the third number are the number of decision trees in the above three data sets, respectively, which can be determined according to the actual situation. For example, among the types and times of historical faults, the probability of deformation faults is relatively small, accounting for 20%, and electrical faults and communication faults each account for 40%. If there are 100 decision trees in total, the first number can be assigned to 20, and the second and third numbers are both 40. It can also be adjusted and determined according to the amount of data in the data set, or, if it is considered that deformation faults are more serious, more attention is given, and the first number is appropriately increased, that is, it can be set according to the actual situation.

[0041] For example, the first number, the second number, and the third number may be determined according to the fault complexity in the corresponding fault data sets. Specifically, in response to the fault complexity in the first fault data set being greater than or equal to the first complexity, the reference value of the first number is increased according to the first adjustment number. In response to the fault complexity in the first fault data set being less than the first complexity, taking the reference value of the first quantity as the first quantity; In response to the fault complexity in the second fault data set being greater than or equal to the second complexity, increasing the reference value of the second quantity according to the second adjustment quantity; In response to the fault complexity in the second fault data set being less than the second complexity, taking the reference value of the second quantity as the second quantity; In response to the fault complexity in the third fault data set being greater than or equal to a third complexity, increasing the reference value of the third quantity according to a third adjustment quantity; In response to the fault complexity in the third fault data set being less than the third complexity, a reference value of the third quantity is taken as the third quantity.

[0042] The data complexity in each data set can be measured from multiple dimensions, and the first quantity adjustment step, the second quantity adjustment step, and the third quantity adjustment step can be determined based on experience. For example, the data complexity of the feature dimension, the discreteness of the fault category, and the feature interaction complexity are calculated respectively. Specifically, the data complexity is calculated according to the first formula: ,in represents the feature dimension complexity, represents the discreteness of fault categories, represents the complexity of feature interaction, is a weight coefficient used to balance the importance of different complexity dimensions.

[0043] is the number of features in the fault data set. For example, if the number of fault features in the first fault data set is 50, then , different fault data sets correspond to The values ​​are different.

[0044] Fault category dispersion It can be calculated using the information entropy formula, ,in Indicates The probability of a fault category appearing in the data set, is the number of fault categories.

[0045] Feature interaction complexity It can be measured by calculating the mutual information or correlation between features. ,in, is the number of features in the fault dataset ( ), Indicates Features and The correlation between features can be calculated by Pearson correlation coefficient, Kendall correlation coefficient or cosine similarity. The first and The above parameters are all numerical values.

[0046] According to the above formula, the complexity of the three fault data sets can be calculated respectively, and compared with the corresponding complexity thresholds (first complexity, second complexity and third complexity). The first complexity, second complexity and third complexity can be set according to preference or determined according to experiments. If it is greater than or equal to the corresponding complexity threshold, it is adjusted according to the pre-set step size (first quantity adjustment step size, second quantity adjustment step size and third quantity adjustment step size). Otherwise, no adjustment is made. The reference value of the first quantity, the reference value of the second quantity and the reference value of the third quantity can be referenced and set according to the number of other decision trees that solve similar problems.

[0047] It can be concluded from the above that the present disclosure introduces a decision tree training module 13, and uses the first fault data set, the second fault data set, and the third fault data set to train different numbers of decision trees respectively, so that each decision tree can better capture the characteristics of a specific fault type, thereby improving the accuracy of overall fault identification. Due to the different data quantities and type distributions in different fault data sets, the trained decision trees also have their own characteristics, which can better cover the various types of faults that communication towers may encounter. This diversified combination of decision trees enables the fault identification device to identify various faults more comprehensively and accurately, and enhances its adaptability in practical applications. This embodiment adjusts the number of decision trees of the first number, the second number, and the third number according to the complexity of the feature dimension, the discreteness of the fault category, and the complexity of the feature interaction, so that the fault identification device can be customized according to different scenarios and needs, so as to better meet the requirements of practical applications and improve the accuracy and reliability of communication tower fault identification.

[0048] Figure 2 is a structural schematic diagram of a second communication tower with a fault identification device provided by an embodiment of the present disclosure. Figure 2 In one embodiment of the present disclosure, the decision tree training module 13 is specifically used to: determining hyperparameters for a first number of decision trees based on a first tuning strategy; determining hyperparameters for a second number of decision trees based on a second tuning strategy; determining hyperparameters of a third number of decision trees based on a third tuning strategy; Among them, the adjustment directions of the first adjustment strategy, the second adjustment strategy and the third adjustment strategy are different.

[0049] In one embodiment of the present disclosure, the hyperparameters include: splitting criterion, maximum depth, minimum sample segmentation, minimum sample leaves, and maximum number of features; The number of structural deformation fault data in the first fault data set accounts for the largest proportion; The decision tree training module 13 is also specifically used for: In response to the degree of influence of the external environment on the structural deformation fault being greater than or equal to a first environmental influence threshold, using information entropy as a splitting criterion for a first number of decision trees; In response to the influence of the external environment on the structural deformation failure being less than a first environmental influence threshold, using Gini impurity as a splitting criterion for a first number of decision trees; In response to the structural complexity of the communication tower being greater than the first complexity, increasing the reference value of the maximum depth of the first number of decision trees according to the first depth step to obtain the maximum depth; Using the reference value of the minimum sample split as the minimum sample split of the first number of decision trees, and using the reference value of the minimum sample leaf as the minimum sample leaf of the first number of decision trees; In response to the feature dimension of the structural deformation fault data being smaller than the first dimension, a maximum feature number of the first number of decision trees is determined as a square root of the number of features.

[0050] In this embodiment, since different decision trees are trained based on different fault data sets and have different fault types (such as the first fault data set mainly for structural deformation faults), different adjustment strategies are adopted to optimize the learning and prediction capabilities of each decision tree set for the corresponding fault type. This differentiated adjustment helps the decision trees better adapt to the characteristics of their respective data sets and the complexity of the corresponding faults.

[0051] The splitting criterion determines how the decision tree selects the best feature when splitting a node. In this embodiment, the splitting criterion can be adjusted according to the influence of the external environment on the structural deformation failure.

[0052] When the influence of the external environment on the structural deformation failure is greater than or equal to the first environmental impact threshold, information entropy is used as the splitting criterion. Information entropy measures the uncertainty of samples in the data set. The use of information entropy can better handle complex feature distributions and uncertain situations. For example, when environmental factors such as strong winds, earthquakes, and extreme temperatures have a greater impact on structural deformation failures, the characteristics of structural deformation failures may become more complex and difficult to predict. Information entropy can more comprehensively consider the information contribution of different features, which helps the decision tree make better splitting decisions.

[0053] When the impact of the external environment on the structural deformation failure is less than the first environmental impact threshold, Gini impurity is used as the splitting criterion. Gini impurity measures the probability of a randomly selected sample from a data set being misclassified. In a relatively stable environment, the characteristics of structural deformation failure may be relatively simple. Gini impurity can more efficiently find the best feature splitting point, allowing the decision tree to learn important features faster.

[0054] The influence of the external environment on the structural deformation failure can be calculated according to the second formula. The second formula is , It indicates the influence of the external environment on the structural deformation failure based on the physical mechanics model. is the number of stress critical locations considered, is the number of deformation critical parts considered. It is The stress generated in each key part under the environmental load, It is The allowable stress of the material in each key part, It is The deformation of key parts under environmental loads, It is The above parameters can be determined according to the construction and maintenance requirements of the tower. The above parameters are all numerical values, and the first environmental impact threshold can be determined according to actual conditions and experiments.

[0055] The structural complexity of a communication tower can be comprehensively judged by the number of simple components, the number of connections between components, or the height, and the first complexity can be determined based on experience. For example, when the number of components of the tower is greater than a preset number of components, the structural complexity of the communication tower is greater than the first complexity, or is set in advance based on experience.

[0056] When the structural complexity of the communication tower is greater than the first complexity, the reference value of the maximum depth is increased according to the first depth step to obtain the final maximum depth. That is, if the structure of the tower is complex (such as the tower structure has multiple layers, multiple connecting parts, etc.), a deeper decision tree is needed to learn the characteristic pattern of its structural deformation failure. By increasing the maximum depth, the decision tree can explore the feature space more deeply to find more refined feature rules to judge the structural deformation failure. However, this adjustment is carried out gradually, and the increase is controlled according to the first depth step to avoid overgrowth of the decision tree and overfitting. The first depth step can be determined based on experiments and can be 3.

[0057] Considering that the first fault data set mainly involves structural deformation faults, its data is relatively stable and reliable. The characteristics of structural deformation faults, such as the inclination of the tower, the deformation of the components, the settlement of the foundation, etc., are usually obtained through more accurate measurement equipment, and the data noise is relatively small. Since the data is relatively clean and there are fewer outliers or noise interference, training the decision tree according to the original reference values ​​of the minimum sample segmentation and the minimum sample leaf may be sufficient to ensure the performance of the model. Using reference values ​​can avoid losing some information that may be useful for structural deformation fault identification due to over-adjustment of these two parameters, and prevent the decision tree from missing important structural deformation feature details when there is sufficient data.

[0058] The reference value of the minimum sample segmentation and the reference value of the minimum sample leaf are relatively general values ​​and can be determined based on experiments or based on values ​​that are commonly used when solving similar problems.

[0059] When the feature dimension of the structural deformation fault data is smaller than the first dimension, the maximum feature number of the first number of decision trees is determined as the square root of the number of features. When the feature dimension is not high, by using the square root of the number of features, the decision tree can consider more feature combinations without increasing the amount of calculation too much, avoiding information loss caused by focusing on only some features, and avoiding the risk of overfitting caused by using too many features. It should be noted that the maximum number of features should be a positive integer, rounded up.

[0060] The above adjustment method is the first adjustment strategy. According to the same idea, the second and third adjustment strategies can be explained as follows: Second adjustment strategy: In response to the influence degree of the external electromagnetic interference on the electrical fault being greater than or equal to the second electromagnetic interference threshold, using information entropy as a splitting criterion for a second number of decision trees; In response to the influence of the external electromagnetic interference on the electrical fault being less than a second electromagnetic interference threshold, using the Gini impurity as a splitting criterion for a second number of decision trees; In response to the complexity of the electrical system being greater than the second complexity, increasing the reference value of the maximum depth of the second number of decision trees according to the second depth step to obtain a maximum depth; In response to a noise level in the electrical fault data being greater than a second noise threshold, increasing the values ​​of the minimum sample split and the minimum sample leaf of the second number of decision trees by respective adjustment amounts; In response to the feature dimension of the electrical fault data being smaller than the second dimension, determining a maximum feature number of the second number of decision trees as a logarithm of the number of features; In response to the feature dimension of the electrical fault data being greater than the second dimension, a maximum number of features of the second number of decision trees is determined as a first ratio of the number of features.

[0061] In this embodiment, the above-mentioned judgment threshold can be determined based on the actual situation and the data in the experimental process. The degree of influence of external electromagnetic interference on electrical faults can be the intensity of electromagnetic interference. The complexity of the electrical system can be indirectly determined according to the complexity of the circuit topology, the number of electrical equipment or the connection method, etc. For example, when the number of electrical equipment is greater than a preset number of equipment, it means that the complexity of the electrical system is greater than the second complexity. Consider the noise level and the proportion of abnormal data in the electrical fault data. If the noise level in the electrical fault data (such as data fluctuations caused by measurement errors, transient interference, etc.) is greater than the second noise threshold, appropriately increase the value of the minimum sample segmentation and the minimum sample leaf. This is to ensure that the decision tree can be based on a sufficient number of reliable samples when splitting nodes and forming leaf nodes, so as to reduce the interference of noise and abnormal data on decision tree learning. If the feature dimension of the electrical fault data (for example, the number of electrical parameters such as voltage, current, power factor, harmonics, etc.) is larger than the second dimension, the maximum feature number is determined as a ratio (i.e., the first ratio) that is smaller than the feature number, so as to avoid overfitting caused by excessive complexity of the decision tree in the high-dimensional feature space; if the feature dimension is smaller than the second dimension, the maximum feature number is determined as the logarithm of the feature number, which should also be a positive integer and rounded up.

[0062] The third adjustment strategy: In response to the degree of influence of the external environment on the communication line failure being greater than or equal to a third environmental interference threshold, using information entropy as a splitting criterion for a third number of decision trees; In response to the influence of the external environment on the communication line failure being less than a third environmental interference threshold, using Gini impurity as a splitting criterion for a third number of decision trees; In response to the complexity of the communication line network being greater than a third complexity, increasing a reference value of a maximum depth of a third number of decision trees according to a third depth step to obtain a maximum depth; In response to a signal fluctuation amplitude in the communication line fault data being greater than a third fluctuation threshold or a data loss ratio being greater than a third loss threshold, increasing the values ​​of the minimum sample split and the minimum sample leaf of the third number of decision trees by corresponding adjustment amounts respectively; In response to the feature dimension of the communication line fault data being smaller than the third dimension, determining the maximum feature number of the third number of decision trees as the square root of the feature number; In response to the feature dimension of the communication line fault data being greater than the third dimension, a maximum feature number of the third number of decision trees is determined as a second ratio of the feature number.

[0063] In this embodiment, the above-mentioned determination threshold can be determined according to the actual situation and the data in the experimental process. The degree of influence of external interference on communication failure can be the intensity of electromagnetic interference. The complexity of the communication line network can be indirectly represented by the line length, the number of branches, the number of access devices, etc. For example, when the number of branches is greater than a preset number, it means that the complexity of the communication line network is greater than the third complexity. Consider the signal fluctuation and data loss in the communication line fault data. If the signal fluctuation amplitude (such as the standard deviation of the signal strength) in the communication line fault data is greater than the third fluctuation threshold, or the data loss ratio is greater than the third loss threshold, the value of the minimum sample segmentation and the minimum sample leaf is appropriately increased. This is to ensure that the decision tree is based on sufficiently stable samples for learning and reduce the adverse effects of signal fluctuations and data loss on the model. If the feature dimension (such as signal strength, signal-to-noise ratio, bit error rate, delay, etc.) of the communication line fault data is greater than the third dimension, the maximum number of features is determined to be less than a ratio of the number of features (such as 0.6 times) to prevent the decision tree from being too complex and causing overfitting; if the feature dimension is less than the third dimension, the maximum number of features is determined to be the square root of the number of features, so that the decision tree can consider more feature combinations without increasing the amount of calculation.

[0064] In this embodiment, different adjustment strategies are set considering that the first fault data set mainly involves structural deformation faults and its data is relatively stable and reliable, but the sensor sensitivity of the second fault data set and the third fault data set is relatively low. It should be noted that the first fault data set, the second fault data set and the third fault data set not only contain fault data, normal data and corresponding labels, but also contain environmental parameters. The environmental parameters are essentially obtained through various sensors, which should exist in themselves and are also a basis for fault identification.

[0065] From the above, it can be concluded that the present disclosure can optimize the learning and prediction capabilities of each decision tree set for the corresponding fault type by setting different hyperparameter adjustment strategies for decision tree sets of different fault types, so that the decision tree can better adapt to the characteristics of the respective data sets and the complexity of the corresponding faults, thereby improving the accuracy of fault identification. This embodiment takes into account the degree of influence of the external environment on the fault, and dynamically adjusts the splitting criterion so that the decision tree can still maintain stable and accurate recognition capabilities in a complex and changing environment. In response to problems such as data noise, signal fluctuations, and data loss, by adjusting hyperparameters such as minimum sample segmentation and minimum sample leaves, the decision tree's ability to handle unstable data is enhanced, the robustness of the present disclosure is improved, and the accuracy and reliability of communication tower fault identification are improved.

[0066] Figure 2 is a structural schematic diagram of a second communication tower with a fault identification device provided by an embodiment of the present disclosure. Figure 2 In one embodiment of the present disclosure, the fault decision module 12 is specifically configured to: In response to the type of the fault feature being a deformation type, increasing the weight of the fault results obtained by the first number of decision trees based on a first deformation weight step, decreasing the weight of the fault results obtained by the second number of decision trees based on a second deformation weight step, and decreasing the weight of the fault results obtained by the third number of decision trees based on a third deformation weight step; In response to the type of the fault feature being electrical, the weight of the fault results obtained by the first number of decision trees is reduced based on a first electrical weight step, the weight of the fault results obtained by the second number of decision trees is increased based on a second electrical weight step, and the weight of the fault results obtained by the third number of decision trees is reduced based on a third electrical weight step; In response to the type of fault feature being communication type, the weight of the fault results obtained by the first number of decision trees is reduced based on the first communication weight step, the weight of the fault results obtained by the second number of decision trees is reduced based on the second communication weight step, and the weight of the fault results obtained by the third number of decision trees is increased based on the third communication weight step.

[0067] In one embodiment of the present disclosure, a communication tower with a fault identification device further includes: a step adjustment module 14; A step size adjustment module 14, configured to adjust the first deformation weight step size, the second deformation weight step size, and the third deformation weight step size based on the data quantity of each fault type in the first fault data set; adjusting the first electrical weight step, the second electrical weight step, and the third electrical weight step based on the data quantity of each fault type in the second fault data set; The first communication weight step size, the second communication weight step size, and the third communication weight step size are adjusted based on the data quantity of each fault type in the third fault data set.

[0068] In one embodiment of the present disclosure, the step size adjustment module 14 is further configured to: In response to the proportion of data of electrical fault types in the first fault data set exceeding the first electrical ratio, and / or the proportion of data of communication fault types in the first fault data set exceeding the first communication ratio, the first deformation weight step is reduced based on the first deformation adjustment step, the second deformation weight step is increased based on the second deformation adjustment step, and the third deformation weight step is increased based on the third deformation adjustment step.

[0069] In this embodiment, because the first number of decision trees are trained based on the first fault data set, and the number of structural deformation fault data in the first fault data set accounts for the largest proportion, when the input is a deformation type fault feature, the results obtained by this part of the decision tree are more valuable for reference, so its weight is increased, and the weights of the other two decision trees are reduced. Similarly, when the input is an electrical fault, the weight of the decision tree trained with the second fault data set should be increased, and the weights of the other two decision trees should be reduced. When the input is a communication type fault, the weight of the decision tree trained with the third fault data set should be increased, and the weights of the other two decision trees should be reduced. The above-mentioned adjustment step size and adjustment base value can be determined based on experience or based on experiments.

[0070] The type of input fault feature can be determined by the type of sensor. For example, inclination sensors, strain sensors or displacement sensors are usually used to monitor the structural deformation of communication towers, such as inclination, strain and displacement of components. When the data of these sensors are abnormal, it is likely that the type of fault feature is structural deformation. Electrical sensors such as current sensors, voltage sensors, and power sensors mainly monitor electrical parameters. When the data of these sensors are abnormal, such as a sudden increase in current (overcurrent) and abnormal voltage fluctuations (too high or too low), it can be determined that the fault feature may belong to electrical faults. Sensors such as signal strength sensors, bit error rate testers, and delay meters are used to monitor the performance of communication lines. When their measurement data is abnormal, such as a decrease in signal strength, an increase in bit error rate, and an increase in delay, the fault feature belongs to communication faults.

[0071] It should be noted that since each fault data set contains all fault types, but the proportions are different, the decision tree trained with any fault data set can identify and monitor various faults.

[0072] Taking into account the different proportions of data in the data set, the reliability of the decision tree obtained by training is also different, so the above-mentioned weight step should be adjusted according to the proportion of data. When the proportion of electrical fault type data in the first fault data set exceeds the first electrical ratio or the proportion of communication fault type data in the first fault data set exceeds the first communication ratio, even if the data of electrical fault type and communication fault type are not the main components of the first fault data set, but the proportion is large (still less than 50%), it will affect the judgment of structural deformation fault, so reduce the first deformation weight step, increase the second deformation weight step and the third deformation weight step.

[0073] Similarly, the second fault data set and the third fault data set are adjusted as follows: Second fault data set: in response to the proportion of deformation fault type data in the second fault data set exceeding the first deformation ratio, and / or the proportion of communication fault type data in the second fault data set exceeding the second communication ratio, the first electrical weight step is reduced based on the first electrical adjustment step, the second electrical weight step is increased based on the second electrical adjustment step, and the third electrical weight step is increased based on the third electrical adjustment step.

[0074] Third fault data set: in response to the proportion of data of deformation fault types in the third fault data set exceeding the second deformation ratio, and / or the proportion of data of electrical fault types in the third fault data set exceeding the second electrical ratio, the first communication weight step is reduced based on the first communication adjustment step, the second communication weight step is increased based on the second communication adjustment step, and the third communication weight step is increased based on the third communication adjustment step.

[0075] The above adjustment step size and adjustment base value may be determined based on experience or experiments.

[0076] It can be concluded from the above that the present disclosure further optimizes the accuracy and efficiency of fault identification by dynamically adjusting the weight step size according to the data proportion of different fault types in each fault data set. By adjusting the weight step size, this embodiment can reduce the interference of the decision tree trained by the data set when identifying other types of faults, thereby improving the accuracy and reliability of communication tower fault identification.

[0077] Figure 2 is a structural schematic diagram of a second communication tower with a fault identification device provided by an embodiment of the present disclosure. Figure 2 , in one embodiment of the present disclosure, a communication tower with a fault identification device further includes: a first signal processing module 15; The first signal processing module 15 is used to send the target fault result to the terminal device.

[0078] In one embodiment of the present disclosure, a communication tower with a fault identification device further includes: A second signal processing module 16; The second signal processing module 16 is used for outputting an alarm signal in response to the target fault result being a fault.

[0079] In this embodiment, the first signal processing module 15 can send the target fault result to the mobile phone of the relevant personnel or the display screen of the central control center, so that the relevant personnel can process and obtain the information of the communication tower in time. The second signal processing module 16 can output an alarm signal when the target fault result is a fault, so as to notify the relevant personnel to handle it in time.

[0080] It can be concluded from the above that after the first signal processing module 15 and the second signal processing module 16 are introduced into the present disclosure, the fault identification device of the communication tower not only has a high-precision fault identification capability, but also has an efficient information transmission and alarm function. This enables relevant personnel to have a more comprehensive understanding of the operating status of the communication tower, respond to and handle faults more quickly, thereby improving the safety and reliability of the communication tower.

[0081] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A communication tower with a fault identification device, characterized in that: include: A feature acquisition module, used to acquire fault features of a communication tower; A fault identification module, used to identify the fault characteristics based on a target number of decision trees to obtain multiple fault results; the number of the fault results is equal to the number of the decision trees; The fault decision module is used to perform weighted calculation on the multiple fault results to obtain a target fault result, wherein the training data sets of the decision tree model are different.

2. A communication tower with a fault identification device as claimed in claim 1, characterized in that: Also includes: Decision tree training module; A decision tree training module, configured to train a first number of decision trees based on a first fault data set; training a second number of decision trees based on a second fault data set; training a third number of decision trees based on a third fault data set; The first fault data set, the second fault data set and the third fault data set have different data quantities for each fault type, and the sum of the first quantity, the second quantity and the third quantity is the target quantity.

3. A communication tower with a fault identification device as claimed in claim 2, characterized in that: The decision tree training module is specifically used for: determining hyperparameters of the first number of decision trees based on a first adjustment strategy; determining hyperparameters of the second number of decision trees based on a second adjustment strategy; determining hyperparameters of the third number of decision trees based on a third adjustment strategy; The first adjustment strategy, the second adjustment strategy and the third adjustment strategy have different adjustment directions.

4. A communication tower with a fault identification device as claimed in claim 3, characterized in that: The hyper parameters include: splitting criterion, maximum depth, minimum sample segmentation, minimum sample leaves and maximum number of features; The number of structural deformation fault data in the first fault data set accounts for the largest proportion; The decision tree training module is further specifically used for: In response to the degree of influence of the external environment on the structural deformation fault being greater than or equal to a first environmental influence threshold, using information entropy as a splitting criterion for a first number of decision trees; In response to the influence of the external environment on the structural deformation failure being less than a first environmental influence threshold, using Gini impurity as a splitting criterion for a first number of decision trees; In response to the structural complexity of the communication tower being greater than the first complexity, increasing the reference value of the maximum depth of the first number of decision trees according to the first depth step to obtain the maximum depth; Using the reference value of the minimum sample split as the minimum sample split of the first number of decision trees, and using the reference value of the minimum sample leaf as the minimum sample leaf of the first number of decision trees; In response to the feature dimension of the structural deformation fault data being smaller than the first dimension, a maximum feature number of the first number of decision trees is determined as a square root of the feature number.

5. A communication tower with a fault identification device as claimed in claim 2, characterized in that: The fault decision module is specifically used for: In response to the type of the fault feature being a deformation type, increasing the weight of the fault results obtained by the first number of decision trees based on a first deformation weight step, decreasing the weight of the fault results obtained by the second number of decision trees based on a second deformation weight step, and decreasing the weight of the fault results obtained by the third number of decision trees based on a third deformation weight step; In response to the type of the fault feature being electrical, reducing the weight of the fault results obtained by the first number of decision trees based on a first electrical weight step, increasing the weight of the fault results obtained by the second number of decision trees based on a second electrical weight step, and reducing the weight of the fault results obtained by the third number of decision trees based on a third electrical weight step; In response to the type of the fault feature being communication type, the weight of the fault results obtained by the first number of decision trees is reduced based on a first communication weight step, the weight of the fault results obtained by the second number of decision trees is reduced based on a second communication weight step, and the weight of the fault results obtained by the third number of decision trees is increased based on a third communication weight step.

6. A communication tower with a fault identification device as claimed in claim 5, characterized in that: Also includes: Step size adjustment module; The step size adjustment module is used to adjust the first deformation weight step size, the second deformation weight step size and the third deformation weight step size based on the data quantity of each fault type in the first fault data set; adjusting the first electrical weight step, the second electrical weight step, and the third electrical weight step based on the data quantity of each fault type in the second fault data set; The first communication weight step size, the second communication weight step size, and the third communication weight step size are adjusted based on the amount of data of each fault type in the third fault data set.

7. A communication tower with a fault identification device as claimed in claim 6, characterized in that: The step size adjustment module is further specifically used for: In response to the proportion of data of electrical fault types in the first fault data set exceeding a first electrical ratio, and / or the proportion of data of communication fault types in the first fault data set exceeding a first communication ratio, the first deformation weight step is reduced based on the first deformation adjustment step, the second deformation weight step is increased based on the second deformation adjustment step, and the third deformation weight step is increased based on the third deformation adjustment step.

8. A communication tower with a fault identification device as claimed in claim 2, characterized in that: Among the first fault data set, the number of structural deformation fault data accounts for the largest proportion, the number of electrical fault data accounts for the largest proportion in the second fault data set, and the number of communication line fault data accounts for the largest proportion in the third fault data set.

9. A communication tower with a fault identification device as claimed in claim 1, characterized in that: Also includes: A first signal processing module; The first signal processing module is used to send the target fault result to the terminal device.

10. A communication tower with a fault identification device as claimed in claim 1, characterized in that: Also includes: A second signal processing module; The second signal processing module is used for outputting an alarm signal in response to the target fault result being a fault.

Images