A Smart Performance Testing Method for Construction Equipment Based on AI Algorithms

Abstract

This invention relates to the field of equipment performance testing technology, specifically to an intelligent testing method for construction equipment performance based on AI algorithms. The method includes: collecting operational data from various monitoring points of a tower crane, then acquiring the features of each monitoring point over a time period, and combining these features into a sample; constructing a sample set containing a predetermined number of samples, including those at the current moment, as the first node of an isolated tree in an isolated forest; acquiring the sparsity and non-stationarity of a feature for each sample; acquiring the anomaly discrimination of a feature, and then acquiring the segmentation probability of a feature for each sample; subsequently segmenting the sample set to obtain two sub-sample sets, and so on, until the depth of the isolated tree reaches a predetermined depth; and acquiring the structural fatigue anomaly level at the current moment based on the constructed isolated trees. This application can effectively monitor the performance of tower cranes.

Description

Technical Field

[0001] This invention relates to the field of equipment performance testing technology, and specifically to an intelligent testing method for construction equipment performance based on AI algorithms. Background Technology

[0002] Construction equipment is crucial in building, municipal, and infrastructure projects, undertaking tasks such as hoisting, excavation, and transportation. Its operational performance directly affects construction progress and safety. Tower cranes, in particular, are vital lifting and transportation equipment on construction sites. Their operation involves various complex movements and is subject to diverse environmental influences, including wind loads and construction collisions. With prolonged use, their critical structures and connections are prone to structural fatigue anomalies. Failure to identify these anomalies in a timely manner can lead to serious safety accidents. Therefore, tower crane structural fatigue detection is a critical aspect of construction equipment safety management.

[0003] Currently, the main method for detecting tower crane structural fatigue using real-time monitoring data is to compare monitoring data (such as vibration amplitude, torque, strain, etc.) to determine whether structural fatigue exists by exceeding the range or deviating from the normal model (such as the "strain-load" curve model). However, tower crane operation monitoring data is easily affected by changes in operating conditions such as load and wind speed, as well as collisions during operation. Traditional methods struggle to distinguish between normal operating condition changes and actual structural fatigue anomalies, leading to inaccurate detection results. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide an intelligent detection method for the performance of construction equipment based on AI algorithms. The specific technical solution adopted is as follows:

[0005] One embodiment of the present invention provides an intelligent detection method for the performance of construction equipment based on AI algorithms, the method comprising:

[0006] The operation data of each monitoring point of the tower crane is collected, and the time period of the collected data is evenly divided into different time periods. Based on the operation data of each monitoring point in a time period, the characteristics of each monitoring point in that time period are obtained, and the characteristics of each monitoring point in that time period are combined into a sample. The time corresponding to the sample is the midpoint of that time period.

[0007] The sample set of the first node of an isolated tree in an isolated forest is formed by a predetermined number of samples containing the current time sample; the sparsity of a feature of a sample is obtained based on the difference between a feature of a sample in the sample set and the feature of other samples; the non-stationarity of a feature of a sample is obtained by using the difference between a feature of a sample in the sample set and the feature of two adjacent time steps.

[0008] The anomaly discrimination degree of a feature is obtained by calculating the standard deviation of a feature in each sample in the sample set; the segmentation probability of a feature in a sample is obtained by using the sparsity, non-stationarity, and anomaly discrimination degree of a feature; the sample set is segmented into two sub-sample sets based on the sample with the highest segmentation probability, and so on, until the depth of the isolated tree reaches the preset depth, thus completing the construction of the isolated tree; the structural fatigue anomaly degree at the current moment is obtained based on each completed isolated tree.

[0009] Preferably, the operational data collected from each monitoring point of the tower crane includes:

[0010] The monitoring points on the tower crane include acceleration monitoring points, torque monitoring points, and strain monitoring points. The acceleration monitoring points include those at the slewing point, the base of the boom, the middle of the tower body, and the end of the boom. The torque monitoring point includes the position of the drum. The strain monitoring points include those at the base of the boom, the base of the tower body, and the middle section of the boom. Acceleration sensors, torque sensors, and strain gauges are installed at each acceleration monitoring point, torque sensors, and strain monitoring point to collect acceleration data from each acceleration monitoring point, torque data from each torque monitoring point, and strain data from each strain monitoring point, which serve as the operating data for each monitoring point.

[0011] Preferably, the characteristics of each monitoring point during a given time period are obtained based on the operational data of each monitoring point during that time period, including:

[0012] For an acceleration monitoring point, the average value and root mean square of each acceleration data point within a time period are obtained as the characteristics of that acceleration monitoring point within that time period.

[0013] For torque monitoring points, the average value and standard deviation of each torque data point within a time period are obtained as the characteristics of the torque monitoring points within that time period.

[0014] For a strain monitoring point, the mean strain and the sum of strain amplitudes of each strain data point within a time period are obtained as the characteristics of that strain monitoring point within that time period.

[0015] Preferably, obtaining the sparsity of a feature of a sample based on the difference between a feature of a sample and the same feature of other samples in the sample set includes:

[0016] The sparsity of a feature is obtained by summing the absolute differences between a feature of a sample and the features of other samples in the sample set, and then normalizing the sum.

[0017] Preferably, the degree of non-stationarity of a feature of a sample is obtained by using the difference between a feature of a sample in the sample set and the same feature of samples at two adjacent time points, including:

[0018] Arrange the samples in the sample set according to time sequence to obtain the sample sequence; obtain the absolute value of the difference between a feature of a sample in the sample sequence and the feature of the same feature of two adjacent time steps, and normalize the minimum value of the two absolute values ​​of the difference to obtain the non-stationarity of the feature of the sample.

[0019] Preferably, the anomaly discrimination degree of a feature is obtained based on the standard deviation of a feature in each sample in the sample set, including:

[0020] The anomaly discrimination degree of a feature is obtained by taking the reciprocal of the sum of the standard deviation of a feature and the hyperparameters for each sample in the sample set and normalizing it.

[0021] Preferably, the segmentation probability of a feature in a sample is obtained by utilizing the sparsity, non-stationarity, and anomaly discriminative power of that feature, including:

[0022] The probability of segmentation of a feature is obtained by weighting the non-stationarity and sparsity of the feature in a sample by taking the difference between the anomaly discrimination degree of a feature, the first preset value and the anomaly discrimination degree of the feature as weights, and summing the non-stationarity and sparsity of the feature in the sample.

[0023] Preferably, the sample set is divided into two subsets based on the sample with the highest probability of segmentation, including:

[0024] The maximum value of a feature among the segmentation probabilities of each sample in the sample set is obtained, and the value of that feature is used as the segmentation threshold. Samples in the sample set whose value of that feature is greater than or equal to the segmentation threshold are grouped into a subset. Samples in the sample set whose value of that feature is less than the segmentation threshold are grouped into a subset.

[0025] Preferably, the structural fatigue anomaly level at the current moment is obtained based on each constructed isolated tree, including:

[0026] Calculate the average path length of the sample in all isolated trees of the isolated forest at the current moment, and denote it as the average path length; substitute the average path length into the isolated forest anomaly scoring formula to obtain the structural fatigue anomaly degree at the current moment.

[0027] Preferably, after obtaining the structural fatigue anomaly level at the current moment based on each constructed isolated tree, the method further includes:

[0028] By analyzing the structural fatigue anomaly at all historical moments using box plots, the normal range of structural fatigue anomaly is obtained. The upper limit of the normal range of structural fatigue anomaly is used as the anomaly warning threshold. If the structural fatigue anomaly at the current moment is greater than or equal to the anomaly warning threshold, a structural fatigue anomaly warning for the tower crane is issued.

[0029] The embodiments of the present invention have at least the following beneficial effects: This application collects the operating data of each monitoring point of the tower crane and evenly divides the time period of the collected data into different time periods; based on the operating data of each monitoring point in a time period, the characteristics of each monitoring point in that time period are obtained, and the characteristics of each monitoring point in that time period are combined into a sample; a preset number of samples containing the current time sample are combined into a sample set of the first node of an isolated tree in an isolated forest; then, based on the distribution density of each sample under each feature and the intensity of fluctuation of each feature in the time series, the degree of conformity of each sample with structural fatigue anomaly is analyzed, and the sparsity, instability and anomaly discrimination of each feature of each sample are obtained, thereby obtaining the segmentation probability of each feature of a sample, thereby adjusting the segmentation position of each isolated tree node when generating the isolated forest, which can more quickly segment out structural fatigue anomaly data points and obtain the degree of structural fatigue anomaly at the current time. Compared with the traditional random segmentation method, it can more effectively distinguish between normal working condition change data and structural fatigue anomaly data, which is conducive to improving the accuracy of tower crane performance detection. Attached Figure Description

[0030] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This application provides a flowchart of a method for intelligent detection of construction equipment performance based on AI algorithms. Detailed Implementation

[0032] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an AI-based intelligent detection method for construction equipment performance proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0034] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent detection method for construction equipment performance based on AI algorithms provided by this invention.

[0035] Example:

[0036] The main application scenario of this invention is as follows: When analyzing the degree of fatigue anomalies in tower crane structures based on real-time monitoring data, the monitoring data is easily interfered with by factors such as load, wind speed, and collisions, generating suspected abnormal data. Traditional detection methods based on normal thresholds are difficult to distinguish between normal operating condition changes and actual structural fatigue anomalies, leading to inaccurate detection results and affecting construction and maintenance efficiency. Therefore, this application combines isolated forest analysis to analyze tower crane monitoring data and thereby obtain information on fatigue anomalies in the tower crane structure.

[0037] Please see Figure 1 The diagram illustrates a flowchart of an intelligent performance detection method for construction equipment based on an AI algorithm, provided by an embodiment of the present invention. The method includes the following steps:

[0038] Step S1: Collect the operation data of each monitoring point of the tower crane, and evenly divide the time period of the collected data to obtain different time periods; based on the operation data of each monitoring point in a time period, obtain the characteristics of each monitoring point in that time period, and combine the characteristics of each monitoring point in that time period into a sample, the time corresponding to the sample is the midpoint of that time period.

[0039] Because the tower crane has a complex stress structure and uneven loads on different parts, it is necessary to install relevant sensors at multiple key locations (monitoring points) on the tower crane. The specific arrangement needs to be adjusted according to the actual size and structure of the tower crane.

[0040] The monitoring points on the tower crane include acceleration monitoring points, torque monitoring points, and strain monitoring points. Acceleration monitoring points include those at the slewing point, the base of the boom, the middle of the tower body, and the end of the boom. Torque monitoring points include those at the drum position. Strain monitoring points include those at the base of the boom, the base of the tower body, and the middle section of the boom. Acceleration sensors, torque sensors, and strain gauges are installed at each acceleration monitoring point, torque monitoring point, and strain monitoring point to collect acceleration data from each acceleration monitoring point, torque data from each torque monitoring point, and strain data from each strain monitoring point, which serve as the operating data for each monitoring point.

[0041] Therefore, the acceleration sensors, torque sensors, and strain gauges installed at each acceleration monitoring point, torque monitoring point, and strain monitoring point can be used to collect the operating data of each monitoring point.

[0042] To more accurately monitor tower crane data fluctuations, the sensor monitoring frequency is uniformly set to 500Hz. Furthermore, to cover operational data under different working conditions, data from the month preceding the current monitoring time is used as historical data; that is, the data collection period is the previous month. Since direct analysis of raw sampled data is significantly affected by noise and other factors, and the training efficiency of the detection model is low, based on the real-time requirements for tower crane structural fatigue anomaly detection in construction scenarios, the data collection period is evenly segmented. Specifically, this application divides the data collection period into 10-second intervals, meaning an anomaly detection is performed every 10 seconds.

[0043] Since fatigue anomalies in tower crane structures are usually persistent, in order to reduce the impact of abnormal data caused by sudden events such as collisions on feature values, features that reflect the overall situation of the data within a time period are selected when extracting features.

[0044] For an acceleration monitoring point, the average value and root mean square of each acceleration data point within a certain time period are obtained as the characteristics of that acceleration monitoring point within that time period. The average value of the acceleration data represents the baseline of acceleration fluctuation within that time period. When structural deformation, loosening, or other conditions occur, the acceleration data will exhibit baseline drift, meaning that the average acceleration will shift. The root mean square of the acceleration data represents the vibration energy within that time period. When fatigue anomalies occur, the overall vibration energy will increase.

[0045] For torque monitoring points, the average value and standard deviation of each torque data point within a certain time period are obtained as the characteristics of the torque monitoring points within that time period. The average value of each torque data point reflects the overall load within the current time window. When the tower crane is under high load for a long time, it may aggravate the structural fatigue anomaly of the tower crane. The standard deviation of each torque data point represents the fluctuation of the torque amplitude, which reflects the instability of the tower crane structure. When the structural fatigue anomaly of the tower crane will aggravate the degree of torque fluctuation.

[0046] For a strain monitoring point, the mean strain and the sum of strain amplitudes of each strain data point within a time period are obtained as characteristics of that strain monitoring point within that time period. The strain data exhibits a sinusoidal change within a time period, with the strain amplitude of one cycle being half the difference between a set of peaks and troughs. The sum of strain amplitudes is the sum of the strain amplitudes corresponding to all cycles within that time window. The mean strain reflects the state of the tower crane structure under static load. When structural fatigue occurs, the same load will produce a larger strain, meaning the mean strain will continuously shift. The sum of strain amplitudes is a characteristic value that combines the strain cycle amplitude and the number of cycles. Structural fatigue often causes an increase in both the strain cycle amplitude and the number of cycles, resulting in a significant increase in the sum of strain amplitudes.

[0047] Therefore, the characteristics of each monitoring point in a given time period can be obtained based on the operational data of each monitoring point in that time period. In addition, since the numerical ranges of the characteristics of each monitoring point in the above time period are different, the characteristics obtained above are normalized for the purpose of comparison and analysis. The characteristics used below are all normalized characteristics.

[0048] Finally, samples need to be constructed based on the characteristics of each monitoring point within a time period. Specifically, the characteristics of each monitoring point within a time period are combined into a multi-dimensional data point, which is a sample (each monitoring point corresponds to two features; in this application, there are a total of 8 monitoring points, so a sample includes 16 features). At this point, the midpoint of the time period can be used as the corresponding time of the sample. This allows for the acquisition of multiple samples for subsequent analysis.

[0049] Step S2: Form a sample set of the first node of an isolated tree in an isolated forest by a preset number of samples containing the current time; obtain the sparsity of a feature of a sample based on the difference between a feature of a sample in the sample set and the feature of other samples; obtain the non-stationarity of a feature of a sample based on the difference between a feature of a sample in the sample set and the feature of samples at two adjacent time points.

[0050] Isolation forests leverage the characteristic that outliers differ significantly from normal data and are fewer in number. They identify outliers by randomly splitting the dataset and determining the path length of each data point within the isolation trees. However, factors such as wind speed changes during crane operation can also cause data to deviate from the normal range, potentially leading to rapid segmentation during random partitioning and identification as outliers. Therefore, to reduce the likelihood of rapid segmentation of non-structural fatigue anomalies, the partitioning position of each node is adjusted based on the degree of similarity between each data point's feature values ​​and structural fatigue anomalies.

[0051] Furthermore, based on the commonly used parameters in the Isolation Forest algorithm, 256 samples (including the current time window) are randomly selected from all samples as a sample set. This sample set is the sample set of the first node of an isolated tree in the Isolation Forest. The first node is the node before splitting, that is, the initial node. The preset number is 256, which is the number of sample sets commonly used in Isolation Forest.

[0052] Then, a feature is randomly selected from the various features of the sample as the splitting condition for the first node. For ease of explanation, we will use the a-th feature as an example.

[0053] Plot a one-dimensional coordinate axis for all samples in the sample set according to the eigenvalue of the a-th feature. Since the eigenvalues ​​of tower crane operating data under normal conditions and the same working conditions often have high similarity, and the severity of structural fatigue anomalies usually increases with usage time, meaning there is no strong stability between structural fatigue anomaly data from different time periods, segmentation should begin from locations with sparser sample distribution. Therefore, the sparsity of each sample's location is calculated based on the distance between the a-th features of each sample on the coordinate axis.

[0054] Therefore, the sparsity of a feature in a sample is obtained based on the difference between a feature of a sample and the same feature of other samples in the sample set. Specifically, the sparsity of a feature is obtained by summing the absolute values ​​of the differences between a feature of a sample and the same feature of other samples in the sample set and normalizing the sum.

[0055] The specific model for calculating sparsity is as follows:

[0056] ,

[0057] in, This represents the sparsity of the a-th feature of the i-th sample, which is also the density sparsity of the i-th sample position under the a-th feature. This indicates the total number of samples contained in the sample set corresponding to the first node, which is 256. This represents the value of the a-th feature of the i-th sample in the sample set corresponding to the first node. This represents the value of the a-th feature of the j-th sample in the sample set corresponding to the first node (here, the j-th sample is any sample other than the i-th sample). This represents the absolute value of the difference between the a-th feature of the i-th sample and the j-th sample. The larger the absolute value of this difference, the farther apart the two samples are on the coordinate axis, that is, the sparser the local density of the i-th sample. norm represents the normalization function.

[0058] Since changes in tower crane operating conditions have a significant impact on the feature values ​​of the features, meaning that changes in operating conditions may generate multiple samples with relatively sparse density, in order to more accurately segment from the samples with structural fatigue anomalies and improve the anomaly score of the samples with structural fatigue anomalies, it is also necessary to select the segmentation position based on the degree of conformity of the features of each sample with respect to the performance of structural fatigue anomalies.

[0059] Therefore, the degree of non-stationarity of a feature of a sample can be obtained by using the difference between a feature of a sample and the same feature of samples at two adjacent time points.

[0060] Specifically, the samples in the sample set are arranged in chronological order to obtain a sample sequence; the absolute value of the difference between a feature of a sample in the sample sequence and the feature of the same feature of two adjacent samples is obtained, and the minimum value of the two absolute differences is normalized to obtain the non-stationarity of the feature of the sample.

[0061] Under normal conditions, the characteristics between adjacent time points should be relatively stable. However, when structural fatigue anomalies occur, the characteristics deviate from the normal values, meaning they differ significantly from the characteristics of adjacent time points. To distinguish between changes in operating conditions and changes in characteristic values ​​caused by structural fatigue (operating conditions typically do not change frequently in consecutive time points, while structural fatigue gradually changes towards an abnormal trend over time), the absolute values ​​of the differences between a characteristic of a sample in the sample sequence and the corresponding characteristic of samples at two adjacent time points are obtained. The minimum of these two absolute differences is then taken as the degree of non-stationarity. Alternatively, if a sample has only one adjacent sample (e.g., the first and last samples in the sample sequence), the degree of non-stationarity is directly represented by the absolute value of the difference between a characteristic of that sample and the corresponding characteristic of samples at adjacent time points.

[0062] Here, we take the a-th feature of each sample as an example to illustrate this. From this, we can obtain the non-stationarity of the a-th feature of each sample in the sample set. The non-stationarity of the a-th feature of the i-th sample is expressed as: .

[0063] Step S3: Obtain the anomaly discrimination degree of a feature based on the standard deviation of a feature in each sample in the sample set; obtain the segmentation probability of a feature in a sample by using the sparsity, non-stationarity, and anomaly discrimination degree of a feature; segment the sample set based on the sample with the highest segmentation probability to obtain two sub-sample sets, and so on, segmenting the sub-sample sets until the depth of the isolated tree reaches the preset depth, thus completing the construction of the isolated tree; obtain the structural fatigue anomaly degree at the current moment based on each constructed isolated tree.

[0064] Since the degree of fluctuation of different features under normal conditions may vary, if the a-th feature itself has large fluctuations under normal conditions, then the degree of instability... The differentiation between normal data and structural fatigue data is relatively low; if the consistency of the a-th feature is high under normal conditions, then the degree of non-stationarity is... The data shows high differentiation between normal and structural fatigue data. Therefore, the standard deviation of the a-th feature for all samples is calculated, and the degree of differentiation between normal and abnormal data obtained above is calculated based on this standard deviation.

[0065] The anomaly discrimination level of a feature is obtained by calculating the standard deviation of a feature in each sample set. Specifically, the anomaly discrimination level of a feature is obtained by taking the reciprocal of the sum of the standard deviation of a feature and the hyperparameters in each sample set, and then normalizing the result.

[0066] The specific calculation model for the anomaly discrimination degree of the a-th feature is as follows:

[0067] ,

[0068] in, Let represent the degree of anomaly discrimination of the a-th feature, indicating the degree to which the non-stationarity of the a-th feature distinguishes between normal and anomalous data; Let represent the standard deviation of the a-th feature in all sample sets, and ε be a hyperparameter used to ensure that the fraction is meaningful; here, ε is specified as 0.01. norm represents the normalization function.

[0069] Furthermore, by combining the density sparsity of each data point location with the temporal non-stationarity of each data point feature, the probability of each data point serving as a segmentation point is calculated. Thus, the segmentation probability of a sample's feature is obtained by utilizing its sparsity, non-stationarity, and anomaly discriminative power.

[0070] Specifically, the anomaly discrimination degree of a feature, the difference between the first preset value and the anomaly discrimination degree of the feature are respectively used as the weights of the non-stationarity degree of the feature in a sample and the sparsity degree of the feature in a sample. The non-stationarity degree and sparsity degree of the feature in a sample are then weighted and summed to obtain the segmentation probability of the feature in a sample.

[0071] The specific calculation model for the segmentation probability of the a-th feature of the i-th sample is as follows:

[0072] ,

[0073] in, Let represent the segmentation probability of the i-th sample in the sample set corresponding to the first node, based on the a-th feature. This indicates the probability that the i-th sample will be a segmentation point under the a-th feature. This represents the degree of non-stationarity of the a-th feature of the i-th sample in the sample set corresponding to the first node. This represents the degree of non-stationarity of the a-th feature in distinguishing between normal and abnormal data, that is, the degree of anomalous discrimination of the a-th sample. This represents the sparsity of the a-th feature of the i-th sample in the sample set corresponding to the first node. The first preset value is 1. and As respectively and The weight, When the time stationarity of the samples is high, the segmentation point is obtained first based on the degree of non-stationarity of the samples in time series. When the density is low, the segmentation point is obtained by prioritizing the density sparsity of the sample on the a-th feature coordinate axis.

[0074] Therefore, the segmentation probability of the a-th feature of each sample can be obtained in the sample set. Furthermore, the sample set can be segmented based on the sample with the highest segmentation probability to obtain two sub-sample sets.

[0075] Specifically, the value of the feature corresponding to the sample with the maximum segmentation probability of a feature in each sample in the sample set is used as the segmentation threshold; samples in the sample set whose feature value is greater than or equal to the segmentation threshold are grouped into a subset; samples in the sample set whose feature value is less than the segmentation threshold are grouped into a subset.

[0076] For example, when segmenting based on the a-th feature, the sample corresponding to the sample with the maximum segmentation probability of the a-th feature among all samples in the sample set has a value of the a-th feature. The segmentation threshold is Therefore, the a-th feature in the sample set is greater than or equal to The samples form a subset, and the subset with the a-th feature less than 0. The samples form a subset, completing the segmentation of the sample set. At this point, the depth of the isolated tree is increased by 1.

[0077] Similarly, for the two obtained subsample sets, a feature is randomly selected as the basis for segmentation according to the above method. Then, the segmentation threshold is obtained and segmentation is performed until the depth of the isolated tree reaches the preset depth of 8, thus completing the construction of the isolated tree. The acquisition of the segmentation threshold and the segmentation method are the same in the subsequent segmentation of the subsample sets corresponding to each node.

[0078] This allows the construction of multiple isolated trees. Based on the number of isolated trees commonly used in existing isolated forest algorithms, this embodiment generates a total of 100 isolated trees to form an isolated forest.

[0079] Finally, the structural fatigue anomaly level at the current moment is obtained based on the constructed isolated trees. Specifically, the average path length of the sample in all isolated trees of the isolated forest at the current moment is calculated and denoted as the average path length; the average path length is then substituted into the isolated forest anomaly scoring formula to obtain the structural fatigue anomaly level at the current moment. The methods for obtaining the path length of the sample in the isolated trees of the isolated forest and the isolated forest anomaly scoring formula are well-known techniques and will not be elaborated upon here.

[0080] After obtaining the current level of structural fatigue anomaly, the levels of structural fatigue anomaly at historical moments can be obtained. Box plots are then used to analyze the levels of structural fatigue anomaly at all historical moments to determine the normal range of structural fatigue anomaly. The upper limit of this normal range is used as the anomaly warning threshold. If the current level of structural fatigue anomaly is greater than or equal to the anomaly warning threshold, a structural fatigue anomaly warning for the tower crane is issued. This completes the performance testing of the tower crane.

[0081] In summary, this application utilizes the isolated forest algorithm to assess the degree of fatigue anomalies in tower crane structures, taking into account the significant differences between abnormal data and normal data, and the relatively small data volume. Specifically, during the generation of the isolated tree, the node segmentation position is adjusted at each node based on the degree of consistency between the data and the structural fatigue anomaly, thereby reducing the likelihood of normal operating condition changes being identified as structural fatigue anomalies. This approach enables accurate identification of fatigue anomalies in tower crane structures, improving construction safety and operation and maintenance efficiency.

[0082] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0083] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent performance detection of construction equipment based on AI algorithms, characterized in that, The method includes: The operation data of each monitoring point of the tower crane is collected, and the time period of the collected data is evenly divided into different time periods. Based on the operation data of each monitoring point in a time period, the characteristics of each monitoring point in that time period are obtained, and the characteristics of each monitoring point in that time period are combined into a sample. The time corresponding to the sample is the midpoint of that time period. The sample set of the first node of an isolated tree in an isolated forest is formed by a predetermined number of samples containing the current time sample; the sparsity of a feature of a sample is obtained based on the difference between a feature of a sample in the sample set and the feature of other samples; the non-stationarity of a feature of a sample is obtained by using the difference between a feature of a sample in the sample set and the feature of two adjacent time steps. The anomaly discrimination degree of a feature is obtained based on the standard deviation of a feature in each sample in the sample set; the segmentation probability of a feature in a sample is obtained by using the sparsity, non-stationarity, and anomaly discrimination degree of a feature; the sample set is segmented into two sub-sample sets based on the sample with the highest segmentation probability, and so on, until the depth of the isolated tree reaches the preset depth, thus completing the construction of the isolated tree; the structural fatigue anomaly degree at the current moment is obtained based on each completed isolated tree. The method of obtaining the segmentation probability of a feature of a sample by utilizing the sparsity, non-stationarity, and anomaly discriminative power of that feature includes: The probability of segmentation of a feature is obtained by weighting the non-stationarity and sparsity of the feature in a sample by taking the difference between the anomaly discrimination degree of a feature, the first preset value and the anomaly discrimination degree of the feature as weights, and summing the non-stationarity and sparsity of the feature in the sample.

2. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, The collected operational data from each monitoring point of the tower crane includes: The monitoring points on the tower crane include acceleration monitoring points, torque monitoring points, and strain monitoring points. The acceleration monitoring points include those at the slewing point, the base of the boom, the middle of the tower body, and the end of the boom. The torque monitoring point includes the position of the drum. The strain monitoring points include those at the base of the boom, the base of the tower body, and the middle section of the boom. Acceleration sensors, torque sensors, and strain gauges are installed at each acceleration monitoring point, torque sensors, and strain monitoring point to collect acceleration data from each acceleration monitoring point, torque data from each torque monitoring point, and strain data from each strain monitoring point, which serve as the operating data for each monitoring point.

3. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, The step of obtaining the characteristics of each monitoring point based on the operational data of each monitoring point within a time period includes: For an acceleration monitoring point, the average value and root mean square of each acceleration data point within a time period are obtained as the characteristics of that acceleration monitoring point within that time period. For torque monitoring points, the average value and standard deviation of each torque data point within a time period are obtained as the characteristics of the torque monitoring points within that time period. For a strain monitoring point, the mean strain and the sum of strain amplitudes of each strain data point within a time period are obtained as the characteristics of that strain monitoring point within that time period.

4. The intelligent performance detection method for construction equipment based on AI algorithm according to claim 1, characterized in that, The step of obtaining the sparsity of a feature of a sample based on the difference between a feature of a sample and the same feature of other samples in the sample set includes: The sparsity of a feature is obtained by summing the absolute differences between a feature of a sample and the features of other samples in the sample set, and then normalizing the sum.

5. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, The method of obtaining the non-stationarity of a feature of a sample by comparing it with the difference of that feature between samples at two adjacent time points includes: Arrange the samples in the sample set according to time sequence to obtain the sample sequence; obtain the absolute value of the difference between a feature of a sample in the sample sequence and the feature of the same feature of two adjacent time steps, and normalize the minimum value of the two absolute values ​​of the difference to obtain the non-stationarity of the feature of the sample.

6. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, The method of obtaining the anomaly discrimination degree of a feature based on the standard deviation of a feature in each sample in the sample set includes: The anomaly discrimination degree of a feature is obtained by taking the reciprocal of the sum of the standard deviation of a feature and the hyperparameters for each sample in the sample set and normalizing it.

7. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, The process of segmenting the sample set into two subsets based on the sample with the highest segmentation probability includes: The maximum value of a feature among the segmentation probabilities of each sample in the sample set is obtained, and the value of that feature is used as the segmentation threshold. Samples in the sample set whose value of that feature is greater than or equal to the segmentation threshold are grouped into a subset. Samples in the sample set whose value of that feature is less than the segmentation threshold are grouped into a subset.

8. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, The step of obtaining the current structural fatigue anomaly level based on each constructed isolated tree includes: Calculate the average path length of the sample in all isolated trees of the isolated forest at the current moment, and denote it as the average path length; substitute the average path length into the isolated forest anomaly scoring formula to obtain the structural fatigue anomaly degree at the current moment.

9. The intelligent detection method for construction equipment performance based on AI algorithm according to claim 1, characterized in that, After obtaining the structural fatigue anomaly level at the current moment based on each constructed isolated tree, the method further includes: By analyzing the structural fatigue anomaly at all historical moments using box plots, the normal range of structural fatigue anomaly is obtained. The upper limit of the normal range of structural fatigue anomaly is used as the anomaly warning threshold. If the structural fatigue anomaly at the current moment is greater than or equal to the anomaly warning threshold, a structural fatigue anomaly warning for the tower crane is issued.

Images