An intelligent recognition and diagnosis system for abnormal working conditions of a cutter suction dredger

Abstract

This invention provides an intelligent identification and diagnosis system for abnormal operating conditions of a cutter suction dredger. The system comprises a SCADA monitoring and data acquisition module that collects operating condition data monitored by the SCADA system and filters key variables; a data preprocessing module that normalizes and standardizes the key variable operating condition data using a conventional preprocessing module, followed by further preprocessing by a candlestick chart time-series feature data generation module based on candlestick chart definitions to obtain candlestick chart time-series feature data; an anomaly diagnosis and annotation module that performs anomaly diagnosis and annotation based on the candlestick chart time-series features, forming an abnormal operating condition database; and an XGBoost abnormal operating condition identification classifier that establishes and trains an XGBoost abnormal operating condition identification model, outputting the optimal XGBoost abnormal operating condition identification classifier. This invention can process large amounts of SCADA monitoring data faster and more accurately, improving the speed of time series processing, reducing the time for anomaly annotation, reducing the impact of single-point anomalies on prediction accuracy, and achieving intelligent identification of abnormal operating conditions.

Description

Technical Field

[0001] This invention relates to the field of safe production of cutter suction dredgers. Background Technology

[0002] Waterways, often referred to as "water highways," are vital arteries connecting major waterways and crucial resources for economic development. To ensure smooth water transport, it is essential to improve the quality and efficiency of dredging projects. Cutter suction dredgers are widely used in dredging projects, playing a significant role in improving navigation conditions and ensuring ship safety. However, as dredgers operate underwater, their complex and variable working conditions make them prone to malfunctions such as suction port blockage, pump blockage, and pipe blockage, severely impacting the continuity of construction and reducing efficiency.

[0003] Currently, the judgment of abnormal operating conditions of cutter suction dredgers mainly relies on the operator's personal experience, lacking scientific identification methods for abnormal conditions, which seriously affects the construction efficiency of cutter suction dredgers. Manual detection of abnormal operating conditions has the disadvantages of subjectivity and lag, which is not conducive to timely detection of abnormal conditions, and the identification results are often inaccurate due to human factors.

[0004] Given the actual situation of cutter suction dredgers, the identification of abnormal operating conditions presents several challenges:

[0005] (1) Single-point anomalies often occur in the timing anomaly detection of cutter suction dredgers. Single-point anomalies usually do not accompany abnormal working conditions. Abnormal working conditions are usually a group of anomalies that last for a period of time. However, point anomalies will affect the accuracy of anomaly identification and need to be identified and processed.

[0006] (2) Establishing an abnormal working condition identification model based on SCADA historical data requires a large amount of abnormal labeling information. However, abnormal labeling relies on the knowledge of experts, which is costly, time-consuming, and also has the problem of insufficient labeled data. Summary of the Invention

[0007] In view of the shortcomings of existing technologies for identifying abnormal working conditions and the need for improvement, this invention provides an intelligent identification and diagnosis method and system for abnormal working conditions of cutter suction dredgers. Its purpose is to solve the technical problems of low efficiency and inaccuracy when using existing manual detection methods to identify abnormal working conditions.

[0008] The technical solution of this invention is as follows:

[0009] A method for intelligent identification and diagnosis of abnormal operating conditions of a cutter suction dredger includes:

[0010] S1: Data collection and screening of key variables;

[0011] S2: Routine data preprocessing;

[0012] S3: Based on the definition of candlestick charts, further preprocessing is used to obtain candlestick chart time series feature data;

[0013] S4: Design an anomaly diagnosis algorithm based on the time series characteristics of candlestick charts, and annotate the construction conditions;

[0014] S5: Establish an abnormal operating condition database;

[0015] S6: Based on the abnormal working condition database, establish an XGBoost abnormal working condition identification model, and output the optimal XGBoost abnormal working condition identification classifier through continuous iterative training of the model.

[0016] Further optimizations include S7 and S8:

[0017] Step S7: Input the real-time collected construction data into the optimal classifier output by S6, and output the real-time construction condition information. If an anomaly occurs, an alarm is triggered, and the alarm anomaly condition data is stored in the anomaly condition database.

[0018] Step S8: Manually flag the alarm information issued in S7. If there are any errors, correct the data in the abnormal operating condition database and update the abnormal operating condition database in step S5.

[0019] Collect construction data from cutter suction dredgers and establish historical and real-time databases.

[0020] An intelligent identification and diagnosis system for abnormal operating conditions of a cutter suction dredger, characterized by:

[0021] It includes a SCADA monitoring and data acquisition module, which is used to collect operating condition data monitored by the SCADA system and filter key variables;

[0022] It includes a data preprocessing module, which includes a routine preprocessing module and a candlestick chart time series feature data generation module. The routine preprocessing module performs normalization and standardization preprocessing operations on the key variable operating condition data, and then the candlestick chart time series feature data generation module further preprocesses the data based on the candlestick chart definition to obtain the candlestick chart time series feature data.

[0023] It includes an anomaly diagnosis and annotation module, which performs anomaly diagnosis and construction condition annotation based on the time series characteristics of K-line charts, thereby forming an anomaly condition database;

[0024] This includes an XGBoost abnormal operating condition identification classifier, which builds an XGBoost abnormal operating condition identification model based on the current abnormal operating condition database, and trains the XGBoost abnormal operating condition identification model iteratively.

[0025] Output the optimal XGBoost abnormal condition identification classifier;

[0026] The construction data collected in real time by the SCADA monitoring and data acquisition system is input into the optimal XGBoost classifier, which outputs real-time identification of construction conditions. If an anomaly occurs, an alarm is triggered, and the alarm's abnormal condition data is stored in the abnormal condition database and processed as follows: the issued alarm information is manually labeled, and if there are errors, the data in the abnormal condition database is corrected to update the abnormal condition database.

[0027] To process large amounts of SCADA monitoring data more quickly and accurately, this invention applies K-line theory from the financial field to the algorithm design framework for identifying abnormal operating conditions on dredging vessels. By redefining the sampling period, it not only significantly improves the speed of time series processing and reduces the time for anomaly labeling, but also solves the problem of single-point anomalies affecting prediction accuracy. Finally, an abnormal operating condition identification model is built using the XGBoost algorithm to achieve intelligent identification of abnormal operating conditions. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the abnormal working condition identification algorithm framework of the present invention.

[0029] Figure 2 This is the definition of the candlestick chart in the algorithm of this invention.

[0030] Figure 3 This is a diagram of the candlestick chart conversion process in the algorithm of this invention (sampling period set to 60 seconds).

[0031] Figure 4 The original time-series graph (a) and the corresponding K-line graph (b) of the output are shown in the example of the algorithm of this invention.

[0032] Figure 5 This is a schematic diagram of the structural framework of the intelligent identification and diagnosis system for abnormal operating conditions of a cutter suction dredger according to the present invention. Detailed Implementation

[0033] The technical solutions provided in this application will be further described below with reference to specific embodiments and accompanying drawings. The advantages and features of this application will become clearer from the following description.

[0034] A method for intelligent identification and diagnosis of abnormal operating conditions of a cutter suction dredger, the entire algorithm is designed as follows:

[0035] Step S1: Data collection and filtering of key variables:

[0036] Collect construction data from cutter suction dredgers and establish historical and real-time databases;

[0037] The SCADA system on the cutter suction dredger collects more than 1,800 variables. In the algorithm design of this invention, key variables of the cutter suction dredger's construction operation system are selected for monitoring, as shown in Table 1.

[0038]

[0039]

[0040] Step S2: Data Preprocessing:

[0041] Perform normalization and standardization preprocessing on historical operating data;

[0042] Normalization preprocessing:

[0043] Where: X i Let X be the i-th variable in Table 1, min(X) i ) is X i The minimum value, max(X) i ) is X i The maximum value.

[0044] Step S3: Based on the candlestick chart definition, further preprocessing is performed to obtain the candlestick chart time series feature data.

[0045] Drawing K-line monitoring charts: For the key variables shown in Table 1, calculate their K variables based on K-line theory and draw K-line time series monitoring charts.

[0046] In the algorithm design of this invention, the candlestick chart is defined as follows: The candlestick chart is a bar-shaped line composed of shadows and a body. The portion of the shadow above the body is called the upper shadow, and the portion below is called the lower shadow. The upper shadow and lower shadow represent the maximum and minimum values ​​of the time sequence for that period, respectively. Figure 2 The candlestick chart definition shown indicates that when the ending value is higher than the starting value, it is represented by a bullish candlestick (red), and when the ending value is lower than the starting value, it is represented by a bearish candlestick (green). Figure 2 In this process, data outside the entity should not exceed 5%, which means that only sampling cases in which the number of entity samples can reach more than 95% of the total samples are selected; otherwise, they are discarded to ensure the reliability and quality of the samples entering the algorithm.

[0047] In the algorithm design of this invention, K-theory states that for X with a time length of k*T... i A time series of variables can be represented as: For X i The value at time k*T.

[0048] If we divide it into several adjacent time intervals of length T, then seq(X) i There are k time periods, and any time period can be used... Let j = 1, 2, 3, ..., k.

[0049]

[0050]

[0051] For any segment of time series data of length L Specifically, its K variable consists of four variables, namely the starting value C of this time series data. open Minimum value C min Maximum value C max End value C close

[0052]

[0053] For time sequence The minimum value, For time sequence The maximum value.

[0054] See Figure 3 As shown, transforming the original time-series monitoring chart with a sampling period of 1 second into a 60-second candlestick monitoring chart not only preserves key information from the original sequence but also reduces the impact of individual outliers, thus improving the efficiency of time-series data processing. See Figure 4 The image shows the original time-series chart of production output and the corresponding candlestick chart of production output.

[0055] Step S4: Design an anomaly diagnosis algorithm based on the time series characteristics of the candlestick chart and mark the construction conditions.

[0056] In the algorithm design of this invention, the candlestick chart represents different changes in the indicators, and the corresponding judgment criteria are shown in Table 2 below:

[0057]

[0058] The indicator for judging a large bullish candlestick is

[0059] The indicator for judging a large bearish candlestick is

[0060] The indicator for judging a bearish candlestick is

[0061] The indicator for judging a bullish candlestick is...

[0062] The indicator for judging a small negative line is...

[0063] The indicator for judging a small positive line is...

[0064] The indicator for judging a one-line line is C. close -C open =0,

[0065] Among them, C open Let C be the initial value of variable K. close C is the final value of variable K. max For the maximum value of variable K, C min It is the minimum value of variable K.

[0066] Four abnormal working conditions for foundations were identified: sludge suction port blockage, sludge pump blockage, sludge discharge pipe blockage, and excessive lateral movement speed. In application, additional abnormal working condition items can be added to further enhance construction stability.

[0067] The rules for determining abnormal operating conditions are designed as shown in Table 3:

[0068]

[0069]

[0070] The abnormal operating condition determination rules include:

[0071] Abnormal operating condition 1: S1 selects variable X 17 X 19 X 23 X 22 After pretreatment by S2 and S3, if the indicators are simultaneously judged to be "large positive line" for suction vacuum, "large negative line" for flow rate, "large negative line" for output, and "large negative line" for underwater pump discharge pressure, that is, when all four conditions occur at the same time, it is diagnosed that an abnormality has occurred at the sludge inlet, and it is determined that the sludge inlet of the underwater pump is blocked.

[0072] Abnormal operating condition 2: S1 selects variable X 17 X 19 X 22 After S2 and S3 pretreatment, if the indicators are simultaneously judged to be "large negative line" for suction vacuum, "large negative line" for flow rate, and "large negative line" for underwater pump discharge pressure, that is, when all three conditions occur at the same time, it is diagnosed as an abnormality in the mud pump and identified as mud pump blockage.

[0073] Abnormal operating condition 3: S1 selection variable X 17 X 19 If, after S2 and S3 pretreatment, the indicators are simultaneously judged to be "large negative line" of suction vacuum and "large negative line" of flow rate, that is, when both situations occur at the same time, it is diagnosed as an abnormality in the sludge discharge pipe and judged as sludge discharge pipe blockage.

[0074] Abnormal operating condition 4: S1 selects variable X 22 X 24 If, after preprocessing with S2 and S3, the indicators are simultaneously judged to be "large negative line" for flow velocity and "large positive line" for concentration, that is, when both situations occur at the same time, the diagnosis is that the transverse velocity is too large.

[0075] Abnormal operating condition 0: Otherwise, the operating condition of the cutter suction dredger is judged as normal.

[0076] For variable X i The corresponding time series data seq(X) i The time interval is uniformly divided into several adjacent time intervals of length T, with intervals shorter than T seconds being discarded. i = 1, 2, 3... 24, corresponding one-to-one with the variables in Table 1. For time series data seq(X) i The kth time period of ).

[0077] According to the abnormal operating condition judgment rules formulated in Table 3, the abnormal operating condition category y of the dredging device for any time period is determined. k Label it, i.e., y k =t, where t∈{0,1,2,3,4}, representing normal, sludge inlet blockage, sludge pump blockage, sludge discharge pipe blockage, and excessive lateral movement speed, respectively. k represents the k-th time period, and the abnormal operating conditions at all times within this time period are set to t. This operation can greatly improve the speed of time series abnormal operating condition judgment and labeling. The time length should be set reasonably; setting it too large reduces the judgment accuracy, while setting it too small consumes too much manpower. This invention sets it to 60 seconds.

[0078] Step S5: Establish an abnormal operating condition database: Input the abnormal operating condition data diagnosed in S4 into the abnormal operating condition database, and randomly select an equal amount of normal operating condition data to input into the abnormal operating condition database to maintain a balance between the proportion of abnormal and normal data.

[0079] Step S6: Based on the abnormal working condition database, establish an XGBoost abnormal working condition identification model, and output the optimal XGBoost classifier by continuously iterating and training the model.

[0080] Input: 70% of the samples are randomly selected as the training dataset, and the training sample dataset D = {(x i ,y i )},

[0081] The input variables are those listed in Table 1, and the output variable is the abnormal operating condition category.

[0082] x j =(x j1 x j2 ...x j24 ), x i Let x be the j-th sample value of the input variable. ji For the i-th variable X i The j-th sample value,

[0083] y i Output the j-th sample value of the variable.

[0084] Output: The optimal XGBoost tree structure, i.e., the optimal classifier.

[0085] The XGBoost tree model is represented as:

[0086]

[0087] in Let x be the predicted value of the output variable for the i-th sample. i For the i-th input sample, f k For the k-th classification tree

[0088] The objective function of XGBoost is:

[0089]

[0090] The objective function (2) consists of two parts. The first part is the loss function l, which represents the difference between the true value and the predicted value, and n is the sample size. The second part is the regularization term, which is used to prevent the model from overfitting.

[0091]

[0092] γ is the regularization parameter for the number of leaf nodes, which suppresses further splitting of nodes; λ is the regularization parameter for the weight of leaf nodes, which prevents the weight of leaf nodes from becoming too large; ω j The score for each leaf node is given. T is the number of leaf nodes.

[0093] The XGBoost algorithm adds only one tree at a time, and through continuous iterative training, it stops updating after finding the set of parameters that minimizes the objective function, and outputs the optimal model structure.

[0094] The remaining 30% of samples are used as the test dataset and input into the optimal tree structure of XGBoost to obtain the predicted classification. Classify the prediction Compared with the measured classification y i The performance is evaluated by comparing the results and calculating the accuracy of the classifiers.

[0095]

[0096] Wherein, TP (True positives): the number of instances that are correctly classified as positive, that is, the number of instances that are actually positive and are classified as positive by the classifier;

[0097] FP (False positives): The number of instances that were incorrectly classified as positive, i.e., the number of instances that were actually negative but were classified as positive by the classifier;

[0098] FN (False negatives): The number of instances that were incorrectly classified as negative, i.e., the number of instances that were actually positive but were classified as negative by the classifier;

[0099] TN (True negatives): The number of instances that are correctly classified as negative, i.e., the number of instances that are actually negative and were classified as negative by the classifier.

[0100] After simulation, the model prediction accuracy is shown in Table 4 below:

[0101]

[0102] Step S7: Input the real-time collected construction data into the optimal classifier output by S6, and output the real-time construction condition information. If an anomaly occurs, an alarm is triggered, and the alarm's abnormal condition data is stored in the abnormal condition database.

[0103] Step S8: Manually flag the alarm information issued in S7. If there are any errors, correct the data in the abnormal operating condition database and update the abnormal operating condition database in step S5.

[0104] Example

[0105] Based on the above algorithm design, the embodiment further provides an intelligent identification and diagnosis system for abnormal operating conditions of a cutter suction dredger, such as... Figure 5 As shown:

[0106] It includes a SCADA monitoring and data acquisition module, which is used to collect operating condition data monitored by the SCADA system and filter key variables;

[0107] It includes a data preprocessing module, which includes a routine preprocessing module and a candlestick chart time series feature data generation module. The routine preprocessing module performs normalization and standardization preprocessing operations on the key variable operating condition data, and then the candlestick chart time series feature data generation module further preprocesses the data based on the candlestick chart definition to obtain the candlestick chart time series feature data.

[0108] It includes an anomaly diagnosis and annotation module, which performs anomaly diagnosis and construction condition annotation based on the time series characteristics of K-line charts, thereby forming an anomaly condition database;

[0109] This includes an XGBoost abnormal operating condition identification classifier, which builds an XGBoost abnormal operating condition identification model based on the current abnormal operating condition database, and trains the XGBoost abnormal operating condition identification model iteratively.

[0110] Output the optimal XGBoost abnormal condition identification classifier;

[0111] The construction data collected in real time by the SCADA monitoring and data acquisition system is input into the optimal XGBoost classifier, which outputs real-time identification of construction conditions. If an anomaly occurs, an alarm is triggered, and the alarm's abnormal condition data is stored in the abnormal condition database and processed as follows: the issued alarm information is manually labeled, and if there are errors, the data in the abnormal condition database is corrected to update the abnormal condition database.

[0112] Furthermore, in the SCADA monitoring and data acquisition system (1), wherein:

[0113] The data collected is the operating data monitored by the SCADA system. Specifically, it involves collecting construction data from the cutter suction dredger and establishing historical and real-time databases.

[0114] Furthermore, in the SCADA monitoring and data acquisition system (1), wherein:

[0115] Selection of key variables:

[0116] Includes TIDE x1, DRAFT_TRNION x2, and BRIDGE_ANGLE x3, which belong to the cable tray system;

[0117] This includes lateral movement speed (TRNS_SPD) x4, wire rope tension (ROPE_TENSION) x5, lateral movement angle (TRNS_ANGLE) x6, lateral movement rope speed (TRNS_ROPE_SPD) x7, and lateral movement hydraulic system pressure (TH_PR) x8, which belongs to the lateral movement system.

[0118] Includes trolley travel (CR_STRK) x9, double pile landing (DOUBLE_PILE_LANDING) x 10 It belongs to the steel pile trolley system;

[0119] Including cutter speed (CUTTER_SPD) X 11 Cutter depth (CUTTER_DEEP) X 12 Cutter power (CUTTER_POWER) X 13 Cutter hydraulic system pressure (CUTTER_CH_PR) X 14 It belongs to the reamer system;

[0120] Including underwater pump discharge pressure (LP_OUTLET_PR)X 15 Submersible pump power (LP_POWER) X 16 Suction vacuum (SUCK_VAC)X 17 It belongs to the underwater pump system;

[0121] Including mud pump speed (DP_SPD) X 18 Sludge pump discharge pressure (DP_OUTLET_SPD) X 19 Mud pump power (DP_POWER) X 20 It belongs to the mud pump system;

[0122] Including density (SLURRY_SPD)X 21 , flow rate (SLURRY_FLOW)X 22 Instantaneous output (OUTPUT) X 23 Concentration (SLURRY_CON)X 24 It belongs to the production measurement system.

[0123] Furthermore, in the preprocessing module, wherein:

[0124] For the selected key variables, normalization and standardization preprocessing operations are performed on these historical operating condition data.

[0125] Normalization preprocessing:

[0126] Where: X i Let X be the i-th variable in Table 1, min(X) i ) is X i The minimum value, max(X) i ) is X i The maximum value.

[0127] Furthermore, in the preprocessing module:

[0128] The K-line chart time series feature data generation module (3) calculates the K variable based on K-line theory for the data obtained by the conventional preprocessing module (2), draws the K-line time series monitoring chart, and transforms the original time series monitoring chart with a sampling period of 1 second into a K-line monitoring chart with a time length of T seconds, thereby obtaining K-line chart time series feature data, namely the original time series chart of production and the corresponding production K-line chart;

[0129] In the K-line time series monitoring chart, the K-line chart is a bar-shaped line composed of shadows and a body. The part of the shadow above the body is called the upper shadow, and the part below is called the lower shadow. The upper shadow and lower shadow represent the maximum and minimum values ​​of the time series, respectively. A higher ending value than the starting value is represented by a bullish K-line, and a lower ending value than the starting value is represented by a bearish K-line. The K-line theory states that for a time series of length k*T... i A time series of variables is represented as: For X i The value at time k*T;

[0130] If we divide it into several adjacent time intervals of length T seconds, then seq(X) i There are k time periods, and any time period is used... Let j = 1, 2, 3, ..., k.

[0131]

[0132]

[0133] For any segment of time series data of length L Specifically, its K variable consists of four variables, namely the starting value C of this time series data. open Minimum value C min Maximum value C max End value C close ,

[0134]

[0135] For time sequence The minimum value, For time sequence The maximum value.

[0136] Furthermore, regarding variable X i The corresponding time series data seq(X) i The time interval is uniformly divided into several adjacent time intervals of length T, and time intervals shorter than T seconds are discarded; i = 1, 2, 3... 24, corresponding one-to-one with the key variables; For time series data seq(X) i The kth time period of ).

[0137] Furthermore, the anomaly diagnosis module (4) includes and operates:

[0138] The software module (4.1) provides different judgment criteria for candlestick chart indicators, specifically including: the judgment indicator for a large bullish candlestick. The indicator for judging a large bearish candlestick is The indicator for judging a bearish candlestick is The indicator for judging a bullish candlestick is... The indicator for judging a small negative line is... The indicator for judging a small positive line is... The indicator for judging a one-line line is C. close -C open =0, where C open Let C be the initial value of variable K. close C is the final value of variable K. max For the maximum value of variable K, C min The minimum value of variable K;

[0139] The software module (4.2) includes four basic abnormal operating condition judgment rules: sludge suction port blockage, sludge pump blockage, sludge discharge pipe blockage, and excessive lateral movement speed. The specific abnormal operating condition judgment rules defined in this software module include:

[0140] Abnormal operating condition 1: S1 selects variable X 17 X 19 X 23 X 22 After pretreatment by S2 and S3, if the indicators are simultaneously judged to be "large positive line" for suction vacuum, "large negative line" for flow rate, "large negative line" for output, and "large negative line" for underwater pump discharge pressure, that is, when all four conditions occur at the same time, it is diagnosed that an abnormality has occurred at the sludge inlet, and it is determined that the sludge inlet of the underwater pump is blocked.

[0141] Abnormal operating condition 2: S1 selects variable X 17 X 19 X 22 After S2 and S3 pretreatment, if the indicators are simultaneously judged to be "large negative line" for suction vacuum, "large negative line" for flow rate, and "large negative line" for underwater pump discharge pressure, that is, when all three conditions occur at the same time, it is diagnosed as an abnormality in the mud pump and identified as mud pump blockage.

[0142] Abnormal operating condition 3: S1 selection variable X 17 X 19 If, after S2 and S3 pretreatment, the indicators are simultaneously judged to be "large negative line" of suction vacuum and "large negative line" of flow rate, that is, when both situations occur at the same time, it is diagnosed as an abnormality in the sludge discharge pipe and judged as sludge discharge pipe blockage.

[0143] Abnormal operating condition 4: S1 selects variable X 22 X 24 If, after preprocessing with S2 and S3, the indicators are simultaneously judged to be "large negative line" for flow velocity and "large positive line" for concentration, that is, when both situations occur at the same time, the diagnosis is that the transverse velocity is too large.

[0144] Abnormal operating condition 0: Otherwise, the operating condition of the cutter suction dredger is judged as normal;

[0145] The abnormal operating condition category labeling software module (4.3) classifies the abnormal operating condition y of the dredging device for any given time period according to the abnormal operating condition judgment rules formulated by the abnormal operating condition judgment rule software module (4.2). k Label it, i.e., y k =t, where t∈{0,1,2,3,4}, representing normal, sludge inlet blockage, sludge pump blockage, sludge discharge pipe blockage, and excessive lateral movement speed, respectively. k represents the kth time period, and the abnormal working conditions at all times within this time period are set to t; output and form an abnormal working condition database (5).

[0146] Furthermore, the abnormal operating condition database (5):

[0147] The abnormal operating condition data diagnosed in the abnormal operating condition category labeling software module (4.3) is input into the abnormal operating condition database (5), and an equal amount of normal operating condition data is randomly selected and input into the abnormal operating condition database (5) to maintain the balance of the proportion of abnormal and normal data.

[0148] Furthermore, the XGBoost abnormal operating condition identification classifier (6) is specifically as follows:

[0149] Input: A randomly selected subset of samples is used as the training dataset. The training sample dataset D = {(x i ,y i )},

[0150] Output: The optimal XGBoost tree structure, which is the optimal classifier.

[0151] Furthermore, the input variables are the selected key variables, and the output variable is the abnormal operating condition category:

[0152] x j =(x j1 x j2 ...x j18 ), x i Let x be the j-th sample value of the input variable. ji For the i-th variable X i The j-th sample value,

[0153] y i Output the j-th sample value of the variable;

[0154] The XGBoost tree model is represented as:

[0155]

[0156] in Let x be the predicted value of the output variable for the i-th sample. i For the i-th input sample, f k This is the k-th classification tree;

[0157] The objective function of XGBoost is:

[0158]

[0159] The objective function (Formula 2) consists of two parts: the first part is the loss function l, which represents the difference between the true value and the predicted value, and n is the sample size; the second part is the regularization term, which is used to prevent the model from overfitting.

[0160]

[0161] γ is the regularization parameter for the number of leaf nodes, which suppresses further splitting of nodes; λ is the regularization parameter for the weight of leaf nodes, which prevents the weight of leaf nodes from becoming too large; ω j The score for each leaf node; T is the number of leaf nodes;

[0162] The XGBoost algorithm adds only one tree at a time, and through continuous iterative training, it stops updating after finding the set of parameters that minimizes the objective function and outputs the optimal model structure.

[0163] The remaining samples are used as the test dataset and input into the optimal tree structure of XGBoost to obtain the predicted classification. Classify the prediction Compared with the measured classification y i The performance is evaluated by comparing the results and calculating the accuracy of the classifiers.

[0164] The above technical solution employs a semi-supervised learning approach, enabling accurate identification of abnormal working condition categories. Typically, abnormal working condition category labels require manual annotation. This invention, by applying K-theory combined with expert knowledge, automatically labels construction data for abnormalities, solving the problem of high manpower costs associated with manual annotation. Simultaneously, it establishes an abnormal working condition identification model based on the XGBoost algorithm, capable of identifying abnormal working conditions in unlabeled real-time data. Through a manual feedback mechanism, the abnormal working condition category labels are continuously corrected, thereby continuously optimizing the abnormal working condition identification model. The identification framework of this invention is scalable; new abnormal working condition categories can be identified simply by adding new expert experience.

[0165] The above description is merely a description of preferred embodiments of this application and is not intended to limit the scope of this application in any way. Any changes or modifications made by those skilled in the art based on the above-disclosed technical content should be considered as equivalent and valid embodiments and fall within the scope of protection of the technical solution of this application.

Claims

1. An intelligent identification and diagnosis system for abnormal operating conditions of a cutter suction dredger, characterized in that: It includes a SCADA monitoring and data acquisition module (1), which is used to collect the operating data monitored by the SCADA system and filter the key variables therein; It includes a data preprocessing module, including a conventional preprocessing module (2) and a K-line chart time series feature data generation module (3); the key variable operating condition data is normalized and standardized by the conventional preprocessing module (2), and then the K-line chart time series feature data is further preprocessed by the K-line chart time series feature data generation module (3) based on the K-line chart definition to obtain the K-line chart time series feature data; It includes an anomaly diagnosis and annotation module (4), which performs anomaly diagnosis and construction condition annotation based on the time series characteristics of K-line chart, thereby forming an anomaly condition database (5). Including the XGBoost abnormal working condition identification classifier (6), the XGBoost abnormal working condition identification model is established based on the current abnormal working condition database (5), and the optimal XGBoost abnormal working condition identification classifier is output by iteratively training the XGBoost abnormal working condition identification model. The construction data collected in real time by the SCADA monitoring and data acquisition system (1) is input into the optimal XGBoost classifier, and the real-time identification of construction conditions is output. If an abnormality occurs, an alarm is triggered, and the abnormal condition data of the alarm is stored in the abnormal condition database (5) and processed as follows: the alarm information is manually labeled, and if there is an error, the data in the abnormal condition database (5) is corrected to update the abnormal condition database (5). The K-line chart time series feature data generation module (3) calculates the K variable based on K-line theory for the data obtained by the conventional preprocessing module (2), draws the K-line time series monitoring chart, and transforms the original time series monitoring chart with a sampling period of 1 second into a K-line monitoring chart with a time length of T seconds, thereby obtaining the K-line chart time series feature data, namely the original time series chart of production and the corresponding production K-line chart; In the K-line time series monitoring chart, the K-line chart is a bar-shaped line composed of shadows and a body. The portion of the shadow above the body is called the upper shadow, and the portion below is called the lower shadow. The upper and lower shadows represent the maximum and minimum values ​​of the K-line time series, respectively. An ending value higher than the starting value is represented by a bullish K-line, and an ending value lower than the starting value is represented by a bearish K-line. The K-line theory states that for a time length of... of A time series of variables is represented as: , for exist The value at time; dividing it into a number of adjacent time periods of a time length seconds, then there are time periods, any one time period being represented by where then For any segment of time series data of length L In terms of K variables, it consists of four variables, namely the start value , the minimum value , the maximum value , and the end value , For time sequence The minimum value, For time sequence The maximum value.

2. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger as claimed in claim 1, characterized in that, The SCADA monitoring and data acquisition system (1) includes: The data collected is the operating data monitored by the SCADA system. Specifically, it involves collecting construction data from the cutter suction dredger and establishing historical and real-time databases.

3. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger as claimed in claim 1, characterized in that, The SCADA monitoring and data acquisition system (1) includes: Selection of key variables: Includes tide level (TIDE) Ear shaft water intake (DRAFT_TRNION) Cable tray angle (BRIDGE_ANGLE) It belongs to the cable tray system; Including lateral speed (TRNS_SPD) Wire rope tension (ROPE_TENSION) Lateral shift angle (TRNS_ANGLE) Lateral rope speed (TRNS_ROPE_SPD) Horizontal hydraulic system pressure (TH_PR) It belongs to the transverse displacement system; including trolley stroke (CR_STRK) , double pile landing (DOUBLE_PILE_LANDING) , belonging to the steel pile trolley system; Including cutter speed (CUTTER_SPD) Cutter depth (CUTTER_DEEP) Cutter power (CUTTER_POWER) Cutter hydraulic system pressure (CUTTER_CH_PR) It belongs to the reamer system; Including underwater pump discharge pressure (LP_OUTLET_PR) Submersible pump power (LP_POWER) Suction vacuum (SUCK_VAC) It belongs to the underwater pump system; Including mud pump speed (DP_SPD) Sludge pump discharge pressure (DP_OUTLET_SPD) Mud pump power (DP_POWER) It belongs to the mud pump system; Including density (SLURRY_SPD) , flow rate (SLURRY_FLOW) Instantaneous output (OUTPUT) Concentration (SLURRY_CON) It belongs to the production measurement system.

4. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger of claim 1, characterized in that, The preprocessing module, wherein: For the selected key variables, normalization and standardization preprocessing operations are performed on these historical operating condition data. Normalization preprocessing: wherein: is the i-th variable in Table 1, is the minimum value of is the maximum value of .​ 5. The intelligent identification and diagnosis system for abnormal operating conditions of a cutter suction dredger as described in claim 1, characterized in that, For variables Corresponding time series data Divide the data evenly into several adjacent time intervals of length T, discarding intervals shorter than T seconds; Each variable corresponds one-to-one with a key variable; For time series data The kth time period.

6. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger of claim 1, characterized in that, The anomaly diagnosis module (4) includes and operates: The software module (4.1) provides judgment criteria for different indicators in candlestick charts, specifically including: the judgment indicator for a large bullish candlestick. The indicator for judging a large bearish candlestick is The indicator for judging a bearish candlestick is... The indicator for judging a bullish candlestick is... The indicator for judging a small negative line is The indicator for judging a small positive line is The indicator for judging a one-line line is ,in, Let K be the initial value. This represents the final value of the variable K. For the maximum value of variable K, The minimum value of variable K; The software module (4.2) includes four basic abnormal operating condition judgment rules: sludge suction port blockage, sludge pump blockage, sludge discharge pipe blockage, and excessive lateral movement speed. The specific abnormal operating condition judgment rules defined in this software module include: Abnormal operating condition 1: S1 selects variable X 17 X 19 X 23 X 22 After S2 and S3 pretreatment, when the indicators are simultaneously judged to be "large positive line" for suction vacuum, "large negative line" for flow rate, "large negative line" for output, and "large negative line" for underwater pump discharge pressure, that is, when all four conditions occur at the same time, it is diagnosed that an abnormality has occurred at the sludge inlet, and it is determined that the sludge inlet of the underwater pump is blocked. Abnormal operating condition 2: S1 Select variable X 17 X 19 X 22 After S2 and S3 pretreatment, if the indicators are simultaneously judged to be "large negative line" for suction vacuum, "large negative line" for flow velocity, and "large negative line" for underwater pump discharge pressure, that is, when all three conditions occur at the same time, it is diagnosed as an abnormality in the mud pump and is judged to be mud pump blockage. Abnormal operating condition 3: S1 selection variable X 17 X 19 If, after pretreatment by S2 and S3, the indicators are simultaneously judged to be "large negative line" for suction vacuum and "large negative line" for flow rate, that is, when both situations occur at the same time, it is diagnosed as an abnormality in the sludge discharge pipe and determined to be a blockage in the sludge discharge pipe. Abnormal operating condition 4: S1 selects variable X 22 X 24 If, after preprocessing with S2 and S3, the indicators are simultaneously judged to be "large negative line" for flow velocity and "large positive line" for concentration, that is, when both situations occur at the same time, the diagnosis is that the transverse velocity is too large. Abnormal operating condition 0: Otherwise, the operating condition of the cutter suction dredger is judged as normal; The abnormal operating condition category labeling software module (4.3) classifies the abnormal operating conditions of the dredging device for any given time period according to the abnormal operating condition judgment rules formulated by the abnormal operating condition judgment rule software module (4.2). To label, that is ,in These represent normal operation, sludge suction port blockage, sludge pump blockage, sludge discharge pipe blockage, and excessive lateral movement speed, respectively. Let k represent the k-th time period, and set the abnormal operating conditions at all times within this time period to 0. Output and form an abnormal operating condition database (5).

7. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger of claim 1, characterized in that, The abnormal operating condition database (5): The abnormal operating condition data diagnosed in the abnormal operating condition category labeling software module (4.3) is input into the abnormal operating condition database (5), and an equal amount of normal operating condition data is randomly selected and input into the abnormal operating condition database (5) to maintain the balance of the proportion of abnormal and normal data.

8. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger of claim 1, characterized in that, The XGBoost abnormal operating condition identification classifier (6) is specifically as follows: Input: Randomly extract a part of the samples as the training data set, and train the sample data set D , Output: The optimal XGBoost tree structure, which is the optimal classifier.

9. The intelligent abnormal condition identification and diagnosis system for a cutter suction dredger of claim 8, characterized in that, The input variables are the selected key variables, and the output variable is the abnormal operating condition category: , Let j be the j-th sample value of the input variable. For the i-th variable The j-th sample value, Output the j-th sample value of the variable; The XGBoost tree model is represented as: (Official 1) in For the predicted value of the output variable of the i-th sample, For the i-th input sample, This is the k-th classification tree; The objective function of XGBoost is: (Official 2) The objective function (Formula 2) consists of two parts, the first part being the loss function. The first part represents the difference between the true value and the predicted value, where n is the sample size; the second part is the regularization term, used to prevent the model from overfitting. (Official 3) This is a regularization parameter for the number of leaf nodes, used to suppress further node splitting. This is a regularization parameter for the weights of leaf nodes, used to prevent leaf node weights from becoming too large. The score for each leaf node; T is the number of leaf nodes; The XGBoost algorithm adds only one tree at a time, and through continuous iterative training, it stops updating after finding the set of parameters that minimizes the objective function and outputs the optimal model structure. The remaining samples are used as the test dataset and input into the optimal tree structure of XGBoost to obtain the predicted classification. , predict classification Compared with measured classification The performance is evaluated by comparing the results and calculating the accuracy of the classifiers.

Images