A ship power operation and maintenance system and operation and maintenance method

Abstract

This invention discloses a ship power operation and maintenance system, characterized by comprising: a parameter setting module, a real-time data acquisition module, a fault analysis module, a ship historical big data module, a fault tree optimization module, and a data terminal module; the fault analysis module is connected to the parameter setting module, the real-time data acquisition module, the fault tree optimization module, and the data terminal module respectively; the ship historical big data module is connected to the fault tree optimization module; the fault analysis module analyzes ship faults using parameters set by the parameter setting module and a fault tree model constructed by the ship historical big data module; the fault tree optimization module optimizes the fault tree model constructed by the fault analysis module using historical data from the ship historical big data module. This invention improves efficiency by performing correlation coefficient analysis on attributes, merging closely related attributes, increasing computational speed, reducing branching and judgment counts, and thus improving efficiency.

Description

Technical Field

[0001] This invention belongs to the field of ship operation and maintenance technology, specifically relating to a ship power operation and maintenance system and method. Background Technology

[0002] As the most important component of marine machinery, the marine propulsion system is a combination of major mechanical equipment for the generation, transmission, and consumption of various forms of energy. Supported by science and industrial technology, modern ships are becoming increasingly larger, and marine propulsion systems are also developing towards larger scale, integration, complexity, and intelligence. Modern marine propulsion systems are increasingly employing advanced technologies, leading to a surge in system size, structure, and technological complexity, making them highly susceptible to malfunctions during operation. The complexity and difficulty of the technology and equipment required for fault repair have increased dramatically, resulting in significantly higher maintenance costs. Existing maintenance services can no longer meet the growing maintenance requirements of modern ships. Poor maintenance and untimely fault handling are significant causes of ship accidents. In the operation and management of engineering vessels, especially large transport ships, dynamic daily management of the vessel is crucial to accurately grasp information such as the vessel's dynamics and the operating status of its machinery. However, relying solely on manual maintenance inevitably leads to errors, posing a risk of economic losses to enterprises or users and causing numerous inconveniences to ship operation and management.

[0003] In view of the above problems, establishing a ship power operation and maintenance system and method to ensure remote operation and maintenance management of ship power systems, ensure the supply of spare parts and components, and enable the monitoring of the operation and maintenance management of ship power equipment has become a key issue that still needs to be addressed in the modernization of ship operation and maintenance support. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a ship power operation and maintenance system and method to solve the technical problems of inaccurate understanding of ship operation and maintenance status and the inability to obtain accurate operation and maintenance methods in existing technologies.

[0005] This invention provides a ship power operation and maintenance system, including: a parameter setting module, a real-time data acquisition module, a fault analysis module, a ship historical big data module, a fault tree optimization module, and a data terminal module;

[0006] The fault analysis module is connected to the parameter setting module, the real-time data acquisition module, the fault tree optimization module, and the data terminal module, respectively; the ship historical big data module is connected to the fault tree optimization module.

[0007] The fault analysis module analyzes ship faults based on the parameters set by the parameter setting module and the fault tree model constructed by the ship historical big data module, and the real-time data collected by the real-time data acquisition module.

[0008] The data terminal module generates alarm signals and operation and maintenance decisions based on the ship fault analysis results.

[0009] The fault tree optimization module optimizes the fault tree model constructed by the fault analysis module using historical data from the ship historical big data module.

[0010] Furthermore, the parameter setting module includes setting basic equipment parameters and setting shore-based parameters; wherein, the basic equipment parameters include a system knowledge base, equipment fault database, and design data provided by the equipment manufacturer; and the shore-based parameters include usage scenario data, reliability data, and maintenance data of the ship's power system provided by the shore.

[0011] Furthermore, the fault analysis module uses the S4000P standard to construct a fault tree model to analyze ship faults.

[0012] Furthermore, the fault tree optimization module constructs a reference fault tree model using historical data from the ship historical big data module, compares the fault model constructed by the fault analysis module with the reference fault tree model, and optimizes the fault model by adding and / or deleting branches.

[0013] The technical solution of this invention is a method for ship power operation and maintenance, the specific method of which is as follows:

[0014] Step 1: Obtain historical fault data, perform data preprocessing, and obtain training set and test set. The training set and test set include diesel engine status attributes.

[0015] Step 2: Randomly extract samples from the training set and test set with replacement to form several sub-training sets and several sub-test sets. Fill in the missing diesel engine state attributes in the sub-training sets and discretize the continuous diesel engine state attributes; fill in the missing diesel engine state attributes in the sub-test sets and discretize the continuous diesel engine state attributes.

[0016] Step 3: Merge similar attributes in the diesel engine state attributes of the sub-training set;

[0017] Step 4: Determine the maximum depth threshold of the fault tree based on the correlation of each attribute in the diesel engine state attributes, and then construct the fault tree using the information gain method based on the maximum depth threshold and the training set;

[0018] Step 5: Prune the constructed fault tree and test the fault tree with the test set;

[0019] Step 6: Input the real-time diesel engine status into the fault tree to diagnose the fault.

[0020] Furthermore, the method for determining the maximum depth threshold in step 4 is as follows:

[0021] When the number of current attributes is less than 5, no maximum depth threshold is applied; when the number of current attributes is greater than or equal to 5, the fault tree threshold is [value missing].

[0022] Where m represents the number of attributes.

[0023] Furthermore, in step 4, when constructing the fault tree, if a new branch needs to be generated, the splitting factor in the information gain rate is first optimized. When the difference in information gain before and after the split meets the preset conditions, a new branch is formed by splitting. When the difference in information gain before and after the split does not meet the preset conditions, the splitting factor in the information gain rate is repeatedly optimized according to the difference in information gain before and after the split until the preset conditions are met, and a new branch is formed by splitting.

[0024] Furthermore, when optimizing the splitting factor term in the information gain ratio, the formula for obtaining the splitting factor term is as follows:

[0025]

[0026] Among them, S k The number of sample data for the current k-th attribute is represented by β; S is the number of samples in the dataset; β k The probability average coefficient, (0<β k <1), where p k p represents the probability of the current attribute. max p represents the maximum probability of the attribute. min This represents the minimum probability of the attribute.

[0027] Furthermore, when repeatedly optimizing the split factor term in the information gain ratio based on the difference in information gain before and after the split, the formula for obtaining the split factor term is as follows:

[0028]

[0029] Among them, S k The number of sample data for the current k-th attribute is represented by β; S is the number of samples in the dataset; β k The probability average coefficient, (0<β k <1), where p k p represents the probability of the current attribute. max p represents the maximum probability of the attribute. min α represents the minimum probability of the attribute; α is the difference between the information gain value before and after the split.

[0030] Furthermore, the preset condition is that the difference between the information gain value before the split and the information gain value after the split is less than a preset value.

[0031] Furthermore, the preset value is obtained by using the squared value of the Pearson correlation attribute coefficient r. 2 To determine the preset value, the value range is between (0, 1).

[0032] Furthermore, the preset value is 0.05.

[0033] The beneficial effects of this invention are:

[0034] This invention explores the relationships between attributes and between attributes and targets by performing correlation coefficient analysis on attributes. Merging closely related attributes increases computation speed and reduces branching and judgment counts, thus reducing the size of the fault tree. Before testing the test set, continuous attribute processing and missing attribute handling help reduce misjudgments and decreased accuracy caused by an incomplete test set. Optimizations to the fault tree algorithm specifically include introducing a probability average coefficient into the splitting factor term to mitigate the bias of information gain when the number of values ​​is large or concentrated. Setting a maximum depth threshold limits the branch depth of the fault tree, reducing the impact of invalid and empty branches. By comparing whether the information gain before and after branching meets a preset value, the branch's condition is determined, addressing situations where splitting effects are unsatisfactory or inaccurate, thereby improving the accuracy of fault tree judgment. Attached Figure Description

[0035] The features and advantages of the invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the drawings:

[0036] Figure 1 This is a system architecture diagram of a specific embodiment of the present invention;

[0037] Figure 2 This is a logical flowchart of the system decision-making method according to a specific embodiment of the present invention;

[0038] Figure 3 This is a system function diagram of a specific embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] The present invention will be further illustrated below with reference to specific embodiments. Those skilled in the art should understand that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Modifications to the present invention in various equivalent forms all fall within the scope defined by the appended claims.

[0041] In the description of this invention, it should be noted that the fault tree feature attributes mentioned include thermal parameters such as burst pressure, compression pressure, exhaust temperature, exhaust pressure, speed, and power, which reflect the working conditions and operating status of each component; vibration signals such as overall engine vibration, local surface vibration, and shaft torsional vibration can effectively diagnose the fuel system, valve train, piston-cylinder liner, connecting rod bearings and sliding main bearings, turbocharger, etc.; noise signals and other signals.

[0042] like Figure 1 As shown, the present invention provides a ship power operation and maintenance system, including: a parameter setting module 1, a real-time data acquisition module 6, a fault analysis module 2, a ship historical big data module 3, a fault tree optimization module 4, and a data terminal module 5.

[0043] Fault analysis module 2 is connected to parameter setting module 1, real-time data acquisition module 6, fault tree optimization module 4 and data terminal module 5 respectively; ship historical big data module 3 is connected to fault tree optimization module 4.

[0044] The fault analysis module 2 analyzes ship faults based on the parameters set by the parameter setting module 1 and the fault tree model constructed by the ship historical big data module 3, and the real-time data collected by the real-time data acquisition module 6.

[0045] Data terminal module 5 generates alarm signals and operation and maintenance decisions based on the ship's fault analysis results; if a fault has already occurred, it detects the fault, locates the fault, performs maintenance as needed, and provides maintenance suggestions.

[0046] The fault tree optimization module 4 constructs a reference fault tree model using historical data from the ship historical big data module 3. It compares the fault model constructed by the fault analysis module 2 with the reference fault tree model and optimizes the fault model by adding and / or deleting branches.

[0047] The parameter setting module 1 includes setting basic equipment parameters and setting shore-based parameters. The basic equipment parameters include the system knowledge base, equipment fault database, and design data provided by the equipment manufacturer. The shore-based parameters include the usage scenario data, reliability data, and maintenance data of the ship's power system provided by the shore.

[0048] Fault Analysis Module 2 uses the S4000P standard to construct a fault tree model for ship fault analysis. First, data preprocessing is performed, and relevant data for preventative maintenance tasks are defined, forming a master task list from the dataset. Based on the master task list and related data, system fault maintenance analysis is performed sequentially, including system failure mode and effects analysis, fault layer analysis, and PMTRI analysis. Structural inspection task analysis and area inspection task analysis are also performed to obtain the fault tree model. The obtained fault tree is then validated to determine if it meets accuracy requirements. If not, the analysis is replanned; otherwise, preventative maintenance analysis is executed.

[0049] like Figure 3 The diagram shows the functions that the system described in the specific embodiment of the present invention can achieve, including: configuration management function, remote operation and maintenance management function, auxiliary management function, maintenance management function, equipment technical data management function, and spare parts management function;

[0050] Configuration management functionality, which specifically includes:

[0051] The maintenance configuration table summary function allows for the centralized configuration of all necessary maintenance operations, tools, and precautions in a categorized manner.

[0052] Dynamic configuration function, including configuration of spare parts and components, and configuration of procurement information summary and statistics;

[0053] Data-related configuration functions allow for unified configuration of equipment information and technical data;

[0054] Remote operation and maintenance management function, which specifically includes:

[0055] The power status monitoring function performs real-time signal acquisition, processing, and detection of the main unit, auxiliary units, power system, and transmission device;

[0056] The energy efficiency management function compares and analyzes the ship's voyage fuel consumption and historical fuel consumption based on real-time ship status data, emissions, fuel consumption, route information, and power system data. It outputs details of the ship's fuel and electricity usage and optimizes the route based on algorithms to achieve energy conservation, emission reduction, and safe and green ships.

[0057] The equipment health function, combining real-time ship operational status data, provides early warnings and assessments of equipment health based on different levels of equipment health models and fault structure trees. For time-related ship failures, comprehensive diagnoses are conducted via remote maintenance mobile terminals from shore-based equipment manufacturers, ship inspection agencies, or experts, providing fault diagnosis, cause analysis, and maintenance recommendations. Simultaneously, an expert knowledge base and maintenance manual are established based on accumulated data.

[0058] The engine room monitoring function collects data from the main and auxiliary equipment on board. Based on the equipment's status, alarm, switching and threshold values, it divides the data into multi-level alarm levels, multi-dimensional data analysis results and multi-dynamic graphs. It also provides real-time statistics of the equipment's historical 3-day trend chart and operating condition comparison, monitors the real-time status of the equipment, and lays the foundation for early warning and maintenance for the judgment and analysis of engine room health.

[0059] The auxiliary management function includes:

[0060] The account access and management function allows you to set permissions for accessing accounts, including permissions such as view only, add / delete content, and cancel account.

[0061] The login and access history function allows users to view all access information of historical users, including file names and access account names.

[0062] Image display assistance features allow you to resize, rotate, move, copy, and maximize images if the display effect is not ideal.

[0063] The file printing output function allows users to print file content, interface content, or select an output link.

[0064] Maintenance management functions, which specifically include:

[0065] The preventative maintenance and inspection function allows users to view information related to preventative maintenance of components, such as maintenance intervals, manpower, and spare parts. The equipment / consumables information query function allows users to query equipment location, quantity, and serial number, and provides maintenance plans and operation manuals.

[0066] The usage and maintenance functions categorize maintenance time intervals, maintenance levels, maintenance classification information, and maintenance type information.

[0067] The maintenance facility information function can provide facility-related maintenance information, including operating procedures, tools and parts required for maintenance, spare parts, and related precautions;

[0068] The repair plan management function displays the required repairs item by item according to the system framework, and deletes and confirms completed repairs.

[0069] The maintenance record function is mainly used to record historical and current maintenance information. The input function records all information related to the current maintenance. The historical maintenance record function allows users to view the maintenance history of the current component; it is presented in a list format for browsing and querying.

[0070] Equipment technical data management function, which specifically includes:

[0071] The equipment development unit information management function can display development unit information, contact information, development time, etc., sorted by ship number;

[0072] The system's initial configuration and main performance management functions include the initial configuration data of the entire system, which is managed according to the ship's power type and can be subcategorized by ship number;

[0073] The system and equipment design data and drawing management function covers the design data of the entire system and equipment. The technical files are also managed according to the ship's power type and can be subcategorized by ship number.

[0074] The equipment training materials management function categorizes the training required for personnel to operate ship equipment into three main subcategories: important, general, and ordinary.

[0075] The equipment operation and maintenance manual management function records the equipment's historical usage time, purpose, personnel, return time, and placement location;

[0076] Spare parts management function, which specifically includes:

[0077] The detailed list function for spare parts includes description information such as principle, function, structure, location, and spare part drawing information.

[0078] The spare parts' current performance status function can be used to track the location and quantity of spare parts, and whether they are brand new or have a fault.

[0079] The failure rate and actual lifespan of critical components and lifespan components are statistically analyzed from the time of commissioning, including the number of failures and remaining service life.

[0080] The equipment has a fault-free operating time and mean time to repair function, which collects and displays the equipment's fault-free operating time and mean time to repair.

[0081] The function of preparing near-term and long-term plans for spare parts procurement allows you to retrieve a list of near-term and long-term plans for spare parts procurement.

[0082] A specific embodiment of the present invention also provides a method for ship power operation and maintenance, the specific steps of which are as follows:

[0083] Step 1: Obtain historical fault data and perform data preprocessing, including normalization and standardization, to remove features with large differences in size or those that easily dominate the target result, thereby enhancing robustness and obtaining data for modeling; randomly divide the data according to the appropriate proportion to obtain training set and test set, which include diesel engine state attributes.

[0084] Step 2: Randomly extract samples from the training set and test set with replacement to form several sub-training sets and several sub-test sets. Fill in the missing diesel engine state attributes in the sub-training sets and discretize the continuous diesel engine state attributes; fill in the missing diesel engine state attributes in the sub-test sets and discretize the continuous diesel engine state attributes.

[0085] Discrete processing of continuous features:

[0086] The continuous feature A of x samples has x values, arranged in ascending order as a1, a2, ..., a3. x The average of two adjacent sample values ​​is used as the split point. There are a total of x-1 split points, where the i-th split point is T. i Represented as: T i =(a i +a i+1 ) / 2.

[0087] Calculate the information gain ratio (IVR) for each of the x-1 points as a binary splitting point. Select the point with the highest IVR as the optimal splitting point for this continuous feature. For example, let's say the point with the highest IVR is 'a'. t Then less than a t The value is category 1, which is greater than a. t The value is category 2, thus achieving the discretization of continuous features.

[0088] Step 3: Merge similar attributes in the diesel engine state attributes of the sub-training set;

[0089] Step 4: Determine the maximum depth threshold of the fault tree based on the correlation of each attribute in the diesel engine state attributes, and then construct the fault tree using the information gain method based on the maximum depth threshold and the training set;

[0090] Method for determining the maximum depth threshold:

[0091] When the number of current attributes is less than 5, no maximum depth threshold is applied; when the number of current attributes is greater than or equal to 5, the fault tree threshold is [value missing].

[0092] Where m represents the number of attributes.

[0093] When a new branch needs to be generated, the splitting factor in the information gain ratio is optimized first. The formula for obtaining the splitting factor is as follows:

[0094]

[0095] Among them, S k The number of sample data for the current k-th attribute is represented by β; S is the number of samples in the dataset; βk The probability average coefficient, (0<β k <1), where p k p represents the probability of the current attribute. max p represents the maximum probability of the attribute. min This represents the minimum probability of the attribute.

[0096] When the difference in information gain before and after the split meets a preset condition, the split forms a new branch. When the difference in information gain before and after the split does not meet the preset condition, the splitting factor in the information gain ratio is repeatedly optimized based on the difference in information gain before and after the split until the preset condition is met, and a new branch is formed. When repeatedly optimizing the splitting factor in the information gain ratio based on the difference in information gain before and after the split, the formula for obtaining the splitting factor is as follows:

[0097]

[0098] Among them, S k The number of sample data for the current k-th attribute is represented by β; S is the number of samples in the dataset; β k The probability average coefficient, (0<β k <1), where p k p represents the probability of the current attribute. max p represents the maximum probability of the attribute. min α represents the minimum probability of the attribute; α is the difference between the information gain value before and after the split.

[0099] The preset condition is that the difference between the information gain value before and after the split is less than a preset value. The preset value is obtained by using the squared value of the Pearson correlation coefficient r. 2 To determine the preset value, the value range is between (0, 1), and the preset value is preferably 0.05.

[0100] Fault tree construction process:

[0101] Suppose S is a sample set consisting of l data points, and the category attribute has x distinct values ​​s1, s2, ..., sn. x If attribute A has k distinct values, then S is the new leaf node in the decision tree, and A is the non-leaf node. The position of the non-leaf node is determined by its value: the higher the value of attribute A, the higher its position; and so on downwards. The total number of values ​​can be represented as:

[0102]

[0103] In the formula, P iThis represents the proportion of the current parameter value within the total sample values. The current value of the sample is calculated based on the attribute value; the value of attribute A is:

[0104]

[0105] in,

[0106]

[0107]

[0108] Finally, the information gain of the sample set S divided by attribute A is:

[0109] G(A) = I(s², s², ..., s²) x )-E(A)

[0110] Based on this optimization algorithm, a simple decision tree model for establishing the operating status of equipment is proposed.

[0111] The optimized C4.5 algorithm uses information gain ratio to select the optimal partitioning attribute.

[0112]

[0113] Step 5: Prune the constructed fault tree and test the fault tree with the test set;

[0114] The specific steps for pruning are as follows:

[0115] For a leaf node covering N samples with E errors, the error rate of that leaf node is (E+0.5) / N. The 0.5 is the penalty factor. For a subtree with L leaf nodes, the estimated misclassification rate of that subtree is:

[0116]

[0117] Among them, E i Let L represent the number of misclassified samples in each leaf node of the subtree, where L is the number of leaf nodes in the subtree, and N is the number of misclassified samples in each leaf node of the subtree. i The number of samples for each leaf node.

[0118] Estimate the mean and standard deviation of the number of misclassifications for this tree:

[0119] E (number of misjudgments in subtree) = N*e

[0120]

[0121] After replacing the subtree with a leaf node, since the subtree is merged into a single leaf node, L = 1. Substituting this into the formula for calculating the false positive rate above, we can obtain the false positive rate for the leaf node. Therefore, the average number of misclassifications for leaf nodes is

[0122] E (number of misjudgments at leaf nodes) = N*e

[0123] Conditions for pruning:

[0124] E(number of misjudgments in the subtree) + std(number of misjudgments in the subtree) > E(number of misjudgments in the leaf node)

[0125] Step 6: Input the real-time diesel engine status into the fault tree to diagnose the fault.

[0126] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for ship power operation and maintenance, characterized in that, The ship power operation and maintenance method includes the following steps: Step 1: Obtain historical fault data, perform data preprocessing, and obtain training set and test set. The training set and test set include diesel engine status attributes. Step 2: Randomly extract samples from the training set and test set with replacement to form several sub-training sets and several sub-test sets. Fill in the missing diesel engine state attributes in the sub-training sets and discretize the continuous diesel engine state attributes; fill in the missing diesel engine state attributes in the sub-test sets and discretize the continuous diesel engine state attributes. Step 3: Merge similar attributes in the diesel engine state attributes of the sub-training set; Step 4: Determine the maximum depth threshold of the fault tree based on the correlation of each attribute in the diesel engine state attributes, and then construct the fault tree using the information gain method based on the maximum depth threshold and the training set. In constructing the fault tree, when a new branch needs to be generated, the splitting factor in the information gain rate is first optimized. When the difference in information gain before and after the split meets the preset conditions, the split is formed to create a new branch. When the difference in information gain before and after the split does not meet the preset conditions, the splitting factor in the information gain rate is repeatedly optimized according to the difference in information gain before and after the split until the preset conditions are met, and the split is formed to create a new branch. When optimizing the splitting factor term in the information gain ratio, the formula for obtaining the splitting factor term is as follows: Among them, S k The number of sample data for the current k-th attribute is represented by β; S is the number of samples in the dataset; β k The probability average coefficient, Where p k p represents the probability of the current attribute. max p represents the maximum probability of the attribute. min This represents the minimum probability of the attribute. When repeatedly optimizing the split factor term in the information gain ratio based on the difference in information gain before and after the split, the formula for obtaining the split factor term is as follows: Among them, S k The number of sample data for the current k-th attribute is represented by β; S is the number of samples in the dataset; β k The probability average coefficient, Where p k p represents the probability of the current attribute. max p represents the maximum probability of the attribute. min α represents the minimum probability of the attribute; α is the difference between the information gain value before and after the split. Step 5: Prune the constructed fault tree and test the fault tree with the test set; Step 6: Input the real-time diesel engine status into the fault tree to diagnose the fault.

2. The ship power operation and maintenance method as described in claim 1, characterized in that, The method for determining the maximum depth critical value in step 4: When the number of current attributes is less than 5, no maximum depth threshold is applied; when the number of current attributes is greater than or equal to 5, the fault tree threshold is [value missing]. Where m represents the number of attributes.

3. The ship power operation and maintenance method as described in claim 1, characterized in that, The preset condition is that the difference between the information gain value before and after the split is less than a preset value. The preset value is obtained by using the squared value of the Pearson correlation coefficient r. 2 To determine the preset value, the value range is between (0,1).

Images