A method, device, equipment and medium for intelligent review of marine operation tasks

Abstract

The present application discloses a method, apparatus, equipment and medium for intelligent review of marine operation tasks, which relates to the field of data processing technology. The method includes: obtaining at least one operation plan document; inputting at least one operation plan document into a recognition model to obtain first operation information; generating a first operation table based on the first operation information; generating a first operation diagram based on the first operation table; inputting the first operation diagram and the first operation table into a first intelligent review model, and outputting a first review result. The present application obtains fine-grained first operation information by identifying at least one operation plan document, and generates a first operation table using the fine-grained first operation information, and then generates a first operation diagram. The first operation table and the first operation diagram are input into the first intelligent review model to obtain a first review result, which effectively verifies the rationality and effectiveness of the operation plan document and the operation task.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, equipment and medium for intelligent review of marine operation tasks. Background Art

[0002] The review of marine operation tasks mainly involves reviewing multiple operation plans prepared by offshore construction personnel. Currently, there are two main methods for reviewing marine operation tasks. The first method is the traditional manual review method, in which specialized reviewers judge whether the operation plan is reasonable and effective based on their experience. This method not only requires the reviewers to have extensive professional experience and knowledge, but also requires them to have real-time access to reliable operation resources and operation area conditions. It is inefficient and has low accuracy. The second method is to obtain an audit model based on artificial intelligence technology training, and verify the operation plan through the audit model. The application of this method in marine operation tasks currently only focuses on reviewing typos and contextual semantics in documents, and does not involve reviewing the effectiveness and rationality of the operation plan in the operation plan document. The review of effectiveness and rationality still relies on specialized reviewers to conduct the review, which is inefficient and has low accuracy. Summary of the Invention

[0003] The main purpose of the embodiments of the present application is to propose an intelligent review method, device, equipment and medium for marine operation tasks, so as to efficiently and accurately review the effectiveness and rationality of marine operation tasks.

[0004] To achieve the above objectives, one aspect of an embodiment of the present application provides an intelligent review method for marine operation tasks, the method comprising the following steps:

[0005] Obtain at least one work plan document;

[0006] Inputting at least one of the operation plan documents into a recognition model to obtain first operation information;

[0007] generating a first operation table according to the first operation information;

[0008] generating a first operation graph according to the first operation table;

[0009] The first operation diagram and the first operation table are input into a first intelligent audit model, and a first audit result is output.

[0010] In some embodiments, inputting at least one of the operation plan documents into a recognition model to obtain first operation information comprises the following steps:

[0011] Inputting at least one of the operation plan documents into the recognition model to identify and obtain the first operation information; wherein the first operation information includes information about a plurality of sub-projects and dependency information between the sub-projects; and each sub-project information includes at least a project identifier, a sub-project identifier, an operation area, an operation volume, an operation vessel, and an estimated number of operation days;

[0012] Generating a first operation table according to the first operation information includes the following steps:

[0013] The first operation table is generated according to a preset template using the first operation information.

[0014] In some embodiments, generating a first operation graph according to the first operation table specifically includes:

[0015] extracting the first job information from the first job table;

[0016] At least one operation branch is constructed with the sub-project identifier as a root node, the project identifier as a first-level leaf node, and the operation vessel as a second-level leaf node;

[0017] The first job graph is constructed according to at least one of the job branches.

[0018] In some embodiments, constructing the first job graph according to at least one of the job branches comprises the following steps:

[0019] Connecting the root nodes in the job branches with edges according to the dependency information between the sub-projects;

[0020] The first operation graph is constructed by taking the operation volume and the operation area as attributes of the first-level leaf node and taking the estimated operation days as attributes of the second-level leaf node.

[0021] In some embodiments, the first intelligent audit model includes a first encoder, a second encoder, a first fuser, and a first-level decoder; the first-level decoder is composed of multiple network blocks and multiple fully connected layers, each of the network blocks includes an LSTM layer, a RELU layer, and a pooling layer;

[0022] The step of inputting the first operation diagram and the first operation table into a first intelligent audit model and outputting a first audit result comprises the following steps:

[0023] Inputting the first job graph into the first encoder to obtain a first encoding vector;

[0024] Inputting the first worksheet into the second encoder to obtain a second encoding vector;

[0025] Inputting the first encoding vector and the second encoding vector into the first fuser to obtain a first fused vector;

[0026] The first fusion vector is input into the first-level decoder to obtain the first audit result.

[0027] In some embodiments, the second encoder includes an extraction sub-model and an encoding sub-model;

[0028] Inputting the first operation table into the second encoder to obtain a second encoding vector includes the following steps:

[0029] Inputting the first operation table into the extraction sub-model to obtain a target key information set;

[0030] The target key information set and the weight information corresponding to the key information are input into the encoding sub-model to obtain a second encoding vector.

[0031] In some embodiments, the method further comprises the following steps:

[0032] Modify the first operation information and the first operation table according to the first audit result to obtain second operation information and a second operation table;

[0033] inputting the second job information into the first job scheduling model to generate first scheduling information;

[0034] generating a first scheduling graph according to the first scheduling information;

[0035] The first scheduling diagram and the second operation table are input into a second intelligent audit model, and a second audit result is output.

[0036] To achieve the above objectives, another aspect of the present application provides an intelligent review device for marine operation tasks, the device comprising:

[0037] An acquisition module, used for acquiring at least one work plan document;

[0038] a recognition module, configured to input at least one of the operation plan documents into a recognition model to obtain first operation information;

[0039] A first generating module, configured to generate a first operation table according to the first operation information;

[0040] A second generating module, configured to generate a first operation graph according to the first operation table;

[0041] The first audit module is used to input the first operation diagram and the first operation table into a first intelligent audit model and output a first audit result.

[0042] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method when executing the computer program.

[0043] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above-mentioned method when executed by a processor.

[0044] The embodiments of the present application include at least the following beneficial effects:

[0045] The present application can obtain at least one work plan document; input at least one work plan document into a recognition model to obtain first work information; generate a first work table based on the first work information; generate a first work diagram based on the first work table; input the first work diagram and the first work table into a first intelligent audit model to output a first audit result. The present application obtains fine-grained first work information by identifying at least one work plan document, and uses the fine-grained first work information to generate a first work table, and then generates a first work diagram. The first work table and the first work diagram are input into the first intelligent audit model to obtain a first audit result, effectively verifying the rationality and effectiveness of the work plan document and the work task. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0047] Figure 1 A flow chart of an intelligent review method for marine operation tasks provided in an embodiment of the present application;

[0048] Figure 2 A schematic diagram of the operation branches provided in the embodiment of the present application;

[0049] Figure 3 A schematic diagram of an operation diagram provided for an embodiment of the present application;

[0050] Figure 4 A schematic diagram of a first intelligent audit model provided in an embodiment of the present application;

[0051] Figure 5 A flowchart of another intelligent review method for marine operation tasks provided in an embodiment of the present application;

[0052] Figure 6 A schematic diagram of a first scheduling diagram provided in an embodiment of the present application;

[0053] Figure 7 A schematic diagram of the structure of an intelligent review device for marine operation tasks provided in an embodiment of the present application;

[0054] Figure 8 A schematic diagram of the structure of another intelligent review device for marine operation tasks provided in an embodiment of the present application;

[0055] Figure 9 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are merely examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0057] It will be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0058] The terms "at least one", "plurality", "each", "any", etc. used in this application include "at least one", "two" or more, "plurality" or "each", "any" or "any one", "each" or "any one" as used herein.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0060] The embodiments of the present application provide a method, device, equipment and medium for intelligent review of marine operation tasks, which relate to the field of data processing technology. The method, device, equipment and medium for intelligent review of marine operation tasks provided by the embodiments of the present application can be applied to a terminal, can also be applied to a server, and can also be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch and a car terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements an intelligent review method for marine operation tasks, etc., but is not limited to the above forms.

[0061] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0062] Reference Figure 1 The embodiment of the present application provides an intelligent review method for marine operation tasks, which may include but is not limited to S1 to S5, as follows:

[0063] S1: Obtain at least one work plan document.

[0064] Each operational plan document contains detailed information about marine operations, including multiple sub-projects. Each sub-project includes at least the project identifier, sub-project identifier, operational area, sub-project workload, the proposed operational vessel, and the estimated number of operational days required for the vessel to perform the sub-project. Sub-projects are designed for marine exploration and may include marine 3D seismic surveys, quasi-3D seismic surveys, 2D seismic surveys, geological sampling, water sampling, gravity and magnetic surveys, thermal flow surveys, ROV surveys, unmanned vessel surveys, and multi-beam surveys. Marine operational information also includes dependencies between sub-projects, including whether sub-projects A and B can be executed simultaneously, whether sub-project A should be executed before sub-project B, and whether sub-project A should be executed after sub-project B.

[0065] S2: Input at least one operation plan document into the recognition model to obtain first operation information.

[0066] Inputting the at least one acquired operation plan document into an intelligent recognition model, and obtaining first operation information through the intelligent recognition model. The intelligent recognition model may utilize an optical character recognition (OCR) network and / or a deep neural network. The operation plan document is input into the recognition model, and the intelligent recognition model is used to recognize and obtain marine operation information as the first operation information.

[0067] In one embodiment, the job plan document contains multimodal data such as text, images, and tables. The intelligent recognition model uses a combined neural network model, which includes a document structure recognition network, a feature extraction network, a feature fusion network, and a decoding network. The feature extraction network includes an image feature extraction subnetwork, a text feature extraction subnetwork, and a table feature extraction subnetwork.

[0068] Optionally, inputting at least one of the operation plan documents into a recognition model to obtain first operation information comprises the following steps:

[0069] Inputting each document image in the job plan document into a document structure recognition network to identify text areas, image areas, and table areas in each document image;

[0070] Input the text area into the text feature extraction subnetwork to extract the first key feature;

[0071] Input the image region into the image feature extraction subnetwork to extract the second key feature;

[0072] Input the table area into the table feature extraction sub-network to extract the third key feature;

[0073] Input the first key feature, the second key feature, and the third key feature into the feature fusion network, and output the key feature through the feature fusion network;

[0074] The key features are input into the decoding network to obtain key text information, and the key text information is used as the first operation information.

[0075] The text feature extraction subnetwork can use BILSTM and BIGRU networks, while the image and table feature extraction subnetworks can use convolutional neural networks. The decoding network can use BILSTM and BIGRU networks. The document structure recognition network uses a commonly used document region partitioning network.

[0076] In another embodiment, the recognition model uses an OCR network and a dictionary matching model. Key fields of the marine operation information are stored in a dictionary model. The operation plan document is recognized as text data through the OCR network. The text data is matched with the dictionary model to obtain the first operation information.

[0077] S3: Generate a first operation table according to the first operation information.

[0078] After identifying and obtaining the first operation information, a first operation table is generated based on the first operation information according to a preset operation table generation template. The first operation table generation template is preconfigured, and the first operation information is populated into the first operation table according to the template fields, thereby generating the first operation table. The template fields include at least a project name field, an operation area field, a subproject name field, a subproject workload field, a field for the intended operation vessel to be used for the subproject, a field for the estimated operation days field, and a dependency field.

[0079] In one embodiment, in order to improve the generation efficiency of the first job table, a parallel and / or multi-threaded method is used to obtain the first job table, specifically: the recognition model and the job table generation template are pre-configured on the cluster nodes; multiple nodes on the cluster nodes used for job plan document recognition and job table generation are selected as the target node set based on the load balancing algorithm; according to the number of job plan documents, multiple job plan documents are sent to each target node in the target node set; each target node uses the recognition model and the job table generation template to generate multiple first temporary job tables; and multiple first temporary job tables are spliced ​​to obtain the first job table.

[0080] For example, there are four work plan documents, where the first work plan document records the work plan for Project A, the second work plan document records the work plan for Project B, the third work plan document records the work plan for Project C, and the fourth work plan document records the work plan for Project D. The generated first work table is shown in Table 1:

[0081]

[0082]

[0083] Table 1

[0084] In Table 1, A, B, C, and D respectively represent project names or project IDs; a, b, c, and d respectively represent subproject names or subproject IDs; Ar1, Ar2, Ar3, and Ar4 respectively represent operation area names or operation area identifiers; and V1, V2, and V3 respectively represent operation vessel names or operation vessel identifiers.

[0085] S4: Generate a first operation diagram according to the first operation table.

[0086] After generating the first operation table, it is necessary to generate a first operation graph based on the first operation table, specifically including: extracting the first operation information in the first operation table; constructing at least one operation branch with the sub-project as the root node, the project as the first-level leaf node, and the operation vessel as the second-level leaf node; and constructing an operation graph based on the at least one operation branch. Figure 2 The diagram shows the job branches constructed according to the first job table shown in Table 1.

[0087] After constructing and obtaining at least one job branch, a first job graph is constructed based on the at least one job branch. Specifically, the root nodes in at least one job branch are connected by edges based on the dependency information between sub-projects. The workload and the job area are used as attributes of the first-level leaf nodes, and the estimated number of working days is used as the attribute of the second-level leaf nodes to construct the first job graph. Among them, V represents a node, E represents an edge, and X represents a node attribute.

[0088] Based on the dependencies between sub-projects, the root nodes in at least one job branch are connected by edges. Specifically, if a dependency exists between sub-projects, bidirectional directed edges and / or unidirectional directed edges are used for connection; if no dependency exists, undirected edges are used for connection. Specifically, when the dependency relationship is that sub-project A should be executed before sub-project B, or sub-project A should be executed after sub-project B, unidirectional directed edges are used for edge connection; when the dependency relationship is that sub-project A and sub-project B can be executed simultaneously, bidirectional directed edges are used for edge connection. Figure 3 Shown is based on Figure 2 The job graph built by the job branches in .

[0089] S5: Input the first operation diagram and the first operation table into the first intelligent audit model and output a first audit result. Optionally, the first audit result includes whether there is abnormal information in the operation information, and if there is abnormal information, indicates the abnormal information.

[0090] In one embodiment, referring to Figure 4The first intelligent audit model includes a first encoder, a second encoder, a first fusion device and a first-level decoder; the first-level decoder consists of multiple network blocks and multiple layers of fully connected layers, and each network block includes an LSTM layer, a RELU layer and a pooling layer.

[0091] Inputting the first operation diagram and the first operation table into the first intelligent review model and outputting a first review result, specifically including:

[0092] S51: Input the first operation graph into the first encoder to obtain a first encoding vector.

[0093] The first encoder uses a graph neural network GNN to input the first job graph into the graph neural network for encoding to obtain a first encoding vector. Specifically, according to the first job graph Construct the adjacency matrix A (1) , weight W1 (1) and node attributes In constructing weight W1 (1) When a unidirectional directed edge is used between nodes, the weight is 1; when a bidirectional directed edge is used between nodes, the weight is 2; when an undirected edge is used between nodes, the weight is 0; when constructing node attributes When , the attribute of the first-level leaf node is workload, and the attribute of the second-level leaf node is expected working days. (1) , weight W1 (1) and node attributes Get the first encoding vector

[0094] S52: Input the first operation table into the second encoder to obtain a second encoding vector.

[0095] The second encoder includes an extraction sub-model and an encoding sub-model; the extraction sub-model is used to extract the key information set from the first job table, and the encoding sub-model is used to encode the key information set according to the key information set and the weight information corresponding to the key information to obtain the second encoding vector

[0096] Specifically, the first job table is input into the second encoder to obtain a second encoding vector, including:

[0097] Input the first job table into the extraction sub-model to obtain the target key information set;

[0098] The target key information set and the weight information corresponding to the key information are input into the encoding sub-model to obtain a second encoding vector.

[0099] In one embodiment, key fields and their corresponding weight information are pre-stored in a database in the form of a dictionary (key-value). The first job table is input into the extraction sub-model, which matches the key fields in the dictionary in the first job table to obtain a set of key information and the corresponding weight information. The key information set and the corresponding weight information for each key information are input into the encoding sub-model to obtain a second encoding vector:

[0100]

[0101] in, Represents the target key information set, W2 (1) Represents weight information.

[0102] S53: Input the first encoding vector and the second encoding vector into the first fuser to obtain a first fused vector.

[0103]

[0104] S54: Input the first fused vector to the first-level decoder to obtain a first audit result. The first-level decoder is composed of multiple network blocks and multiple fully connected layers, each network block including an LSTM layer, a RELU layer, and a pooling layer.

[0105] y (1) =Decoder_BLOCK_LSTM(v (1) );

[0106] By using multiple network blocks in the decoder, the input feature vector can be reduced in multiple levels, key entity information can be extracted, and computational efficiency can be improved. At the same time, the use of multi-level LSTM network layers can obtain more comprehensive contextual semantic information, thereby improving computational accuracy.

[0107] For example, the first-level decoder includes three network blocks and two fully connected layers. Each network block includes an LSTM layer, a RELU layer, and a pooling layer. The pooling layer uses maximum pooling or average pooling. The first fusion vector is input to the first-level decoder to obtain a first audit result, which specifically includes:

[0108] The first fusion vector is input into the first network block to obtain the first network block vector, specifically:

[0109]

[0110] The first network block vector is input into the second network block to obtain the second network block vector, specifically:

[0111]

[0112] The third network block vector is input into the third network block to obtain the third network block vector, specifically:

[0113]

[0114] The first network block vector, the second network block vector, and the third network block vector are input into the first fully connected layer to obtain the first fully connected layer vector, specifically:

[0115]

[0116] The first fully connected layer vector is input into the second fully connected layer to obtain the first audit result, which is specifically:

[0117]

[0118] The first intelligent audit model can detect anomalies in the first operation information and indicate the presence of such anomalies when anomalies are detected. The anomaly information includes whether the dependencies between sub-items in the marine operation information are reasonable, whether the operation vessels used to execute the sub-items are reasonable, whether the estimated operation days to complete the sub-items are reasonable, etc. For example, the first operation table 1 and the first operation diagram Figure 3 The data is input into the first intelligent audit model, and an anomaly is detected in the operation information. The anomaly is that project identifier C and sub-project identifier c can still be executed using the operation vessel V3. At the same time, due to the actual operation environment of the operation area Ar3, the estimated operation time of sub-project c and sub-project d has obvious deviations and should be appropriately extended.

[0119] Before using the first intelligent audit model to audit marine operations information, it is necessary to train the model using training samples. The training samples include manually annotated historical sample documents and operational area information obtained through web crawling. A loss function is set, and when the loss function reaches a target threshold, the first intelligent audit model is trained.

[0120] This embodiment obtains first operation information by intelligently identifying multiple marine operation plan documents, constructs a first operation table and a first operation diagram with dependency and association relationships based on the first operation information, inputs the first operation table and the first operation diagram into a pre-trained first intelligent audit model to obtain an audit result, thereby improving the audit efficiency and accuracy of marine operation information.

[0121] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0122] (1) The embodiment of the present application proposes an intelligent audit method for marine operation plan documents and marine operation tasks. The method obtains fine-grained operation information by identifying multiple operation plan documents, and uses the fine-grained operation information to generate a first operation table and a first operation diagram. The first operation table and the first operation diagram are input into the intelligent audit model to obtain the audit results, which effectively verifies the rationality and effectiveness of the operation plan documents and the operation tasks. At the same time, the present application uses the first intelligent audit model trained and constructed with prior knowledge and rich external knowledge to perform intelligent audits on the operation tasks, further improving the efficiency and accuracy of the marine operation task audits.

[0123] (2) The embodiment of the present application generates a first ocean scheduling diagram based on the operation information, and before the first scheduling diagram is generated, the operation information is audited by a first intelligent audit model, and the operation information is adjusted based on the audit results, thereby improving the accuracy of the operation scheduling information. After the operation scheduling diagram is generated, the generated scheduling information is audited by a second intelligent audit model, thereby further improving the accuracy of the operation scheduling.

[0124] (3) When generating an operation diagram, the embodiment of the present application takes into account the affiliation, dependency and importance between the various information in the operation information, takes the sub-project as the root node, the project as the first-level leaf node, and the operation vessel as the second-level leaf node, and constructs at least one operation branch. Then, using the dependency information, the root nodes in the operation branch are connected, the operation volume is used as the attribute of the first-level leaf node, and the operation days are used as the attribute of the second-level leaf node. It can not only visually display the core information in the operation plan document, but also provide important data support for intelligent review.

[0125] Reference Figure 5 The embodiment of the present application further proposes a solution for generating a first scheduling diagram based on the operation information and performing an intelligent audit on the first scheduling diagram based on the second intelligent audit model, which specifically includes the following steps S6 to S9:

[0126] S6: Modify the first operation information and the first operation table according to the first audit result to obtain second operation information and a second operation table.

[0127] After the first intelligent audit model completes the intelligent audit of the marine operation task, the first operation information and the first operation table are modified based on the first audit results to obtain second operation information and a second operation table. During the modification, the anomalies in the first operation information and the first operation table are adjusted based on the indication of the anomaly information.

[0128] S7: Input the second job information into the first job scheduling model to generate first scheduling information.

[0129] A first job scheduling model is pre-constructed, and the first job scheduling model includes a scheduling encoding network, a scheduling generation network, and a scheduling decision network; the scheduling generation network includes multiple generation sub-networks, and the multiple generation sub-networks are used to generate scheduling information corresponding to sub-project information.

[0130] After the first job scheduling model is trained, the first job scheduling model and the second job information are used to generate first scheduling information, specifically including:

[0131] S71: Input multiple sub-item information into the scheduling coding network to obtain multiple sub-item vectors.

[0132] S72: Input multiple sub-project vectors into the scheduling generation network, and generate multiple first scheduling sub-information corresponding to the multiple sub-projects through the generation network; the first scheduling sub-information includes operation vessel information, operation day information and window label information.

[0133] In the actual production process, considering the influence of various factors such as weather and waves, the operation cycle is divided into multiple continuous operation windows on a monthly basis. For example, March to September of the current year is divided into the first window, and September of the current year to March of the following year is divided into the second window. The first window label and the second window label are used to identify the first window and the second window respectively.

[0134] When constructing the scheduling generative network, multiple generative subnetworks are built for each of the work vessels. For example, a first generative subnetwork is constructed for work vessel V1, a second generative subnetwork is constructed for work vessel V2, and a third generative subnetwork is constructed for work vessel V3. Labels are assigned to each of the generative subnetworks based on the sub-project categories that the work vessels can perform. The first generative subnetwork is assigned label a, the second generative subnetwork is assigned labels b, c, and d, and the third generative subnetwork is assigned labels a, b, and d. The generative subnetworks employ a generative adversarial network model.

[0135] Inputting the multiple sub-item vectors into the schedule generation network, and generating multiple first schedule sub-information corresponding to the multiple sub-items through the schedule generation network, specifically includes:

[0136] Based on the sub-item identifier and the generated sub-network label, the corresponding generated sub-network is matched. For example, the first and third generated sub-networks are matched based on sub-item category a, the second and third generated sub-networks are matched based on sub-item category b, and the second generated sub-network is matched based on sub-item category c.

[0137] Input the sub-item vectors into the matching generative sub-networks. For example, the sub-item vectors corresponding to the sub-item information (A, a, Ar1, 5000) are input into the first and third generative sub-networks, the sub-item information (B, b, Ar2, 300) are input into the second and third generative sub-networks, and the sub-item information (B, c, Ar2, 3000) is input into the second generative sub-network.

[0138] The first scheduling sub-information is outputted by generating a sub-network. The first scheduling sub-information includes a project identifier, a sub-project identifier, an operation area information, an operation vessel information, a first window label and first operation day information, a second window label and second operation day information.

[0139] S73: Inputting the dependency data between the plurality of first scheduling sub-information and the sub-items into the scheduling decision network to generate first scheduling information.

[0140] After obtaining the first scheduling sub-information, the dependency relationship between multiple first scheduling sub-information and sub-projects is input into the scheduling decision network, and the first scheduling information is output through the scheduling decision network; the first scheduling information includes the execution sequence corresponding to the operation vessel, operation project identifier, project sub-identifier, operation area and project sub-identifier.

[0141] The scheduling decision network adopts a multi-network fusion structure based on dependency relationships, including an input layer, an embedding layer, a feature extraction layer, a first gating layer, a multi-block shared MLP layer, a linear layer, and an output layer, where:

[0142] The input layer is used to receive a plurality of first scheduling sub-information and dependency relationship information between sub-items;

[0143] The embedding layer is used to convert the first scheduling sub-information and the corresponding dependency information into feature vectors to obtain the sub-information feature vector and the dependency feature vector;

[0144] The fusion layer is used to fuse the sub-information feature vector and the dependency feature vector to obtain a fused feature vector;

[0145] Utilize multiple feature extraction networks in the feature extraction layer to extract features from the fused feature vector to obtain multiple initial feature vectors; wherein the feature extraction layer may adopt a deep neural network model;

[0146] The first gating layer is used to generate first weight data corresponding to each feature extraction network, and perform weighted summation on the initial feature vectors output by each feature extraction network to obtain a first combined feature vector;

[0147] The multi-block shared MLP layer is used to perform task migration processing on the first combined feature vector to obtain a first migration feature vector;

[0148] The linear layer is used to transform the first migration feature vector to obtain first scheduling information;

[0149] The output layer is used to output the first scheduling information.

[0150] In one embodiment, the multi-block shared MLP layer is a multi-level block network structure, each block includes m MLPs and a shared layer, the shared layer of the i-th block is used to extract the output features of the m MLPs in the current block, obtain m knowledge features, and convert the MLP i k1 The output features of MLP i k2 The extracted knowledge features are fused as MLP i+1 k1 The input is , where m is the number of operating vessels, MLP represents the multi-layer perceptron model, k2 = 1:m and k2 ≠ k1.

[0151] For example, the output of the MLPs in the i-th block is: X_MLP i 1 , ..., X_MLP i k1 , ..., X_MLP i m .

[0152] The input is shared to the shared layer, which extracts knowledge and obtains the knowledge feature data of the i-th level:

[0153] Y_MLP i 1 =EXTRACT(X_MLP i 1 ),……,Y_MLP i k1 =EXTRACT(X_MLP i k1 ),……,Y_MLP i m =EXTRACT(XMLP i m );

[0154] The feature data input to the MLPs in the i+1th level is:

[0155] {Input_MLP i+1 1} = concat(X_MLP i 1 ,Y_MLP i 2 ,……,Y_MLP ik1 ,……,Y_MLP i m );

[0156] {Input_MLP i+1 k1} = concat(X_MLP i k1 ,Y_MLP i 1 ,……,Y_MLP i k1-1 ,Y_MLP i k1+1 ,……,Y_MLP i m );

[0157] {Input_MLP i+1 m} = concat(X_MLP i m ,Y_MLP i 1 ,……,Y_MLP i k1 ,……,Y_MLP i m-1 ).

[0158] Through the first scheduling model, the first scheduling information is obtained as shown in Table 2:

[0159] Ship number V1 V2 V3 (V1, A, a, Ar1, 1) (V2, C, c, d, Ar3, 1) (V3, D, d, Ar4, 1) (V1, C, a, Ar3, 2) (V2, B, c, Ar2, 2) (V3, B, a, Ar2, 1) (V3, B, b, Ar2, 3)

[0160] Table 2

[0161] S8: Generate a first scheduling graph according to the first scheduling information.

[0162] After generating the first scheduling information, the first scheduling graph is constructed with the operating vessel as the root node, the triple consisting of the operation project identifier, project sub-identifier, and operation area as the leaf node, and the execution order corresponding to the project sub-identifier as the constraint condition. like Figure 6 As shown, Figure 6 The solid lines in the figure represent the execution order of sub-items for the same vessel, and the dotted lines represent the execution order of sub-items between different vessels. Figure 6 In the example, vessel V3 first completes the work area Ar2, project identifier B, and sub-project identifier a. Next, vessel V3 completes the work area Ar2, project identifier B, and sub-project identifier c. Finally, vessel V3 completes the work area Ar2, project identifier B, and sub-project identifier b. By constructing the first scheduling diagram, the work scheduling plan can be clearly reflected.

[0163] S9: Input the first scheduling diagram and the second operation table into the second intelligent audit model, and output a second audit result.

[0164] The second intelligent audit model adopts an architecture similar to the first intelligent audit model, which includes a third encoder, a fourth encoder, a second fusion unit and a second-level decoder; the second-level decoder is also composed of multiple network blocks and multiple layers of fully connected layers, each network block includes an LSTM layer, a RELU layer and a pooling layer.

[0165] Input the first scheduling diagram and the second operation table into the second intelligent audit model, and output a second audit result, specifically including:

[0166] S91: Input the first scheduling graph into the third encoder to obtain a third encoding vector.

[0167] The third encoder uses a convolutional neural network to input the first scheduling graph into the convolutional neural network for encoding, and obtains the third encoding vector:

[0168]

[0169] S92: Input the first operation table into the fourth encoder to obtain a fourth encoding vector.

[0170] The fourth encoder uses the same architecture as the second encoder, and also includes an extraction sub-model and an encoding sub-model; the extraction sub-model is used to extract the key information set from the second job table, and the encoding sub-model is used to encode the key information set according to the key information set and the weight information corresponding to the key information to obtain the fourth encoding vector

[0171]

[0172] in, Represents the target key information set, W2 (1) Represents weight information.

[0173] S93: Input the third encoding vector and the fourth encoding vector into the second fuser to obtain a second fused vector:

[0174]

[0175] S94: Input the second fusion vector to the second-level decoder to obtain a second audit result:

[0176] y (2) =Decoder_BLOCK_LSTM(v (2) ).

[0177] For example, the second-level decoder includes three network blocks and two fully connected layers. Each network block includes an LSTM layer, a RELU layer, and a pooling layer. The pooling layer uses maximum pooling or average pooling. The second fusion vector is input to the second-level decoder to obtain a second audit result, which specifically includes:

[0178] The second fusion vector is input into the first network block to obtain the first network block vector, specifically:

[0179]

[0180] The first network block vector is input into the second network block to obtain the second network block vector, specifically:

[0181]

[0182] The third network block vector is input into the third network block to obtain the third network block vector, specifically:

[0183]

[0184] The first network block vector, the second network block vector, and the third network block vector are input into the first fully connected layer to obtain the first fully connected layer vector, specifically:

[0185]

[0186] The first fully connected layer vector is input into the second fully connected layer to obtain the second audit result, which is specifically:

[0187]

[0188] The second intelligent audit model detects anomalies in the job schedule and indicates the presence of any anomalies. This includes information such as whether the execution order of sub-items in the job information is reasonable, whether the vessels used to execute sub-items are reasonable, and whether any sub-items are missing.

[0189] In this embodiment, before the marine operation scheduling information is generated, the operation information is audited by the first intelligent audit model, and the operation information is adjusted based on the audit results, thereby improving the accuracy of the scheduling information generation; at the same time, after the marine operation scheduling information is generated, the generated scheduling information is audited by the second intelligent audit model, thereby further improving the accuracy of the scheduling information generation.

[0190] Reference Figure 7 The present application also provides an intelligent marine operation task review device, which includes:

[0191] An acquisition module 701 is used to acquire at least one job document;

[0192] The recognition module 702 is configured to input at least one of the job documents into a recognition model to obtain first job information;

[0193] A first generating module 703 is configured to generate a first operation table according to the first operation information;

[0194] A second generating module 704 is configured to generate a first operation graph according to the first operation table;

[0195] The first review module 705 is configured to input the first operation diagram and the first operation table into a first intelligent review model and output a first review result.

[0196] The identification module 702 identifies at least one job plan document and obtains first job information, which specifically includes:

[0197] At least one operation plan document is input into the recognition model to obtain first operation information; the first operation information includes at least one project information, each project information includes multiple sub-project information and dependency information between sub-projects; the sub-project information includes at least project identification, sub-project identification, operation area, operation volume, operation vessel and operation days.

[0198] The first generating module 703 generates a first operation table according to the first operation information, specifically including: generating the first operation table according to a preset template based on the first operation information.

[0199] The second generating module 704 generates a first operation graph according to the first operation table, specifically including:

[0200] Extracting first job information from the first job table;

[0201] At least one operation branch is constructed with the sub-project as the root node, the project as the first-level leaf node, and the operation vessel as the second-level leaf node;

[0202] Construct a job graph based on at least one job branch, specifically: connect the root nodes in the job branches based on dependency information; use the job volume and job area as attributes of the first-level leaf nodes, and use the job days as attributes of the second-level leaf nodes to construct the job graph.

[0203] The first intelligent audit model in the first audit module 705 includes a first encoder, a second encoder, a first fuser, and a first-level decoder; the first-level decoder is composed of multiple network blocks, each of which includes an LSTM network, a RELU layer, and a pooling layer;

[0204] The first review module 705 obtains a first review result according to the first operation diagram and the first operation table, specifically including:

[0205] Inputting the first job graph into the first encoder to obtain a first encoding vector;

[0206] Inputting the first worksheet into the second encoder to obtain a second encoding vector; the second encoder includes an extraction sub-model and an encoding sub-model; inputting the first worksheet into the second encoder to obtain the second encoding vector, specifically: inputting the first worksheet into the extraction sub-model to obtain a target key information set; inputting the target key information set and weight information corresponding to the key information into the encoding sub-model to obtain the second encoding vector;

[0207] Inputting the first encoding vector and the second encoding vector into a first fuser to obtain a first fused vector;

[0208] The first fusion vector is input into the first-level decoder to obtain a first audit result.

[0209] In another embodiment, referring to Figure 8 The embodiment of the present application proposes an intelligent review device for marine operation tasks, further comprising:

[0210] A correction module 706 is configured to correct the first operation information and the first operation table according to the first review result to obtain second operation information and the second operation table;

[0211] The third generating module 707 is used to input the second job information into the first job scheduling model to generate first scheduling information;

[0212] A fourth generating module 708 is configured to generate a first scheduling graph according to the first scheduling information;

[0213] The second audit module 709 is configured to input the first scheduling diagram and the second operation table into a second intelligent audit model and output a second audit result.

[0214] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0215] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of the present application. The electronic device can be any smart terminal, such as a tablet computer or an in-vehicle computer.

[0216] It can be understood that the contents of the above method embodiments are all applicable to the embodiments of the present device, the functions specifically implemented by the embodiments of the present device are the same as those of the method of the present application, and the beneficial effects achieved are also the same as those achieved by the method of the present application.

[0217] See also Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0218] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0219] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the methods of the embodiments of this application.

[0220] Input / output interface 903, used to implement information input and output;

[0221] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0222] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0223] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other via a bus 905 within the device.

[0224] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method of the present application is implemented.

[0225] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiment, the functions specifically implemented by the present storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0226] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0227] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0228] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0229] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0230] Those skilled in the art will appreciate that all or some of the steps, devices, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0231] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the numbers used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or equipment.

[0232] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0233] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules described above is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0234] The modules described above as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0235] In addition, the functional modules in the various embodiments of the present application may be integrated into a processing module, or each module may exist physically separately, or two or more modules may be integrated into a single module. The above-mentioned integrated modules may be implemented in the form of hardware or software functional modules.

[0236] If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store programs.

[0237] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. An intelligent audit method for marine operation tasks, characterized in that: The method comprises the following steps: Obtain at least one work plan document; Inputting at least one of the operation plan documents into a recognition model to obtain first operation information; generating a first operation table according to the first operation information; generating a first operation graph according to the first operation table; Inputting the first operation diagram and the first operation table into a first intelligent review model, and outputting a first review result; Generating a first operation table according to the first operation information includes the following steps: Generate the first operation table according to the preset template using the first operation information; A first job table is obtained using a parallel and / or multi-threaded method, specifically: pre-configuring a recognition model and a job table generation template on a cluster node; selecting, based on a load balancing algorithm, multiple nodes on the cluster node for job plan document recognition and job table generation as a target node set; sending multiple job plan documents to each target node in the target node set according to the number of job plan documents; each target node generates multiple first temporary job tables using the recognition model and the job table generation template; and concatenating the multiple first temporary job tables to obtain a first job table. Generating a first operation graph according to the first operation table specifically includes: extracting the first job information from the first job table; At least one operation branch is constructed with the sub-project identifier as the root node, the project identifier as the first-level leaf node, and the operation vessel as the second-level leaf node; constructing the first operation graph according to at least one of the operation branches; The step of constructing the first job graph according to at least one job branch comprises the following steps: Connecting the root nodes in the job branches with edges based on dependency information between the sub-projects; The first operation graph is constructed by taking the operation volume and operation area as the attributes of the first-level leaf nodes and taking the estimated operation days as the attributes of the second-level leaf nodes.

2. The intelligent audit method for marine operation tasks according to claim 1 is characterized in that: The step of inputting at least one of the operation plan documents into the recognition model to obtain first operation information comprises the following steps: At least one operation plan document is input into the recognition model to identify and obtain the first operation information; wherein, the first operation information includes multiple sub-project information and the dependency information between each sub-project; each sub-project information includes at least the project identifier, the sub-project identifier, the operation area, the operation volume, the operation vessel and the expected operation days.

3. The intelligent audit method for marine operation tasks according to claim 1, characterized in that: The first intelligent audit model includes a first encoder, a second encoder, a first fuser, and a first-level decoder; the first-level decoder is composed of multiple network blocks and multiple layers of fully connected layers, each of which includes an LSTM layer, a RELU layer, and a pooling layer; The step of inputting the first operation diagram and the first operation table into a first intelligent audit model and outputting a first audit result comprises the following steps: Inputting the first job graph into the first encoder to obtain a first encoding vector; Inputting the first worksheet into the second encoder to obtain a second encoding vector; Inputting the first encoding vector and the second encoding vector into the first fuser to obtain a first fused vector; The first fusion vector is input into the first-level decoder to obtain the first audit result.

4. The intelligent audit method for marine operation tasks according to claim 3 is characterized in that: The second encoder includes an extraction sub-model and an encoding sub-model; Inputting the first operation table into the second encoder to obtain a second encoding vector includes the following steps: Inputting the first operation table into the extraction sub-model to obtain a target key information set; The target key information set and the weight information corresponding to the key information are input into the encoding sub-model to obtain a second encoding vector.

5. The intelligent audit method for marine operation tasks according to any one of claims 1 to 4, characterized in that: The method further comprises the following steps: Modify the first operation information and the first operation table according to the first audit result to obtain second operation information and a second operation table; inputting the second job information into the first job scheduling model to generate first scheduling information; generating a first scheduling graph according to the first scheduling information; The first scheduling diagram and the second operation table are input into a second intelligent audit model, and a second audit result is output.

6. An intelligent audit device for marine operation tasks, characterized in that: The device comprises: An acquisition module, used for acquiring at least one work plan document; a recognition module, configured to input at least one of the operation plan documents into a recognition model to obtain first operation information; A first generating module, configured to generate a first operation table according to the first operation information; A second generating module, configured to generate a first operation graph according to the first operation table; A first review module, configured to input the first operation diagram and the first operation table into a first intelligent review model and output a first review result; Generating a first operation table according to the first operation information includes the following steps: Generate the first operation table according to the preset template using the first operation information; A first job table is obtained using a parallel and / or multi-threaded method, specifically: pre-configuring a recognition model and a job table generation template on a cluster node; selecting, based on a load balancing algorithm, multiple nodes on the cluster node for job plan document recognition and job table generation as a target node set; sending multiple job plan documents to each target node in the target node set according to the number of job plan documents; each target node generates multiple first temporary job tables using the recognition model and the job table generation template; and concatenating the multiple first temporary job tables to obtain a first job table. Generating a first operation graph according to the first operation table specifically includes: extracting the first job information from the first job table; At least one operation branch is constructed with the sub-project identifier as the root node, the project identifier as the first-level leaf node, and the operation vessel as the second-level leaf node; constructing the first operation graph according to at least one of the operation branches; The step of constructing the first job graph according to at least one job branch comprises the following steps: Connecting the root nodes in the job branches with edges based on dependency information between the sub-projects; The first operation graph is constructed by taking the operation volume and operation area as the attributes of the first-level leaf nodes and taking the estimated operation days as the attributes of the second-level leaf nodes.

7. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images