Risk assessment method and system for ship classification society and classification society test based on large language model

Abstract

The invention belongs to the technical field of data processing, and provides a risk assessment method and system for a ship classification society and classification society inspection based on a large language model, and the method comprises the steps: S1, pre-training a large language model: pre-training the large language model based on a preset prompt word and a risk assessment data set, and obtaining a ship risk assessment large model; and S2, performing classification society risk and classification society inspection risk assessment on the ship based on the large language model to generate a first risk assessment result, a second risk assessment result and a third risk assessment result. According to the scheme, deep processing is carried out on existing ship operation information in combination with various third-party data according to a preset risk rule mainly by calling a large language model, corresponding risk scores are accurately calculated, and risk assessment suggestions and countermeasures with reference values are generated.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and in particular relates to a method and system for constructing an expected credit loss model for the shipping leasing industry. Background Art

[0002] Traditional manual assessments offer high accuracy and comprehensive recommendations, but the sheer volume of ship data makes them impractical to perform with human effort alone. Existing machine learning methods calculate scores based on the ship's classification society and the risk rules used by the classification society's surveys. However, scoring rules vary for different ship factors, requiring separate models for each assessment. Furthermore, machine learning methods cannot accurately provide risk assessment recommendations tailored to each ship's circumstances. Summary of the Invention

[0003] In order to solve the technical problem that due to the huge amount of ship data, it is impossible to conduct real-time risk assessment of ship classification societies and classification society inspections, the present invention proposes a risk assessment method and system for ship classification societies and classification society inspections based on a large language model. The large language model is used to evaluate ship data. Only by modifying the prompt words can the scores of different factors be calculated and evaluation suggestions be obtained, which greatly reduces manpower consumption.

[0004] The specific plan is as follows:

[0005] A risk assessment method for ship classification societies and classification society inspections based on a large language model,

[0006] S1: Pre-training a large language model: Contextually train the large language model based on preset prompt words and a risk assessment dataset to obtain a large model for ship risk assessment; the preset prompt words are generated based on classification society risk rules and classification society inspection risk rules; the preset prompt words include: risk rule prompt words, calculation score prompt words, and risk assessment prompt words;

[0007] S2: Conduct classification society risk and classification society inspection risk assessment on the ship based on the large language model: the user inputs basic information of the ship into the ship risk assessment large model described in S1, and in the first processing stage of the ship risk assessment large model, generates a first risk assessment result of the classification society risk and the classification society inspection risk based on the risk rule prompt words and the basic information of the ship; uses the first risk assessment result as input to the second processing stage of the ship risk assessment large model, and calculates risk scores of the classification society risk and the classification society inspection risk respectively based on the calculation score prompt words as the second risk assessment result; uses the first risk assessment result as input to the third processing stage of the ship risk assessment large model, and generates assessment suggestions of the classification society risk and the classification society inspection risk based on the risk assessment prompt words as the third risk assessment result;

[0008] S3: User display: Display the second risk assessment result and the third risk assessment result on the user interface.

[0009] Preferably, the preset prompt words are added or updated based on preset risk rules.

[0010] Preferably, the risk rule prompt words are input into the large language model from the classification society's risk rules and the classification society's inspection risk rules, and the large language model is generated after polishing.

[0011] Preferably, in step S1, the risk assessment data set includes: basic information of the ship input by the user and a fixed-format output result that the large language model is expected to answer based on the basic information of the ship; the output result includes: a first risk assessment result, a second risk assessment result and a third risk assessment result.

[0012] Preferably, the risk rule prompt words include: guiding the large language model to extract relevant content of the classification society risk factor and classification society inspection risk factor of the corresponding ship from the database based on the user query content, and generate a first risk assessment result in a fixed format; the classification society risk factor is set based on the classification society risk rule, and the classification society inspection risk factor is set based on the classification society inspection risk rule.

[0013] Preferably, the first risk assessment result includes the private data of the ship and is not displayed to the user, while the second risk assessment result and the third risk assessment result hide the private data of the ship and are displayed to the user.

[0014] Preferably, in step S2, the basic ship information input into the ship risk assessment model includes at least one of: ship name or MMSI number.

[0015] Preferably, the database includes: ship name, MMSI number, classification society risk factor, classification society inspection risk factor and corresponding values.

[0016] Preferably, the calculation of the scoring prompt words includes: guiding the large language model to calculate the classification society score and the classification society inspection score of the ship based on the first risk assessment result, and judging whether to generate a risk prompt according to the score, and generating a second risk assessment result in a fixed format; the calculation of the scoring prompt words is based on the scoring rules in the classification society risk rules and the classification society inspection risk rules and whether the risk prompt is generated based on the scoring rules.

[0017] Preferably, the risk assessment prompt word includes: guiding the large language model to generate a risk assessment suggestion as a third risk assessment result based on the first risk assessment result.

[0018] A risk assessment system for ship classification societies and classification society inspections based on a large language model, including:

[0019] The pre-trained large language model module includes a preset prompt word submodule and a model pre-training submodule. The model pre-training submodule pre-trains a large language model based on preset prompt words and a risk assessment dataset to obtain a large ship risk assessment model. The preset prompt word submodule generates preset prompt words based on classification society risk rules and classification society inspection risk rules. The preset prompt words include risk rule prompt words, calculation score prompt words, and risk assessment prompt words.

[0020] Ship risk assessment module: A large ship risk assessment model is constructed based on the classification society risk and classification society inspection risk assessment of the ship, including: a first processing submodule, a second processing submodule and a third processing submodule;

[0021] The first processing submodule generates a first risk assessment result of classification society risk and classification society inspection risk based on the risk rule prompt word and the basic information of the ship input by the user;

[0022] The second processing submodule takes the first risk assessment result as input and calculates risk scores of the classification society risk and the classification society inspection risk respectively as second risk assessment results based on the calculation score prompt words;

[0023] The third processing submodule takes the first risk assessment result as input and generates assessment suggestions of classification society risk and classification society inspection risk as a third risk assessment result based on the risk assessment prompt words;

[0024] User display interface: used to display the second risk assessment results and the third risk assessment results.

[0025] Beneficial effects

[0026] This paper proposes a risk assessment method and system for ship classification societies and classification society inspections based on a large language model. This solution primarily utilizes the large language model to combine existing ship operation information with various third-party data, deeply processing it according to pre-set risk rules. This method accurately calculates corresponding risk scores and generates valuable risk assessment recommendations and countermeasures. The accuracy of risk assessment for classification societies reached 91%, and for classification society inspections reached 89%.

[0027] Compared with the traditional method of relying on manual evaluation and suggestions, the present invention uses a large language model to significantly improve work efficiency and greatly reduce the investment and waste of human resources. Compared with existing machine learning calculation and scoring technologies, due to the differences in ship-related factors, the matching rules are also different, and it is often necessary to build multiple specialized machine learning algorithm models to deal with different situations. The large language model used in the present invention only needs to flexibly adjust the input preset prompt words and generate corresponding preset prompt words based on different risk rules. It can complete the risk scoring calculation task for different ship factors and rules under the same model framework. At the same time, the large language model also has the function of outputting risk assessment suggestions, further expanding the scope of application and practicality.

[0028] In the present invention, the large model of ship risk assessment is used as a core component. By setting different processing stages in the large model, it can accurately analyze the complex relationship between ship information and risk rules, thereby achieving efficient and accurate risk assessment. Different processing stages are divided in the large model of ship risk assessment. In the first processing stage, a first risk assessment result of classification society risk and classification society inspection risk is generated based on the risk rule prompt words and the basic information of the ship; the first risk assessment result is used as the input of the second processing stage, and the risk scores of classification society risk and classification society inspection risk are calculated based on the calculation score prompt words as the second risk assessment result; the first risk assessment result is used as the input of the third processing stage, and the assessment suggestions of classification society risk and classification society inspection risk are generated based on the risk assessment prompt words as the third risk assessment result. By splitting the prompt words into different stages for processing, the word length of a single question and answer of the large model is reduced, and the accuracy of the output result of the large language model is avoided to be reduced due to the excessive length of a single input.

[0029] Secondly, in the first processing stage, to improve the accuracy of the results, the display calculation of ship data, including private ship data, is performed. This private data is output as the result of the first processing stage to the second and third processing stages to obtain risk scores and risk assessment recommendations. However, the second and third risk assessment results do not display private data, thus effectively preventing the leakage of this private data.

[0030] At the same time, the large-scale ship risk assessment model utilizes multiple rounds of dialogue rather than a single conversation, separating risk scoring from risk assessment opinions. This significantly improves the accuracy of risk scoring and the rationality of assessment recommendations. This approach not only optimizes the assessment process but also enhances the reliability and practicality of the assessment results, providing strong technical support for ship risk management. This paper examines the two risk indicators—the ship's classification society and the classification society's inspection—to form a comprehensive ship risk assessment method and system based on a large artificial intelligence language model.

[0031] The large ship risk assessment model of the present invention processes the second and third processing stages in parallel, and uses a concurrent approach to increase the output speed by about 5 times. Compared with the existing technology, the concurrent approach of the present invention greatly optimizes the assessment efficiency, meets the needs of rapid ship risk assessment, and is an important innovative means to improve assessment performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flow chart of a risk assessment method for ship classification societies and classification society inspections based on a large language model in an embodiment.

[0033] Figure 2 A structural diagram of a risk assessment system for ship classification societies and classification society inspections based on a large language model in an embodiment.

[0034] Figure 3 A data flow diagram of a risk assessment system for ship classification societies and classification society inspections based on a large language model in an embodiment.

[0035] Figure 4 System user display interface data flow diagram in the embodiment.

[0036] Figure 5 A user interface rendering of a risk assessment system for ship classification societies and classification society inspections based on a large language model in an embodiment.

[0037] Figure 6 A diagram showing the second and third risk assessment results of some classification society inspections of a risk assessment system for ship classification societies and classification society inspections based on a large language model in an embodiment.

[0038] Figure 7 A diagram showing the second and third risk assessment results of some classification societies in a risk assessment system for ship classification societies and classification society inspections based on a large language model in an embodiment. DETAILED DESCRIPTION

[0039] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0040] like Figure 1 As shown in the figure, a risk assessment method for ship classification societies and classification society inspections based on a large language model is proposed.

[0041] S1: Pre-training a large language model: Pre-training a large language model based on preset prompt words and a risk assessment dataset to obtain a large model for ship risk assessment; the preset prompt words are generated based on classification society risk rules and classification society inspection risk rules; the preset prompt words include: risk rule prompt words, calculation score prompt words, and risk assessment prompt words;

[0042] S2: Conduct classification society risk and classification society inspection risk assessment on the ship based on the large language model: the user inputs basic information of the ship into the ship risk assessment large model described in S1, and in the first processing stage of the ship risk assessment large model, generates a first risk assessment result of the classification society risk and the classification society inspection risk based on the risk rule prompt words and the basic information of the ship; uses the first risk assessment result as input to the second processing stage of the ship risk assessment large model, and calculates risk scores of the classification society risk and the classification society inspection risk respectively based on the calculation score prompt words as the second risk assessment result; uses the first risk assessment result as input to the third processing stage of the ship risk assessment large model, and generates assessment suggestions of the classification society risk and the classification society inspection risk based on the risk assessment prompt words as the third risk assessment result;

[0043] S3: User display: Display the second risk assessment result and the third risk assessment result on the user interface.

[0044] Preferably, the preset prompt words are added or updated based on preset risk rules.

[0045] Preferably, the risk rule prompt words are input into the large language model from the classification society's risk rules and the classification society's inspection risk rules, and the large language model is generated after polishing.

[0046] As shown in Table 1, the classification society risk rules are:

[0047] Table 1

[0048]

[0049]

[0050]

[0051] *Note: For dual-class ships, the overall ship rating will be increased by one level, for example, B will be increased to B+.

[0052] The classification society risk rules also include additional points for classification society changes.

[0053] Based on the above classification society scores, additional points will be awarded to ships that frequently change classification societies. Classification society indicators Total score = classification society score + additional score for classification society changes , the calculation rules for additional points of classification society changes are as follows:

[0054] Table 2

[0055]

[0056] *Note: If a ship joins multiple classification societies at the same time, and one classification society changes while the others remain unchanged, it is not considered a classification society change and the above change rules are not applicable.

[0057] The classification society survey risk rules include:

[0058] 1. Risk Type

[0059] (1) Dock inspection (dock inspection)

[0060] Surveys conducted in dry dock or on the raft are called dock surveys. During the special survey cycle every five years, at least two dock surveys should be conducted. One of these surveys should be conducted in conjunction with the special survey. In all cases, the interval between any two surveys should not exceed 36 months.

[0061] (2) Special inspection (special inspection)

[0062] The hull and machinery (including electrical equipment) are normally to be specially surveyed at intervals of 5 years in order to renew the classification certificate.

[0063] (3) Shelf inspection

[0064] During the lay-up period, all overdue post-construction surveys (including dry dock surveys and special surveys) will be extended to the date of re-operation.

[0065] 2. Rule Logic

[0066] Based on the above definitions of various tests and the existing data, the indicator scoring rules are formulated as shown in Table 3:

[0067] Table 3

[0068]

[0069] Preferably, in step S1, the risk assessment data set includes: basic information of the ship input by the user and a fixed-format output result that the large language model is expected to answer based on the basic information of the ship; the output result includes: a first risk assessment result, a second risk assessment result and a third risk assessment result.

[0070] Preferably, the risk rule prompt words include: guiding the large language model to extract relevant content of the classification society risk factor and classification society inspection risk factor of the corresponding ship from the database based on the user query content, and generate a first risk assessment result in a fixed format; the classification society risk factor is set based on the classification society risk rule, and the classification society inspection risk factor is set based on the classification society inspection risk rule.

[0071] Preferably, the first risk assessment result includes ship privacy data and is not displayed to the user, and the second risk assessment result and the third risk assessment result are risk assessment results, do not include ship privacy data, and are displayed to the user.

[0072] Preferably, in step S2, the basic ship information input into the ship risk assessment model includes at least one of: ship name or MMSI number.

[0073] Preferably, the database includes: ship name, MMSI number, classification society risk factor, classification society inspection risk factor and corresponding values.

[0074] The collection content and storage format of the database are as follows:

[0075] (1) Data collection: First, data collection is carried out, covering all kinds of relevant information of the ship.

[0076] (2) Data classification and storage format: The collected data is accurately classified according to the imo number and ship name. For classification society risk-related factors, such as the ship flag, classification society, and the number of countries passed through each year, the json format is used for storage. The example is as follows:

[0077] {"ship_flag":"XX","classification_society":"XX"}

[0078] Among them, "ship_flag" represents the ship flag and "classification_society" represents the classification society. This JSON format facilitates structured storage and subsequent calls of data.

[0079] Classification society inspection risk factor storage: Classification society inspection risk-related factors, such as docking inspection date, special inspection date, etc., are also stored in JSON format. The example is as follows:

[0080] {"dockingSurvey":"XXXX-XX-XX","specialSurvey":"XXXX-XX-XX"}

[0081] “DockingSurvey” indicates the date of docking survey, and “specialSurvey” indicates the date of special survey, ensuring the clear presentation of survey-related data.

[0082] In addition to the JSON format, other structured data storage formats such as XML can also be used. The choice is based on the actual data characteristics and system compatibility. For example, the XML format may be more advantageous when processing complex hierarchical data, as shown below:

[0083] <ship><ship_flag> XX< / ship_flag><classification_society> XX< / classification_society><ann ual_countries_passed> XX< / annual_countries_passed>< / ship>

[0084] Preferably, the calculation of the scoring prompt words includes: guiding the large language model to calculate the classification society score and the classification society inspection score of the ship based on the first risk assessment result, and judging whether to generate a risk prompt according to the score, and generating a second risk assessment result in a fixed format; the calculation of the scoring prompt words is based on the scoring rules in the classification society risk rules and the classification society inspection risk rules and whether the risk prompt is generated based on the scoring rules.

[0085] Preferably, the risk assessment prompt word includes: guiding the large language model to generate a risk assessment suggestion as a third risk assessment result based on the first risk assessment result.

[0086] like Figure 2-4 As shown, a risk assessment system for ship classification societies and classification society inspections based on a large language model includes:

[0087] The pre-trained large language model module includes a preset prompt word submodule and a model pre-training submodule. The model pre-training submodule performs contextual training on the large language model based on the preset prompt words and the risk assessment dataset to obtain a large ship risk assessment model. The preset prompt word submodule generates preset prompt words based on the classification society's risk rules and classification society's inspection risk rules. The preset prompt words include risk rule prompt words, calculation score prompt words, and risk assessment prompt words.

[0088] Ship risk assessment module: A large ship risk assessment model is constructed based on the classification society risk and classification society inspection risk assessment of the ship, including: a first processing submodule, a second processing submodule and a third processing submodule;

[0089] The first processing submodule generates a first risk assessment result of classification society risk and classification society inspection risk based on the risk rule prompt word and the basic information of the ship input by the user;

[0090] The second processing submodule takes the first risk assessment result as input and calculates risk scores of the classification society risk and the classification society inspection risk respectively as second risk assessment results based on the calculation score prompt words;

[0091] The third processing submodule takes the first risk assessment result as input and generates assessment suggestions of classification society risk and classification society inspection risk as a third risk assessment result based on the risk assessment prompt words;

[0092] like Figure 5-7 As shown, the user display interface is used to display the second risk assessment result and the third risk assessment result.

[0093] Example 2: Taking a specific ship as an example:

[0094] Vessel Information

[0095] [{"endDate":"20231031","list":[{"classsociety":"LR","dockingSurvey":"2027-05-23","specia lSurvey":"2029-05-23"}],"riskLevel":1.0,"startDate":"20230930"}]

[0096] (1) Pre-training large language model:

[0097] Question: The risk rules are as follows: The scoring rules are as follows: Special surveys due after the assessment date are assigned a score of 7 and a risk indication; docking surveys due after the assessment date are assigned a score of 6, also indicating a risk indication; no data from both special and docking surveys is assigned a score of 5 and a risk indication; both special and docking surveys are not overdue or at least one of the surveys was conducted normally is assigned a score of 1, with no risk indication. Ships not affiliated with a classification society are assigned a score of 1 and may be non-operating or not requiring classification. There is a ship with class number 9591698 and name SHIDAI 8. Its information is as follows: [{{"endDate":"20231031","list":[{{"classsociety":"CC","dockingSurvey":"2025-09-09","specialS urvey":"2027-09-17"}}],"riskLevel":1.0,"startDate":"20230930"}}]. Among them, startdate and enddate are invalid data. Now, according to the above rules and ship information, assuming that the evaluation date is 2024-11-01, it is expired three months before the evaluation date, that is, before 2024-08-01. Please compare the docking survey and special survey with the evaluation date, give the classification society survey risk score, and give the classification society survey risk assessment of the ship and provide the current ship's classification society and ship name.

[0098] Answer: "According to the vessel's data, the special survey (SS) is due on September 17, 2027, the dry dock survey (DOCK) is due on September 9, 2025, and the assessment date is November 1, 2024. Both the SS and the dry dock survey are within the SS and DOCK survey deadlines. According to the scoring rules, both the SS and the dry dock survey are normal, with a risk level of 1, and no risk warning is required. The vessel has been registered with CC Classification Society, the survey data is complete, and the subsequent survey arrangements are reasonable, indicating that the vessel is in a compliant operating state."

[0099] S2: Classification society risk and classification society inspection risk assessment of ships based on large language models:

[0100] (1) The user inputs basic ship information into the ship risk assessment model described in S1: There is a ship whose ship class number is {name}, whose ship name is {row['ship_name']} and whose information is as follows: {row['boat']}. Based on the document rules and the above ship information, the classification society survey risk of the ship is evaluated.

[0101] (2) Classification Society Survey Risk Assessment in the First Processing Phase

[0102] ①Refine and query ship information

[0103] **Vessor Information:**

[0104] -Ship Class Number: 9938822

[0105] -Ship name: CL YINGTAN

[0106] -Classification Society: LR

[0107] -Dock inspection (docking inspection) due date: 2027-05-23

[0108] -Special inspection (special inspection) due date: 2029-05-23

[0109] -Evaluation date: 2024-11-01

[0110] ② Generate assessment rules based on risk rule prompt words:

[0111] - Both the special inspection and dock inspection are not overdue or at least one of them is carried out normally, with a score of 1 and no risk warning.

[0112] - If the special inspection due date is after the assessment date, the score will be 7 and the risk will be indicated.

[0113] - If the dock inspection due date is after the assessment date, a score of 6 is given, which also indicates risk.

[0114] - Both the special inspection and dock inspection have no data and are rated 5 points and indicate risks.

[0115] - Vessels not affiliated with a classification society are assigned a score of 1 and may be non-operating vessels or vessels not requiring classification.

[0116] ③ The large model generates evaluation steps based on risk rule prompt words:

[0117] 1. **Confirmed evaluation date: **2024-11-01

[0118] 2.**Check the due dates for docking inspection and special inspection:**

[0119] -Dock inspection due date: 2027-05-23

[0120] -Special inspection due date: 2029-05-23

[0121] 3. **Compare Due Date to Valuation Date:**

[0122] -Docking inspection: 2027-05-23 > 2024-11-01, not expired

[0123] - Special inspection: 2029-05-23 > 2024-11-01, not expired

[0124] 4. **Scoring according to the rules:**

[0125] - Both the special inspection and dock inspection are not overdue, which is normal status. The risk level is 1 point and there is no risk warning.

[0126] ④The risk rule prompt word generates the first risk assessment result:

[0127] -**Risk score:** 1 point

[0128] -**Risk Warning:** No risk warning

[0129] -**Ship Classification Society Survey Risk Assessment:** The vessel has joined the LR classification society, the survey data is complete, the special survey and dock survey are not overdue, and the subsequent survey arrangements are reasonable, and it is in a compliant operating status.

[0130] -**Classification Society:**LR

[0131] -**Ship Name:**CL YINGTAN

[0132] Based on the above assessment, the ship's classification society survey risk rating is 1 point, and it is currently in a low-risk status.

[0133] 2. The second processing stage of the ship risk assessment model for classification society inspection risk assessment:

[0134] (1) Inputs include: the results of the first risk assessment in the first processing phase;

[0135] **Vessor Information:**

[0136] -Ship Class Number: 9938822

[0137] -Ship name: CL YINGTAN

[0138] -Classification Society: LR

[0139] -Dock inspection (docking inspection) due date: 2027-05-23

[0140] -Special inspection (special inspection) due date: 2029-05-23

[0141] -Evaluation date: 2024-11-01

[0142] (2) Generate and refine scoring rules based on the calculated scoring prompt words:

[0143] - Both the special inspection and dock inspection are not overdue or at least one of them is carried out normally, with a score of 1 and no risk warning.

[0144] - If the special inspection due date is after the assessment date, the score will be 7 and the risk will be indicated.

[0145] - If the dock inspection due date is after the assessment date, a score of 6 is given, which also indicates risk.

[0146] - Both the special inspection and dock inspection have no data and are rated 5 points and indicate risks.

[0147] - Vessels not affiliated with a classification society are assigned a score of 1 and may be non-operating vessels or vessels not requiring classification.

[0148] (3) Evaluation steps based on calculating scoring prompt words:

[0149] 1. **Confirmed evaluation date: **2024-11-01

[0150] 2.**Check the due dates for docking inspection and special inspection:**

[0151] -Dock inspection due date: 2027-05-23

[0152] -Special inspection due date: 2029-05-23

[0153] 3. **Compare Due Date to Valuation Date:**

[0154] -Docking inspection: 2027-05-23 > 2024-11-01, not expired

[0155] - Special inspection: 2029-05-23 > 2024-11-01, not expired

[0156] 4. **Scoring according to the rules:**

[0157] - Both the special inspection and dock inspection are not overdue, which is normal status. The risk level is 1 point and there is no risk warning.

[0158] (4) Generate a second risk assessment result based on the calculated scoring prompt words:

[0159] -**Risk score:** 1 point

[0160] -**Risk Warning:** No risk warning

[0161] -**Ship Classification Society Survey Risk Assessment:** The vessel has joined the LR classification society, the survey data is complete, the special survey and dock survey are not overdue, and the subsequent survey arrangements are reasonable, and it is in a compliant operating status.

[0162] -**Classification Society:**LR

[0163] -**Ship Name:**CL YINGTAN

[0164] Based on the above assessment, the vessel's classification society survey risk rating is 1, placing it in a low-risk state. Based on the above, please find the vessel's classification society survey score, outputting only the numerical value.

[0165] 3. The third processing stage of the ship risk assessment model:

[0166] (1) Risk assessment recommendation input includes: the first risk assessment results of the first processing stage;

[0167] ###Classification Society Inspection Risk Assessment

[0168] **Vessor Information:**

[0169] -Ship Class Number: 9938822

[0170] -Ship name: CL YINGTAN

[0171] -Classification Society: LR

[0172] -Dock inspection (docking inspection) due date: 2027-05-23

[0173] -Special inspection (special inspection) due date: 2029-05-23

[0174] -Evaluation date: 2024-11-01

[0175] (2) Evaluation rules based on risk assessment prompt words:

[0176] - Both the special inspection and dock inspection are not overdue or at least one of them is carried out normally, with a score of 1 and no risk warning.

[0177] - If the special inspection due date is after the assessment date, the score will be 7 and the risk will be indicated.

[0178] - If the dock inspection due date is after the assessment date, a score of 6 is given, which also indicates risk.

[0179] - Both the special inspection and dock inspection have no data and are rated 5 points and indicate risks.

[0180] - Vessels not affiliated with a classification society are assigned a score of 1 and may be non-operating vessels or vessels not requiring classification.

[0181] (3) Assessment steps based on risk assessment prompts:

[0182] 1. **Confirmed evaluation date: **2024-11-01

[0183] 2.**Check the due dates for docking inspection and special inspection:**

[0184] -Dock inspection due date: 2027-05-23

[0185] -Special inspection due date: 2029-05-23

[0186] 3. **Compare Due Date to Valuation Date:**

[0187] -Docking inspection: 2027-05-23 > 2024-11-01, not expired

[0188] - Special inspection: 2029-05-23 > 2024-11-01, not expired

[0189] 4. **Scoring according to the rules:**

[0190] - Both the special inspection and dock inspection are not overdue, which is normal status. The risk level is 1 point and there is no risk warning.

[0191] 5. Generate assessment suggestions based on the risk assessment prompt words and preliminary scoring results

[0192] -**Risk score:** 1 point

[0193] -**Risk Warning:** No risk warning

[0194] -**Ship Classification Society Survey Risk Assessment:** The vessel has joined the LR classification society, the survey data is complete, the special survey and dock survey are not overdue, and the subsequent survey arrangements are reasonable, and it is in a compliant operating status.

[0195] -**Classification Society:**LR

[0196] -**Ship Name:**CL YINGTAN

[0197] Based on the above assessment, the vessel's classification society survey risk rating is 1, placing it in a low-risk state. Based on the above information and the vessel's information, please provide recommendations for the vessel's classification society survey risk assessment. Please output the results in a single paragraph.

[0198] (5) Generating the third risk assessment result: It is recommended to maintain the existing classification society survey management for this ship, ensure the timely execution of special surveys and docking surveys, and regularly review the survey plan to respond to any potential changes. Considering the normal status of the classification society survey, this indicates that the ship is currently in good technical condition and compliance, and no immediate action is required. However, it should continue to pay attention to the classification society's survey requirements and the latest guidelines to maintain its low risk status.

[0199] Finally, it should be noted that the above description is only used to illustrate the technical solution of the present invention and is not intended to limit it. Although the present invention has been described in detail with reference to the preferred arrangement scheme, those skilled in the art should understand that the technical solution of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A risk assessment method for ship classification societies and classification society inspections based on a large language model, characterized in that: S1: Pre-training large language model: Contextual training of the large language model is performed based on preset prompt words and risk assessment datasets to obtain a large model for ship risk assessment; The preset prompt words are generated based on classification society risk rules and classification society inspection risk rules; The preset prompt words include: risk rule prompt words, calculation score prompt words and risk assessment prompt words; S2: Conduct classification society risk and classification society inspection risk assessment on the ship based on the large language model: the user inputs basic information of the ship into the ship risk assessment large model described in S1, and in the first processing stage of the ship risk assessment large model, generates a first risk assessment result of the classification society risk and the classification society inspection risk based on the risk rule prompt words and the basic information of the ship; uses the first risk assessment result as input to the second processing stage of the ship risk assessment large model, and calculates risk scores of the classification society risk and the classification society inspection risk respectively based on the calculation score prompt words as the second risk assessment result; uses the first risk assessment result as input to the third processing stage of the ship risk assessment large model, and generates assessment suggestions of the classification society risk and the classification society inspection risk based on the risk assessment prompt words as the third risk assessment result; S3: User display: Display the second risk assessment result and the third risk assessment result on the user interface.

2. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: The preset prompt words are added or updated based on preset risk rules.

3. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: The risk rule prompt words are input into the large language model from the classification society's risk rules and the classification society's inspection risk rules, and the large language model is generated after polishing.

4. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: In step S1, the risk assessment data set includes: basic ship information input by the user and a fixed-format output result expected from the large language model for the basic ship information; the output result includes: a first risk assessment result, a second risk assessment result, and a third risk assessment result.

5. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: The risk rule prompt words include: guiding the large language model to extract relevant content of the classification society risk factor and the classification society inspection risk factor of the corresponding ship from the database based on the user query content, and generate a first risk assessment result in a fixed format; the classification society risk factor is set based on the classification society risk rules, and the classification society inspection risk factor is set based on the classification society inspection risk rules.

6. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: The first risk assessment result includes the private data of the ship and is not displayed to the user. The second risk assessment result and the third risk assessment result hide the private data of the ship and are displayed to the user.

7. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: In step S2, the basic ship information input into the ship risk assessment model includes at least one of: ship name or MMSI number.

8. The risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 5 is characterized in that: The database includes: ship name, MMSI number, classification society risk factor, classification society inspection risk factor and corresponding values.

9. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: The calculation of the scoring prompt words includes: guiding the large language model to calculate the classification society score and the classification society inspection score of the ship based on the first risk assessment result, and determining whether to generate a risk prompt based on the score, and generating a second risk assessment result in a fixed format; the calculation of the scoring prompt words is based on the scoring rules in the classification society risk rules and the classification society inspection risk rules and whether the risk prompt is generated based on the scoring rules.

10. A risk assessment method for ship classification societies and classification society inspections based on a large language model according to claim 1, characterized in that: The risk assessment prompt word includes: guiding the large language model to generate a risk assessment suggestion as a third risk assessment result based on the first risk assessment result.

11. A risk assessment system for ship classification societies and classification society inspections based on a large language model, characterized in that: The method according to any one of claims 1 to 10 comprises: The pre-trained large language model module includes a preset prompt word submodule and a model pre-training submodule. The model pre-training submodule performs contextual training on the large language model based on the preset prompt words and the risk assessment dataset to obtain a large ship risk assessment model. The preset prompt word submodule generates preset prompt words based on the classification society's risk rules and classification society's inspection risk rules. The preset prompt words include risk rule prompt words, calculation score prompt words, and risk assessment prompt words. Ship risk assessment module: A large ship risk assessment model is constructed based on the classification society risk and classification society inspection risk assessment of the ship, including: a first processing submodule, a second processing submodule and a third processing submodule; The first processing submodule generates a first risk assessment result of classification society risk and classification society inspection risk based on the risk rule prompt word and the basic information of the ship input by the user; The second processing submodule takes the first risk assessment result as input and calculates risk scores of the classification society risk and the classification society inspection risk respectively as second risk assessment results based on the calculation score prompt words; The third processing submodule takes the first risk assessment result as input and generates assessment suggestions of classification society risk and classification society inspection risk as a third risk assessment result based on the risk assessment prompt words; User display interface: used to display the second risk assessment results and the third risk assessment results.

12. A risk assessment system for ship classification societies and classification society inspections based on a large language model, characterized in that: The system adopts a parallel strategy and outputs the risk assessment results of the large ship risk assessment model through a multi-threaded mode.

Images