Operation risk assessment method for few-person conditional autonomous navigation ship in different operation modes

Abstract

The invention discloses an operation risk assessment method for a few-person conditional autonomous navigation ship in different operation modes, which is used for analyzing an operation system control structure of the few-person conditional autonomous navigation ship in different operation modes, and identifying unsafe control behaviors and cause scenes thereof by using a system theoretical process analysis method. Defining an analysis purpose; the method comprises the following steps: selecting a cause scene of an operation unsafe control behavior of a few-person conditional autonomous navigation ship in a remote control mode as an input node, the unsafe control behavior as an intermediate node, and an operation risk level corresponding to the remote control mode adopted by the ship as an output node, and constructing a Bayesian network model; and performing parameter learning and probability calculation, and dynamically evaluating the operation risk. And on the basis of the Bayesian network model, index parameters are adjusted according to requirements, the posterior probability of the adjusted target node is calculated, and the operation risk level is evaluated.

Description

Technical Field

[0001] The present invention relates to the field of ship risk analysis, and in particular to a method for assessing the operational risks of a conditionally autonomous ship with minimal human traffic under different operating modes. Background Art

[0002] Conditionally autonomous ships with fewer crew members are the product of the deep integration of the shipping industry and modern information technology. As an emerging development in the shipping sector, their operational systems encompass multiple aspects, including ship automation systems, crew operations, and shore-based control centers. The interactions between these components create a high level of complexity. Currently, research on conditionally autonomous ships with fewer crew members is still in its developmental stages, particularly regarding their operations, where a comprehensive and scientific analytical methodology has yet to be established.

[0003] Traditional ship risk analysis methods are primarily based on experience and statistical laws, focusing on static descriptions and analyses of basic ship navigation operations and crew responsibilities. However, these methods have significant limitations when applied to ships with reduced crew capacity. They tend to overlook the interactions and synergies between various systems and insufficiently consider the ship's adaptability and emergency response capabilities in complex dynamic environments, making them difficult to effectively apply to the actual conditions of ships with reduced crew capacity. Furthermore, existing research lacks a dynamic quantification method for systemic risks during multi-mode switching, particularly in the analysis of the coupling effects of the dynamic allocation of human and machine rights and responsibilities and cross-system failure propagation. Therefore, there is an urgent need to develop an operational risk assessment method for ships with reduced crew capacity under different operating modes to ensure the safe operation of ships with reduced crew capacity under different operating modes and promote the intelligent development of the shipping industry. Summary of the Invention

[0004] In order to solve the operational difficulties currently faced by conditionally autonomous ships with fewer crew members, the purpose of the present invention is to provide a method for assessing the operational risks of conditionally autonomous ships with fewer crew members under different operating modes, so as to achieve precise control of the operational risks of conditionally autonomous ships with fewer crew members under different operating modes.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A method for assessing the operational risk of conditionally autonomous ships with reduced crew capacity under different operating modes includes the following steps:

[0007] S1. Analyze the operational system control structure of a conditionally autonomous vessel with reduced crew capacity under different operating modes, and apply systems theory and process analysis to identify unsafe control behaviors and their causal scenarios. Define the analysis objective.

[0008] Clarify the losses that may occur during the operation of ships with reduced crew and conditional autonomous navigation; identify the system-level hazards that lead to the said losses; determine the safety constraints that the system needs to meet to avoid the occurrence of hazards; and further refine and classify the identified hazards.

[0009] Build control structures.

[0010] Clarify the various components of the ship operation system with reduced manpower and conditional autonomous navigation (ship automation system, crew operation, shore-based control center, etc.); clarify the interaction and control logic between the various components; and clarify the distribution of rights and responsibilities of each party under different operating modes.

[0011] Comprehensively identify unsafe control behaviors during ship operations, including situations where control instructions are erroneous, missing, delayed, or inappropriate, and conduct in-depth analysis of the causes of unsafe control behaviors to develop causal scenarios.

[0012] S2. Select the causal scenarios of unsafe control behaviors in the operation of unmanned conditionally autonomous ships under different operating modes as input nodes, unsafe control behaviors as intermediate nodes, and the operational risk levels corresponding to different operating modes adopted by ships as output nodes. Construct a Bayesian network model, perform parameter learning and probability calculation, and dynamically evaluate operational risks.

[0013] Determine the node variables and hierarchical relationships of the Bayesian network.

[0014] Input node definition: Based on the causal scenario identified by S1, it is converted into the root node of the Bayesian network.

[0015] Intermediate node construction: The unsafe control actions (UCAs) obtained based on the system theory process analysis method are used as intermediate nodes to establish a causal relationship with the input nodes.

[0016] Output node setting: Based on the different operating modes classified by China Classification Society for autonomous ships (such as assisted navigation, remote control, remote supervision, and full autonomy), define the output nodes of the operational risk level and quantify the risk level. The risk level can be divided into different states such as safe, relatively safe, and dangerous.

[0017] Carry out causal relationship mapping, establish a directed acyclic graph between nodes based on the causal chain obtained by the system theory process analysis method, and construct a Bayesian network topology structure.

[0018] The probability of occurrence of causal scenarios is obtained from historical ship data, and the prior probability of input nodes is calculated. For nodes lacking historical data, expert opinions are aggregated through the fuzzy Delphi method, and qualitative descriptions are converted into probability intervals to obtain the prior probability of input nodes.

[0019] Perform posterior probability inference based on Bayesian theorem and assess risk levels in real time.

[0020] With reference to historical ship data and combined with the experience of experts in the shipping field, the Bayesian network model is trained on parameters. Based on the real-time data of the input nodes, the probability of ship operation risks being at different levels under different operating modes is calculated.

[0021] Determine the nodes and structure of the Bayesian network model: Assume that there are n nodes in the Bayesian network model, represented by X1, X2, ..., X n , where X n is the output node representing the ship operation risk level, and the other nodes are input nodes.

[0022] The dependency relationship between nodes is represented by a directed acyclic graph, where each node X i Has its parent node set Pa(X i ).

[0023] Determine the conditional probability table (CPT) of each node: For each node X i , its conditional probability table P(X i |Pa(X i ) indicates that at its parent node Pa(X i ) takes different values, node X i The probability of taking different values.

[0024] Calculate the output node X using the total probability formula n The prior probability of being in different states.

[0025] Assume that the input nodes X1, X2, ..., X n-1 The different values ​​or states of constitute a partition of the sample space E1, E2, ..., E n , for the output node X n For a certain risk level A, there are:

[0026]

[0027] Among them, P(E i ) is event E i The probability of occurrence, that is, the various causes or conditions that affect the risk level of ship operation E i The probability of occurrence can be determined through previous statistical data, equipment maintenance records and other information; P(A|E i ) is in E i The probability of event A occurring under the conditions of occurrence, that is, the probability of ship operation risk being at level A given certain influencing factors.

[0028] S3, based on the Bayesian network model, adjusting the index parameters according to the demand, calculating the posterior probability of the adjusted target node, and evaluating the ship operation risk level under the operation mode change.

[0029] According to the actual demand of ship operation, the specific input node index parameters in the Bayesian network model are adjusted, and after adjustment, the Bayesian network model is re-run.

[0030] The same Bayesian inference algorithm and probability calculation method as in S2 are used to calculate the probability of the adjusted ship operation risk at different levels. Mainly calculate the posterior probability P(X n |E) of the output node when the new data source (observation data) E appears.

[0031] Joint probability calculation: for a Bayesian network with n nodes X1, X2, …, X n , the joint probability distribution can be expressed as the product of the conditional probabilities of each node under the condition of its parent nodes, that is:

[0032]

[0033] Where Pa(X i ) represents the parent node set of node X i .

[0034] Assume that the real-time data of the input node is:

[0035]

[0036] Then we can get:

[0037]

[0038] Posterior probability calculation: when a new data source (observation data) E appears, according to Bayes' theorem, the formula is obtained:

[0039]

[0040] Where P(X n ) is the prior probability of event X n , that is, the probability of X n before observing data E; P(E|X n ) is the likelihood, which represents the probability of observing data E under the assumption that X n is true; P(E) is the probability of evidence E.

[0041] Since P(E) takes the same value for all X n , it can be calculated by P(E,X n =xnj ) is normalized to obtain the posterior probability P(X n =x nj |E), that is:

[0042]

[0043] S4. Use information entropy to conduct uncertainty assessment on the operational risk level results, analyze the confidence interval of the posterior probability of the operational risk level, and evaluate the reliability of the risk level determination.

[0044] Determine the random variable and its value range: take the ship operation risk as the random variable X, and determine the value space of the random variable X.

[0045] In practical applications, ship operation risks can be divided into different levels, such as safe, relatively safe, critical, dangerous, etc. These different levels constitute the value space χ.

[0046] Calculate the probability of the ship operation risk being in different states: Based on the Bayesian network model constructed by S3, the probability value is obtained through Bayesian reasoning and probability calculation.

[0047] Assume that X has k values, namely x1, x2, ..., x k , the corresponding probabilities are P(x1), P(x2),…, P(x k ), the probability value reflects the possibility of ship operation risk being at various levels under the current data and model structure.

[0048] Calculate using the information entropy formula: Substitute the probability values ​​of each value into the information entropy formula of ship operation risk X:

[0049]

[0050] Among them, P(x i ) is X and its value is x i probability.

[0051] If the H(X) value is large, it means that the uncertainty of the obtained ship operation risk level is high; if the H(X) value is small, it means that the certainty of the operation risk level is relatively high and the assessment result is relatively reliable.

[0052] The present invention provides a method for assessing the operational risk of a conditionally autonomous ship with few personnel under different operating modes. Compared with the prior art, the method has the following advantages:

[0053] By applying the system theory process analysis method to conduct an in-depth analysis of the control structure of the operation system of autonomous ships with few people, it is possible to comprehensively identify unsafe control behaviors and their causes during the ship operation process, covering various situations such as erroneous, missing, delayed or inappropriate control instructions, thereby constructing a scientific and comprehensive ship operation safety indicator system, making safety risk assessment more comprehensive and accurate.

[0054] Deeply integrate the system theory process analysis method with the Bayesian network to achieve dynamic mapping of "operation mode-causal scenario-risk level".

[0055] For the four modes defined by classification societies, namely assisted navigation, remote control, remote supervision, and full autonomy, differentiated network topologies are constructed by building a Bayesian network model.

[0056] The causal scenarios of unsafe control behaviors of conditionally autonomous ships with few people in different operating modes are selected as input nodes, and the operational risk level of the ship corresponding to the operating mode is used as the output node. Parameter learning and probability calculation are performed to dynamically evaluate the degree of operational risk. Based on the Bayesian network model, the present invention can adjust the index parameters as needed according to the actual needs of ship operation, and calculate the posterior probability of the adjusted target node to evaluate the operational risk level of conditionally autonomous ships with few people in different operating modes, thereby dynamically evaluating the operational risk. Compared with traditional methods, the evaluation results of the present invention are more accurate and reliable, and can provide strong data support and decision-making basis for ship safety management under different operating modes.

[0057] Furthermore, this dynamic adjustment mechanism enables the present invention to flexibly adapt to the safety assessment needs of ships in different operational scenarios and environmental conditions, demonstrating its strong adaptability and practicality. By calculating the information entropy value of the posterior probability of ship operational risk under different operating modes and providing confidence intervals for the posterior probability, the reliability of the risk level is assessed. The magnitude of the information entropy value intuitively reflects the uncertainty of the ship's operational risk level, thereby providing more comprehensive and in-depth risk information for ship safety management. This further enhances the reliability and credibility of risk assessment results, enabling ship operators to more accurately grasp the ship's safety status. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0059] Figure 1Flow chart for the method of the present application

[0060] Figure 2 Flow chart for the embodiment of the present application

[0061] Figure 3 Flow chart for the embodiment of the present application using the system theory process analysis method to analyze the unsafe control behavior of the few-people conditional autonomous navigation ship operation. DETAILED DESCRIPTION

[0062] In order to make the technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application:

[0063] The embodiment provides a few-people conditional autonomous navigation ship operation risk assessment method, sets a ship operation mode as a remote control mode, and accordingly expands the few-people conditional autonomous navigation ship operation risk level assessment analysis in the remote control mode based on a case background.

[0064] The embodiment specifically includes the following steps:

[0065] S1, analyzing the operation system control structure of the few-people conditional autonomous navigation ship in the remote control mode, and identifying unsafe control behavior and cause scenarios by using a system theory process analysis method.

[0066] In the embodiment, as shown in Figure 1 , there are four steps in total:

[0067] Specifically, step one: defining the analysis purpose.

[0068] In the embodiment, the ship operation mode is selected as the remote control mode, the losses occurring in the few-people conditional autonomous navigation ship operation process in the remote control mode are determined, the system level hazards causing the losses are identified, the safety constraint conditions required for the system to avoid the hazards are determined, and the identified hazards are further refined and classified.

[0069] Step two: establishing a hierarchical control structure model.

[0070] The components (ship automation system, crew operation, shore-based control center, etc.) of the few-people conditional autonomous navigation ship operation system in the remote control mode are determined, the interaction relationship and control logic between the components are determined, and the right and responsibility distribution of each party in different operation modes is determined, mainly including the following elements: controller, control behavior, actuator, sensor, feedback, and controlled process.

[0071] Step three: identifying unsafe control behavior.

[0072] In remote control mode, the following four situations occur, indicating that the control behavior may have safety hazards: (1) Failure to provide control behavior leads to danger; (2) Provided control behavior leads to danger; (3) Provided control behavior that may be safe but the node provided is too early, too late, or in the wrong order; (4) Control behavior lasts too long or stops too early.

[0073] Step 4: Develop causal scenarios.

[0074] Possible causal scenarios in remote control mode include: (1) unsafe controller behavior; (2) unsafe control paths; (3) unsafe controlled processes; and (4) inappropriate feedback and information.

[0075] S2. Select the causal scenarios of unsafe control behaviors in the operation of unmanned conditionally autonomous ships in remote control mode as input nodes, unsafe control behaviors as intermediate nodes, and the operational risk level corresponding to the remote control mode adopted by the ship as output nodes. Construct a Bayesian network model, perform parameter learning and probability calculation, and dynamically evaluate operational risks.

[0076] In a specific embodiment, the cause scenario of the unsafe control behavior of the ship in the remote control mode is the input node. Assume that the input node is The ship operation risk level is set as the output node N S The node status, i.e., the operational risk level, is divided into three states: low S1, medium S2, and high S3.

[0077] According to the causal scenarios of unsafe control behaviors in ship operation, the topological structure of the Bayesian network model is constructed. To output node N S There is one directed edge.

[0078] Combining expert opinions and the occurrence frequency of causal scenarios extracted from historical ship data, the Bayesian network model is trained to determine the conditional probability table of each node. Under different conditions (condition X and condition Y), the probability of being at different risk levels is expressed as wait.

[0079] The total probability formula is used to calculate the prior probability of the output node. Assume that the factors affecting the risk level of ship operation are node A, with the values ​​of normal a1 and abnormal a2, and node B, with the values ​​of good b1 and bad b2). According to the total probability formula:

[0080]

[0081] Calculate the risk level of the ship at different risk levels S i The prior probability of .

[0082] S3. Based on the Bayesian network model, adjust the indicator parameters as needed, calculate the posterior probability of the adjusted target node, and evaluate the operational risk level.

[0083] In the embodiment, according to the actual needs of ship operation in the remote control mode, the specific input node index parameters in the Bayesian network model are adjusted in a targeted manner, and the input node Danger level from Reduce to This optimized situation is used as evidence E.

[0084] Calculating Joint Probability: Assumptions The parent node is According to the joint probability formula, we have:

[0085]

[0086] Calculate the posterior probability: P(E,N S =S i ) is normalized and the posterior probability is obtained as:

[0087]

[0088] By calculation, we can find that the posterior probability P(N S =S1|E)(the ship operation risk level is low) is improved, and the posterior probability P(N S =S3|E)(high ship operation risk level) reduced.

[0089] Compare the posterior probability P(N S =S1|E), P(N S =S2|E), P(N S =S3|E) size.

[0090] Due to the input node parameter adjustment, P(N S =S1|E) becomes the largest, it is judged that the current ship operation risk level has changed from the previous possible high or medium state to a low state.

[0091] S4. Uncertainty assessment based on information entropy is introduced to provide the confidence interval of the posterior probability of ship operation risk under remote control mode and evaluate the reliability of the risk level.

[0092] According to the information entropy formula:

[0093]

[0094] Assessing the uncertainty of the risk level, in the embodiment, it is assumed that the information entropy H(N S ) is calculated, compared with before optimization, the posterior probability distribution is more concentrated in the lower state, the information entropy H(N S ) is reduced, indicating that the uncertainty is reduced, and the evaluation result is more reliable.

[0095] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art in the technical range disclosed by the present application, according to the technical scheme and the inventive concept of the present application, equivalent replacement or change, should be covered in the protection scope of the present application.

Claims

1. A method for assessing the operational risk of conditionally autonomous ships with limited crew under different operation modes, characterized by include: Analyze the operational system control structure of conditionally autonomous ships with reduced crew capacity under different operating modes, and use system theory and process analysis methods to identify unsafe control behaviors and their causal scenarios; The system uses the causal scenarios of unsafe control behaviors of conditionally autonomous ships with reduced crew in different operating modes as input nodes, the unsafe control behaviors as intermediate nodes, and the operational risk levels corresponding to the different operating modes adopted by the ship as output nodes to construct a Bayesian network model, perform parameter learning and probability calculation on the Bayesian network model, and dynamically evaluate the operational risk of the Bayesian network model. Adjust the indicator parameters of the Bayesian network model according to needs, calculate the posterior probability of the adjusted target node, and evaluate the operational risk level of the conditionally autonomous ship with reduced crew under different operating modes; The information entropy is used to evaluate the uncertainty of the operational risk level results, the confidence interval of the posterior probability of the operational risk level is analyzed, and the reliability of the operational risk level determination is evaluated.

2. The method for assessing operational risks of a conditionally autonomous ship with reduced crew capacity under different operating modes according to claim 1 is characterized by: Identify the losses that may occur during the operation of conditionally autonomous ships with reduced crews, identify the system-level hazards that lead to the losses, determine the safety constraints that the system needs to meet to avoid the hazards, and refine and classify the identified hazards; Delineate the various components of the operating system for conditionally autonomous ships with reduced crews, including the ship automation system, crew operations, and shore-based control centers, clarify the interaction and control logic between the various components, and clarify the allocation of rights and responsibilities of each party under different operating modes; Comprehensively identify unsafe control behaviors during ship operations, including incorrect, missing, delayed or inappropriate control instructions, analyze the causes of unsafe control behaviors, and develop causal scenarios.

3. The method for assessing operational risks of a conditionally autonomous ship with reduced crew capacity under different operating modes according to claim 1 is characterized by: When building a Bayesian network model: First, the Bayesian network node variables and hierarchical relationships are determined. The input node is defined and converted into the root node of the Bayesian network model based on the identified causal scenario. When constructing the intermediate node, the unsafe control behavior analyzed through the system theory process analysis method is used as the intermediate node and a causal relationship is established with the input node. Setting output nodes: Divide autonomous ships into different operating modes, define output nodes of operational risk levels, and quantify risk levels; Carry out causal relationship mapping, analyze and derive the causal chain based on the system theory process analysis method, establish a directed acyclic graph between nodes, and construct a Bayesian network topology structure; The frequency of occurrence of causal scenarios is extracted from historical ship data, and the prior probability of the root node is calculated. The nodes lacking historical data are analyzed by aggregating expert opinions through the fuzzy Delphi method, and the qualitative description is converted into probability intervals. The posterior probability inference is performed based on the Bayesian theorem to assess the risk level in real time.

4. The method for assessing operational risks of a conditionally autonomous ship with reduced crew capacity under different operating modes according to claim 1 is characterized by: When the Bayesian network model is used to evaluate the operational risk level of autonomous ships: Determine the nodes and structure of the Bayesian network model: Assume that there are n nodes in the Bayesian network model, represented by X1, X2, ..., X n , where X n is the output node representing the ship operation risk level, and the other nodes are input nodes. The dependency relationship between the nodes is represented by a directed acyclic graph. Each node X i Has its parent node set Pa(X i ); Determine the conditional probability table (CPT) of each node: For each node X i , its conditional probability table P(X i |Pa(X i ) indicates that at its parent node Pa(X i ) takes different values, node X i The probability of taking different values; Use the total probability formula to calculate the prior probability of the output node; Calculate the output node X using the total probability formula n The prior probability of being in different states, assuming that the input nodes X1, X2, ..., X n-1 The different values ​​or states of constitute a partition of the sample space E1, E2, ..., E n , for the output node X n A certain level A is expressed as: Among them, P(E i ) is event E i The probability of occurrence, P(A|E i ) is in E i The probability of event A occurring under the conditions of occurrence, that is, the probability of ship operation risk being at level A given certain influencing factors.

5. The method for assessing operational risks of a conditionally autonomous ship with reduced crew capacity under different operating modes according to claim 1 is characterized by: According to the actual needs of ship operation, the input node index parameters in the Bayesian network model are adjusted in a targeted manner, and then the model is re-run. The Bayesian inference algorithm and probability calculation method are used to calculate the probability of ship operation risks at different levels after adjustment, and the posterior probability P(X) of the output node is calculated when the data source E appears. n |E): Calculate the joint probability: For n nodes X1, X2, ..., X n The joint probability distribution of the Bayesian network is expressed as the product of the conditional probabilities of each node under the condition of its parent node, which is expressed as: Among them, Pa(X i ) represents node X i The parent node collection Assume that the real-time data of the input node is but Calculate the posterior probability: When a new data source E appears, the formula is obtained according to Bayes' theorem P(X n ) is event X n The prior probability of occurrence, that is, before the data E is observed, X n The probability of P(E|X n ) is the likelihood, indicating that under the assumption X n The probability of observing data E when the probability of data E is established; P(E) is the probability of evidence E; Since P(E) is n The same value is obtained by n =x nj ) is normalized to obtain the posterior probability P(X n =x nj |E), i.e.

6. The method for assessing operational risks of a conditionally autonomous ship with reduced crew capacity under different operating modes according to claim 1 is characterized by: When assessing the reliability of risk level determinations: Determine the random variable and its value range: take the ship operation risk level as the random variable X, and determine the value space of the random variable X; Calculate the probability of the ship operation risk level being in different states: Based on the Bayesian network model, the probability value is obtained through Bayesian reasoning and probability calculation. Assume that X has k values, namely x1, x2,…, x k , the corresponding probabilities are P(x1), P(x2),…, P(x k ), where the probability value reflects the likelihood of the ship operation mode being at each risk level under the current data and model structure; Calculate using the information entropy formula: Substitute the probability values ​​of each value into the information entropy formula of the ship operation risk level X Among them, P(x i ) is X and its value is x i If the H(X) value is large, it means that the uncertainty of the ship operation risk level is high; if the H(X) value is small, it means that the certainty of the risk level is relatively high and the assessment result is relatively reliable.

Images