An improved knotted-continuous bayesian network model based cabin upset event emergency decision method

Abstract

The application discloses a kind of cabin jolt event emergency decision-making methods based on improved knot-continuous bayesian network model, comprising the following steps: according to the preset jolt event emergency handling process, confirm disposal node and node relationship;According to the disposal node and the node relationship, construct knot model, and record the node state under multiple time slices before and after jolt event through the knot model, obtain dynamic knot model;The dynamic knot model is converted into continuous bayesian network model, and the transition probability of any node under different time slices is calculated by preset transition mode;The application can realize the simulation of the state of each node before and after jolt event changes with time by improving knot model to form continuous bayesian network model, and can realize the dynamic analysis of each node.

Description

Technical Field

[0001] This invention relates to the field of emergency decision-making technology, and more specifically to an emergency decision-making method for cabin turbulence events based on an improved bowtie-continuous Bayesian network model. Background Technology

[0002] Turbulence refers to the weather disasters that cause aircraft to be impacted by disturbed airflow during flight, resulting in inaccurate instruments and operational difficulties.

[0003] Current research both domestically and internationally focuses on pre-flight turbulence warning technologies and post-flight emergency response methods. On one hand, turbulence warning technology focuses on pre-flight analysis of cabin turbulence causes from a meteorological perspective, leading to pre-flight turbulence prediction and early warning. However, while existing turbulence warning technologies can identify most convective turbulence, they struggle to accurately warn of clear-air turbulence. Existing technologies treat cabin emergencies as static events, focusing on post-incident case analysis and response strategies, neglecting the impact of event severity and ongoing response processes. On the other hand, post-flight emergency response methods focus on personnel evacuation, emergency decision-making, and causal analysis after cabin turbulence, fires, and other emergencies during flight. Existing technologies lack specific research on turbulence events themselves, particularly regarding the inability to accurately warn of sudden turbulence, the inability to dynamically monitor the situation after turbulence occurs, and the lack of critical emergency strategies that consider dynamic changes. Summary of the Invention

[0004] In view of this, the present invention provides an emergency decision-making method for cabin turbulence events based on an improved bowtie-continuous Bayesian network model. It takes into account the characteristics of turbulence-related emergencies, such as lack of warning, degree of impact, and time variation. By constructing a dynamic evaluation model, it scientifically analyzes the key nodes of the event and emergency strategies to reduce the consequences of the event, such as casualties. It can provide a methodological reference for similar emergencies in confined spaces, where rescue is difficult and dynamic handling is required.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model includes the following steps:

[0007] Confirm the handling nodes and node relationships.

[0008] A bowtie model is constructed based on the handling nodes and the node relationships, and the node states at different times during the turbulence event and the degree of turbulence of the turbulence event are recorded through the bowtie model.

[0009] The bow tie model corresponding to different times is transformed into a continuous Bayesian network model under multiple time slices.

[0010] Initialize the initial probability of each node in the continuous Bayesian network model according to the degree of turbulence; calculate the transition probability of adjacent time slices according to the initial probability, and obtain the node probability distribution of each time slice by combining the observations; identify key nodes according to the node probability distribution.

[0011] Preferably, the method of combining observations to obtain the node probability distribution for each time slice includes:

[0012] Calculate the joint transition probability for a given time slice based on the node transition probabilities.

[0013] Based on the joint transition probability, the observation values ​​from the start to the corresponding time slice are introduced, and the node probability corresponding to each node is calculated to obtain the node probability distribution under the corresponding time slice.

[0014] Preferably, the initial probability is a prior probability or a posterior probability under the observed value.

[0015] When the initial probability is the prior probability, the dynamic changes of each node before and after the occurrence of the turbulence event are predicted by the node probability distribution.

[0016] When the initial probability is the posterior probability under the observed value, the cause of the event is analyzed through the node probability.

[0017] Preferably, the step of obtaining the prior probability includes:

[0018] Obtain statistical data on the occurrence of events at each node and calculate the first prior probability;

[0019] Obtain the configuration data for the occurrence of events at each node and calculate the second prior probability;

[0020] The final prior probability is determined by combining the first prior probability and the second prior probability.

[0021] Preferably, the transition probability is calculated as follows:

[0022]

[0023] Among them, X t Let X be the set of nodes in the transition network corresponding to time slice t. t+Δt Let be the set of nodes on time slice t+Δt; For the i-th node in the t-th time slice, For nodes The parent node, N is the total number of nodes in each time slice.

[0024] Preferably, the joint transition probability is calculated as follows:

[0025]

[0026] in, Let X be the node state of node n in the transition network corresponding to time slice t, and let X0 represent the set of nodes in the initial network.

[0027] Preferably, the step of transforming the bow tie model corresponding to different times into a continuous Bayesian network model under multiple time slices specifically includes:

[0028]

[0029] Among them, A d Let A represent the set of all nodes in the improved bowtie model, where n is the set of parent nodes and top events, n∈(1,2…n); m For intermediate events, m represents the intermediate event number, and a... z For the base event, z represents the base time number; A is the top event, and DBN consists of the initial network X0 and the dynamic network X. t Composition, where X0 represents the initial set of nodes at t=0; X t X represents the set of network nodes that are dynamically generated during time slice t; z (0) Represents the root node, X m (0) Represents an intermediate node; X (0) Represents a leaf node; Δt∈(1,2…t) represents the time variation.

[0030] Preferably, the bow tie model is constructed based on the handling node and the node relationship, including: taking the occurrence of turbulence as the top event A, and confirming the various levels of events of the factors causing the event and the various levels of events of the decision-making process for the event based on the top event A.

[0031] Preferably, the events at each level of the event occurrence decision include multi-level representative events of turbulence, set sequentially according to the order of the degree of turbulence after it occurs.

[0032] An emergency decision-making system for cabin turbulence events based on an improved bowtie-continuous Bayesian network model includes a computer system for implementing the aforementioned emergency decision-making method.

[0033] As can be seen from the above technical solution, compared with the prior art, this invention discloses an emergency decision-making method for cabin turbulence events based on an improved bowtie-continuous Bayesian network model. By analyzing the emergency response process for cabin turbulence events, an improved bowtie model reflecting structured node relationships is constructed. Mapping conditions and transformation rules are established to form a continuous Bayesian network model. In the model construction, the characteristics of turbulence-type emergencies, such as lack of warning, degree of impact, and time variation, are considered, forming a dynamic model. This helps to scientifically analyze the key nodes and emergency strategies of the event, so as to reduce the consequences of the event, such as casualties. It can provide a methodological reference for similar emergencies in confined spaces, where rescue is difficult, and dynamic handling. At the same time, through the analysis of key nodes, non-critical nodes in the emergency response process can be reflected indirectly, which is also helpful for optimizing the emergency response process. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0035] Figure 1 A schematic diagram of an emergency decision-making method for cabin turbulence events based on an improved bowtie-continuous Bayesian network model provided by the present invention.

[0036] Figure 2 This is a schematic diagram of the improved bow tie model structure in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram illustrating the transformation method from the improved bowtie model to the continuous Bayesian network model in an embodiment of the present invention.

[0038] Figure 4 This is a schematic diagram of the structure of the transformed Bayesian network model in an embodiment of the present invention. Detailed Implementation

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

[0040] Example 1

[0041] This invention discloses an emergency decision-making method for cabin turbulence events based on an improved bowtie-continuous Bayesian network model, comprising the following steps:

[0042] S1: Confirm the processing node and node relationships.

[0043] S2: Construct a bow tie model based on the handling nodes and node relationships, and record the node status at different times during the turbulence event and the degree of turbulence of the turbulence event through the bow tie model.

[0044] S3: Based on the bow tie model corresponding to different times, transform it into a continuous Bayesian network model under multiple time slices;

[0045] S4: Initialize the initial probability of each node in the continuous Bayesian network model according to the degree of turbulence; take the node state recorded by the bow tie model at different times as the observation value; calculate the node probability distribution under each time slice according to the initial probability and the observation value, and identify the key nodes according to the node probability distribution.

[0046] In this embodiment, the present invention introduces time slices into the standard bow tie model, increasing its dynamism. This allows for the analysis of the changes in the node states of each node over time before and after a turbulence event using a continuous Bayesian network model. The continuous Bayesian network model includes an initial node network and a transition node network. When dynamically analyzing a turbulence event using this model, the transition network analyzes the transition probabilities of each node and updates the network based on different node observations, thereby clarifying the dynamic impact of each node on the turbulence event.

[0047] For example, when the transition probabilities of nodes A111 (captain's misjudgment of turbulence level), A121 (crew fatigue), and A123 (crew's failure to take proper personal protective measures) are known, it can be predicted that in the following period, if the probability of the captain's misjudgment increases, the probability of the crew's emergency response failure will also increase. If the probability of the crew being fatigued is high, the probability of failing to take proper personal protective measures will also increase accordingly.

[0048] Furthermore, based on the continuous Bayesian network model, forward and backward reasoning functions can be achieved by combining prior or posterior probabilities. Specifically, this includes constructing a continuous Bayesian network for cabin turbulence emergency response using GeNle software, with prior probability P... A Data analysis is conducted by first performing forward reasoning to predict the dynamic changes in cabin turbulence emergency response time, helping decision-makers to take early intervention measures; secondly, backward reasoning is performed to obtain the posterior probability of each key node when the risk occurs at time t, analyze the key causes of turbulence events, and obtain a continuous Bayesian network for cabin turbulence emergency response.

[0049] Forward reasoning starts from known evidence and propagates forward along the structure of the Bayesian network to gradually infer the probability distribution of other nodes. This reasoning method can be used to predict the state of unknown variables or to estimate the state of the remaining variables when some variables are known.

[0050] In the forward reasoning process, the nodes in the continuous Bayesian network model need to be initialized first based on the prior probabilities of each node; where the prior probability refers to the initial probability distribution of the nodes under a certain degree of turbulence.

[0051] After obtaining the initial probability distribution, the transition probability is calculated based on the initial probability distribution. Then, by introducing the observed values, the probability distribution of the current time slice is updated. This allows us to obtain the dynamic changes of each node before and after the occurrence of the turbulence.

[0052] Backward reasoning is the process of starting from the target node and propagating backward along the Bayesian network structure to gradually obtain the necessary information. This reasoning method is typically used to find the possible causes of a given result.

[0053] In the process of reverse reasoning, it is necessary to provide the posterior probability for a given observation, and then combine the observation with the transition probability to infer the probability distribution of nodes before the observation, thereby determining the cause of the observed state. Therefore, the key time slices and handling nodes can be obtained based on the occurrence of the turbulence event.

[0054] Next, the various steps in this invention will be described in detail.

[0055] For S1, to clarify the handling nodes and their relationships, it is necessary to define the emergency response process for cabin turbulence events. In one embodiment, turbulence events can be divided into three stages based on the handling time: pre-flight, in-flight, and post-flight. Key personnel involved in turbulence handling include passengers, cabin crew, flight crew, and airline AOC personnel. Furthermore, to clarify the entire process of incident causation in in-flight cabin emergency handling, the handling nodes and their relationships are identified from four aspects: weather conditions, mechanical equipment, safety supervision, and human factors.

[0056] like Figure 2For S2, when constructing the improved bowtie model, the overall risk factors are first identified from four aspects: human, machine, environment, and management. Cabin turbulence leading to personnel injury is taken as the top event. The intermediate and basic events leading to the top event are analyzed from top to bottom, and 23 key handling nodes are selected to construct an accident tree. The top event is the occurrence of turbulence event A. The first layer of intermediate events are: A1 Human factors; A2 Management factors; A3 Environmental factors; A4 Onboard equipment factors. The second layer includes: A11 Captain's emergency response failure; A12 Cabin crew's emergency response failure; A13 Passenger emergency response failure; A21 Internal causes; A22 External causes; A31 Severe weather conditions; A32 Inability for the aircraft to bypass; A41 Fixture failure; A42 Weather radar failure. The third layer includes: A5 Pre-flight phase; A6 In-flight phase. The fourth level includes: A111 misjudgment of turbulence level; A112 captain's failure to properly secure aircraft; A113 failure to broadcast turbulence; A114 failure to hold a tripartite meeting; A115 captain's fatigue; A121 flight attendant's fatigue; A122 failure to check and secure aircraft items; A123 failure to provide personal protection; A124 ineffective handling of service issues; A125 not wearing seatbelts; A132 not in seat; A133 not wearing seatbelts; A211 inadequate crew training; A212 excessive service by flight attendants; A213 crew psychological problems; A221 deficiencies in regulations; A222 insufficient supervision. The fifth level includes: A1241 failure to secure the catering cart; A1242 failure to secure oneself; A1321 failure to hold onto the handrail; A1322 failure to secure oneself.

[0057] Furthermore, taking turbulence A as the initial event, and based on the chronological order of turbulence events, the impact of turbulence severity, failure barriers in handling procedures, and the severity of event consequences, a seven-layer event tree was constructed for handling decisions. According to accident statistics and relevant laws and regulations, the turbulence severity was further divided into seven levels according to its evolution after occurrence: whether the turbulence stopped, whether there were any casualties, the severity of the casualties, whether to make an emergency landing, whether the landing was successful, whether rescue efforts were coordinated, and whether the incident was handled collaboratively. Finally, an improved bowtie model for cabin turbulence emergency response was obtained, namely the dynamic bowtie model.

[0058] For S3, the first step is to improve the transformation from the bowtie model to a continuous Bayesian model, such as... Figure 3 The mapping transformation from the improved bowtie model to the continuous Bayesian network model needs to satisfy the following conditions. ： 1. Nodes are in one-to-one correspondence, that is, a node in the bow tie model corresponds to a node in the Bayesian network model; the hierarchical relationship of each node in the bow tie model corresponds to the root-leaf node relationship in the Bayesian network model; 2. The transformation process has a certain time order. According to the preset time slice interval Δt, a transformation is performed every Δt time slices.

[0059] Specifically, the mapping expression is:

[0060]

[0061] In equation (1), A d Let represent the set of all nodes in the improved bowtie model, where n is the set of parent nodes and top events, n∈(1,2…n). A m For intermediate events, m represents the intermediate event number, and A... z Let z represent the base event, and A represent the top event. DBN consists of the initial network X0 and the dynamic network X... t Composition, where X0 represents the initial set of nodes at t=0; X t X represents the set of network nodes that are dynamically generated during time slice t; z (0) Represents the root node, X m (0) Represents an intermediate node; X (0) Represents a leaf node. Δt∈(1,2…t) represents the time variation.

[0062] The relationships between nodes in the transformed Bayesian network model are as follows: Figure 4 As shown in the diagram, the arrows point from causal factors to result factors, i.e., from parent nodes to child nodes. Nodes without parent nodes are root nodes, and each root node is represented by a directed arc "1". The time slice interval is 1. In DBN, the node status is described using appropriate qualitative language based on the node name, such as "yes" or "no".

[0063] Next, the prior probabilities are substituted into the constructed node graph for further calculation.

[0064] Specifically, the transition rules between different time slices in the continuous Bayesian network model are as follows: Since the transition network of the continuous Bayesian network for cabin turbulence is a Bayesian network model spanning multiple time slices, it satisfies the Markov assumption of stationarity. Emergency response decisions for cabin turbulence are only affected by the previous time slice; that is, the probability of a node at time t is only affected by time t-1 and is independent of the time slices before t-1. Therefore, the conditional probabilities of nodes in the network model transformed from the improved bowtie model are exactly the same as in the initial network, and the transition probabilities remain unchanged throughout the entire continuous Bayesian network model. The transition probabilities between any two adjacent time slices in the continuous Bayesian network are expressed as follows:

[0065]

[0066] In equation (2), X is in the Bayesian model after the improved bowtie model mapping. t+Δt Let X be the set of nodes on time slice t+Δt. t Let t be the set of nodes on time slice t; Let be the value of the i-th node in the (t+Δt)-th time slice. For nodes The parent node of a node can be located in the same or adjacent time slices. N is the total number of nodes in each time slice, N∈(1,2…n); according to the Markov chain principle, nodes cannot be connected across time slices.

[0067] To further implement the above technical solution, based on the prior probability P of the initial network X0... A The transition probability P(X) of the transition network t |X t-1 The joint transition probability of any node in the continuous Bayesian network is calculated as follows:

[0068]

[0069] In equation (3), This refers to the situation of node n in time slice t. This is the transition probability of the initial network. During calculation, the parent nodes of the initial network use the prior probability P of the initial network. A ; It represents the transition probability of the transition network.

[0070] In this embodiment, the node probability distribution under any time slice can be calculated by combining the joint transition probabilities with the corresponding observation values.

[0071] To further implement the above technical solution, the steps for obtaining prior probabilities include: obtaining statistical data on the occurrence of events at each node and calculating a first prior probability; obtaining configuration data on the occurrence of events at each node and calculating a second prior probability. The first prior probability and the second prior probability are respectively the direct prior probability and the supplementary prior probability.

[0072] Specifically, the prior probability P required for continuous Bayesian computation A It consists of direct prior probabilities that can be directly statistically obtained and supplementary prior probabilities. The supplementary prior probabilities are obtained by the triangular fuzzy number pair expert evaluation method.

[0073] Among them, the direct prior probability is the statistical probability P(C n )as follows:

[0074]

[0075] In equation (4), C a P(C) represents the number of times the nth node occurs under statistical circumstances. a ) represents the probability of the nth node occurring, C all This refers to the number of times each node occurs, i = 1, 2, 3...n (n is the total number of events).

[0076] Supplementing the prior probability of the node, i.e., the fuzzy probability P G The calculations are as follows: First, eight senior experts in airline cabin management were invited to assign weights to their experience levels based on industry relevance factors such as years of experience, flight time, and industry qualifications. Second, a questionnaire was created based on the turbulence handling process for moderate and severe turbulence. For each issue, five linguistic variables were defined based on the probability of occurrence, ranging from never occurring to always occurring, for the experts to score. Finally, the fuzzy total G for each node was obtained through the questionnaire. i :

[0077]

[0078] In the formula, G i Let W be the total number of fuzzy numbers, i = 1, 2, 3…n. i The weight G of the i-th expert ij Let B be the score for event i in the j-th pair, and B be the total number of experts. Substituting the obtained total fuzzy count into equation (6), we get the triangular fuzzy number G for the relevant nodes.

[0079]

[0080] In equation (6), G = (k, l, u) is the triangular fuzzy number, and its corresponding membership function μ G (A) represents the upper and lower limits of the triangular fuzzy number, respectively, and l is the value when the membership degree of the fuzzy number is 1.

[0081] Based on the triangular fuzzy number obtained through statistical calculation, the final fuzzy probability is:

[0082]

[0083] To facilitate subsequent simulations, different weights can be assigned to statistical probabilities and fuzzy probabilities, such as statistical probability P(C n The weight is assigned to 0.7, and the prior probability P obtained from the triangular fuzzy number is... G Assign a weight of 0.3. Then, obtain all the final prior probabilities.

[0084] Understandably, for prior probabilities, in addition to using the weighted sum of statistical probability and fuzzy probability as the final prior probability, one can also choose either statistical probability or fuzzy probability alone as the final prior probability, depending on the actual situation.

[0085] In this embodiment, the present invention achieves full-process monitoring of cabin turbulence events by constructing an improved bowtie-continuous Bayesian network model, grasps the changes and trends of each node over time, and analyzes turbulence events by calculating node probabilities, performing data simulation and modeling.

[0086] For example, the prior probabilities under moderate and severe turbulence are used for initialization, respectively, and the initialization parameters are shown in the table below:

[0087]

[0088] Based on the calculated prior probabilities of different nodes under moderate and severe turbulence, the node states are set according to the actual situation of the nodes. Then, dynamic indicators are selected to consider the time-varying characteristics of the nodes, making them dynamic nodes. The number of time slices is then appropriately set using GeNle software. The calculation logic of GeNle software is the transition probability and joint probability mentioned above.

[0089] First, a dynamic Bayesian network for cabin turbulence emergency response was constructed using GeNle software, and the graph was input with prior probabilities for data analysis. Through data simulation, the probability distribution of each node was obtained. In the experiment, the variation every 2 minutes of time slices was significant; therefore, node results were analyzed every two time slices starting from t=1. When t≥8, the degree of variation almost overlapped with that at t=7, and was no longer significant. Thus, the probability distribution results of each node under moderate and severe turbulence at time slices t=1, t=3, t=5, and t=7 were obtained.

[0090] (1) Emergency response is time-sensitive, with an optimal response time. Under moderate turbulence, the probability of node failure increases with response time. Similarly, the change in the failure probability of response nodes under severe turbulence shows the same trend as under moderate turbulence. For example, at t=1, the failure probability of emergency response nodes under severe turbulence is higher than that under moderate turbulence, but as time increases, at t=7, the failure probabilities of the two are almost equal. The failure probabilities of response under moderate and severe turbulence are almost the same at t=5 and t=7, tending to stabilize, indicating that the optimal emergency response time is before t=5. (2) Emergency response strategies are affected by the degree of turbulence and are different. At t=1, comparing nodes under moderate and severe turbulence, the probabilities of flight attendants not taking proper protective measures (A123), passengers and flight attendants not fastening their seat belts (A125 and A133), not having a fixed dining car (A1241), and adverse weather conditions (A31) change significantly, with changes exceeding 0.2. This indicates that inadequate personnel restraint measures were the main reason for the failure of emergency response after the degree of turbulence increased.

[0091] (3) Human factors accounted for the highest probability of emergency response failures at each stage. Among all contributing factors, A123, A1241, A1242, A133, and A212 were the key points leading to the failure of cabin turbulence response, with high failure probabilities at different time points and under different degrees of turbulence. This indicates that passengers not fastening their seatbelts and excessive service from cabin crew were the key reasons for the failures.

[0092] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0093] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model, characterized in that, Includes the following steps: The handling points and relationships of each point are identified from four aspects: weather conditions, machinery and equipment, safety supervision, and human factors. An improved bowtie model is constructed based on the handling nodes and the node relationships, and the node states at different times during the turbulence event and the degree of turbulence of the turbulence event are recorded through the improved bowtie model. An improved bowtie model is constructed based on the handling nodes and the node relationships, including: taking the occurrence of turbulence as the top event A, and confirming the various levels of events of the factors that cause the event and the various levels of events of the event occurrence decision based on the top event A; the various levels of events of the event occurrence decision include multi-level turbulence representative events set in sequence according to the time order of the occurrence of turbulence events and the order of the degree of evolution after the occurrence of turbulence. The improved bow tie model corresponding to different times is transformed into a continuous Bayesian network model under multiple time slices; during the transformation process, the nodes in the improved bow tie model correspond one-to-one with the nodes in the Bayesian network model; the hierarchical relationship of each node in the improved bow tie model corresponds to the root and leaf node relationship in the Bayesian network model; and the model is periodically transformed according to the preset time slice interval. Initialize the initial probability of each node in the continuous Bayesian network model according to the degree of turbulence; The transition probabilities of adjacent time slices are calculated based on the initial probabilities, and the node probability distribution of each time slice is obtained by combining the observations. Key nodes are identified based on the node probability distribution.

2. The cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model according to claim 1, characterized in that, The combined observations yield the node probability distribution for each time slice, including: Calculate the joint transition probability for a given time slice based on the node transition probability; Based on the joint transition probability, the observation values ​​corresponding to the start of the time slice are introduced, and the node probability corresponding to each node is calculated to obtain the node probability distribution under the corresponding time slice.

3. A cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model according to claim 1 or 2, characterized in that, The initial probability is either the prior probability or the posterior probability under the observed value; When the initial probability is the prior probability, the dynamic changes of each node before and after the occurrence of the turbulence event are predicted by the node probability distribution. When the initial probability is the posterior probability under the observed value, the cause of the event is analyzed by node probability analysis.

4. The cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model according to claim 3, characterized in that, The steps for obtaining the prior probability include: Obtain statistical data on the occurrence of events at each node and calculate the first prior probability; Obtain the configuration data for the occurrence of events at each node and calculate the second prior probability; The final prior probability is determined by combining the first prior probability and the second prior probability.

5. The cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model according to claim 2, characterized in that, The transition probability is calculated as follows: ； in, For time slice intervals, for A time slice corresponds to the set of nodes in the transition network. for The set of nodes on a time slice; No. The first time slice 1 node For nodes The parent node, The total number of nodes in each time slice.

6. The cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model according to claim 5, characterized in that, The joint transition probability is calculated as follows: ； in, No. The first time slice 1 node This represents the initial set of nodes in the network.

7. The cabin turbulence event emergency decision-making method based on an improved bowtie-continuous Bayesian network model according to claim 1, characterized in that, The process of transforming the improved bowtie model at different times into a continuous Bayesian network model across multiple time slices specifically includes: ； in, Represents the set of all nodes in the improved bowtie model; This is an intermediate event. The number representing the intermediate event. As the bottom event, The number representing the bottom event; For the top event, the dynamic Bayesian network is formed by the initial network. and transfer network composition, Representative at The network set consisting of the initial nodes at that time; Representative at A time slice is a network set composed of dynamic nodes; Represents the root node. Represents an intermediate node; Represents a leaf node; Indicates the change over time.

8. A cabin turbulence event emergency decision-making system based on an improved bowtie-continuous Bayesian network model, characterized in that, Includes a computer system for implementing the emergency decision-making method according to any one of claims 1-7.

Images