Passenger evacuation time prediction method and device under ship inclination condition

Abstract

The invention provides a passenger evacuation time prediction method and device under a ship inclination condition, and belongs to the field of artificial intelligence, and the method comprises the steps: taking a training set as the input of a vector weighted average algorithm, and employing a reverse learning mode to generate a vector set during iteration; the training set comprises personnel categories and evacuation speed and evacuation time of each category of personnel; determining a candidate value of the next iteration based on a product of the candidate value of the current iteration and a preset coefficient, and determining a Mean Rule value based on the candidate value of the next iteration; and updating a vector position based on a Mean Rule value, optimizing the weight of the DELM model based on the vector of the updated position, training the model based on a training set to obtain a passenger evacuation time prediction model, and obtaining the total time required for evacuation based on the model. According to the method, reverse learning and dynamic candidate value generation are adopted to solve the problem that an existing passenger evacuation time prediction method under the ship inclination condition is low in prediction precision, speed and reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and particularly relates to a method and device for predicting the evacuation time of passengers in the case of ship inclination. Background Art

[0002] When a ship is damaged and inclined, it is necessary to predict the overall evacuation time of the passengers inside the ship in a timely manner for rescue. There are many passengers on the ship. When the passenger information data is lacking and the parameters are mismatched, the accuracy of the prediction of the personnel evacuation time is greatly reduced, and it will also lead to a low speed and reliability of the prediction of the evacuation time. Summary of the Invention

[0003] In view of this, it is necessary to provide a method and device for predicting the evacuation time of passengers in the case of ship inclination to solve the technical problems of low prediction accuracy, speed and reliability existing in the existing method for predicting the evacuation time of passengers in the case of ship inclination.

[0004] To solve the above problems, in a first aspect, the present invention provides a method for predicting the evacuation time of passengers in the case of ship inclination, including: Taking the training set as the input of the vector weighted average algorithm, performing iterative solution, and generating a vector set in a reverse learning manner during iteration; the training set includes the personnel categories on the ship, as well as the evacuation speed and evacuation time of each category of personnel; Determining the candidate value for the next iteration based on the product of the candidate value for the current iteration and a preset coefficient, and determining the MeanRule value based on the candidate value for the next iteration; Updating the vector positions in the vector set based on the MeanRule value, determining the weights of the DELM model based on the vectors at the updated positions, and training the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model; Inputting the test set into the passenger evacuation time prediction model to obtain the total time required for the evacuation of the personnel on the ship.

[0005] In a possible implementation manner, taking the training set as the input of the vector weighted average algorithm and performing iterative solution includes: Taking the training set, weight boundary, maximum number of iterations, weight dimension and fitness function as the input of the vector weighted average algorithm, and performing iterative solution; Wherein, the fitness function is the sum of the mean square errors of the evacuation times of each category of personnel in the training set and the mean square errors of the evacuation times of each category of personnel in the test set, and the test set is used to test the trained DELM model.

[0006] In a possible implementation manner, generating a vector set in a reverse learning manner during iteration includes: During iteration, a reverse vector is generated based on the current vector in the vector set of the current iteration, and a vector closer to the global optimal solution is determined from the current vector and the reverse vector. The current vector in the vector set of the current iteration is replaced with the vector closer to the global optimal solution to obtain the vector set for the next iteration; wherein, the initial vector set is the training set.

[0007] In a possible implementation manner, determining a better vector from the current vector and the reverse vector includes: Evaluating the fitness function values corresponding to the current vector and the reverse vector, and determining a better vector from the current vector and the reverse vector based on the fitness function values.

[0008] In a possible implementation manner, generating a reverse vector based on the current vector in the vector set of the current iteration includes: Subtracting the current vector from the sum of the maximum value and the minimum value of the current vector in the vector set of the current iteration to obtain the reverse vector.

[0009] In a possible implementation manner, weighting the vector DELM model based on the vector at the updated position includes: Merging the vectors at the updated position to generate a new vector; Determining a local search operator, and determining the optimal fitness of the new vector based on the local search operator; Updating the global optimal fitness based on the optimal fitness of the new vector in two adjacent iterations; When it is determined that the latest global optimal fitness meets the preset iteration condition, weighting the DELM model based on the last generated new vector.

[0010] In a possible implementation manner, the calculation formula of the candidate value is as follows:

[0011]

[0012] where, is the candidate value for the i th iteration, Maxg represents the maximum number of iterations, i is the current iteration number, is for the i +1th iteration candidate value.

[0013] In a possible implementation manner, the calculation formula of the MeanRule value is as follows:

[0014]

[0015]

[0016]

[0017]

[0018]

[0019]

[0020]

[0021]

[0022]

[0023]

[0024] Among them, is the objective function; and are MeanRule functions in; , and are integers randomly selected from [1, N P ; , and are all independent variables; rand has a value range of [0, 1]; is the optimal solution vector in the g-th generation population; is the sub-optimal solution vector in the g-th generation population; is the worst solution vector in the g-th generation population; is a random number in [0, 0.5]; , , , , and are 6 different weighting functions respectively; exp is the natural exponential function.

[0025] In a second aspect, the present invention also provides a device for predicting the passenger evacuation time in the case of ship inclination, including: An inverse learning module, which is used to take the training set as the input of the vector weighted average algorithm, perform iterative solution, and generate a vector set in an inverse learning manner during iteration; the training set includes the personnel categories on the ship, as well as the evacuation speed and evacuation time of each category of personnel; A dynamic candidate module, which is used to determine the candidate value for the next iteration based on the candidate value of the current iteration and the product of a preset coefficient, and determine the MeanRule value based on the candidate value of the next iteration; A model training module, which is used to update the vector positions in the vector set based on the MeanRule value, determine the weights of the DELM model based on the vectors at the updated positions, and train the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model; A prediction module, which is used to input a test set into the passenger evacuation time prediction model to obtain the total time required for the evacuation of personnel on the ship.

[0026] In a third aspect, the present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions, and when the programs or instructions are executed by a processor, the steps in the method for predicting the passenger evacuation time in the case of ship inclination described in any one of the above can be implemented.

[0027] The beneficial effects of adopting the above implementation method are as follows: The present invention adopts a DELM model obtained by stacking multiple deep learning machines, which can better realize the mapping of data features, helps to improve the prediction accuracy. By using the training set as the input of the vector weighted average algorithm, iterative solution is carried out, and a vector set is generated in a reverse learning manner during iteration, and then the weights of the DELM model are optimized through this vector set, which can overcome the problem of unstable network output of the DELM model. Moreover, generating a vector set in a reverse learning manner can also expand the search space and enhance population diversity, thereby optimizing the accuracy and reliability of evacuation time prediction. Furthermore, the present invention improves the existing vector weighted average algorithm through a reverse learning manner, avoiding the problems of the vector weighted average algorithm falling into local optimal solutions and slow convergence speed, thus solving the technical problems of low prediction accuracy, speed and reliability existing in the existing methods for predicting the passenger evacuation time in the case of ship inclination. Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0029] Figure 1 It is a flowchart of an embodiment of the method for predicting the passenger evacuation time in the case of ship inclination provided by the present invention; Figure 2 It is a flowchart of another embodiment of the method for predicting the passenger evacuation time in the case of ship inclination provided by the present invention; Figure 3Flowchart of the vector weighted average algorithm provided by the present invention; Figure 4 Prediction result graph in an embodiment of the method for predicting passenger evacuation time in case of ship inclination provided by the present invention; Figure 5 Fitness curve graph in an embodiment of the method for predicting passenger evacuation time in case of ship inclination provided by the present invention; Figure 6 Principle block diagram of an embodiment of the device for predicting passenger evacuation time in case of ship inclination provided by the present invention; Figure 7 Structural schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed implementation manners

[0030] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0031] In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0032] In the embodiments of the present invention, the terms "include" and "have" and any of their deformations are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment including a series of steps or modules does not necessarily have to be limited to those clearly listed steps or modules, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or equipment.

[0033] In the embodiments of the present invention, the naming or numbering of steps does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The already named or numbered process steps can be changed in the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0034] Referring to "embodiments" herein means that the specific features, structures or characteristics described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The phrase appears in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0035] The present invention provides a method and device for predicting the passenger evacuation time in the case of ship inclination, which will be described separately below.

[0036] As Figure 1 shown, the present invention provides a method for predicting the passenger evacuation time in the case of ship inclination, including: S101. Taking the training set as the input of the vector weighted average algorithm, performing iterative solution, and generating a vector set in the way of reverse learning during iteration; the training set includes the personnel categories on the ship, as well as the evacuation speed and evacuation time of each category of personnel.

[0037] It can be understood that, in order to obtain the training set, the present invention uses Pathfinder software to simulate the personnel evacuation process in the cruise ship inclination scenario. There is a reaction time from the evacuation alarm to the start of personnel evacuation. The reaction time of personnel during the day is calculated as:

[0038] where is the standard deviation; is the individual reaction time; is the mean value.

[0039] The position update of personnel is described as:

[0040]

[0041] where is the velocity vector in the personnel evacuation direction; is the velocity of personnel in the current direction; is the evacuation speed of personnel at the next moment; is the position of personnel at the next moment; is the position of personnel at the current moment; is the maximum acceleration of personnel; is the time step.

[0042] Next, taking a large cruise ship as the research object, a real-time personnel information data set required by the present invention is constructed. Importing the cruise ship CAD drawings into the Pathfinder software, establishing a cruise ship personnel evacuation simulation model, and then setting according to the personnel reaction time stipulated by IMO, simulating different evacuation scenarios when the cruise ship inclines, so as to obtain the training set.

[0043] S102. Determining the candidate value for the next iteration based on the product of the candidate value of the current iteration and the preset coefficient, and determining the MeanRule value of the vector weighted average algorithm based on the candidate value of the next iteration.

[0044] It can be understood that for the candidate value corresponding to the first iteration, it can be determined based on the maximum number of iterations. During subsequent iteration processes, the candidate value for the next iteration is determined according to the candidate value of the current iteration. The way of dynamically adjusting the candidate value is the dynamic candidate mechanism provided by the present invention.

[0045] S103. Update the vector positions in the vector set based on the MeanRule value, determine the weights of the DELM model based on the vectors at the updated positions, and train the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model.

[0046] It can be understood that the DELM model is obtained by stacking multiple ELM (extreme learning machine) models. MeanRule is a function in the vector weighted average algorithm (MINFO). The present invention adjusts the MeanRule value through the dynamic candidate mechanism, thereby updating the vector positions in the vector set, and further optimizing the weights of the DELM model to overcome the problem that the unreasonable weight setting of the DELM model leads to unstable network output.

[0047] S104. Input the test set into the passenger evacuation time prediction model to obtain the total time required for the evacuation of the personnel on the ship.

[0048] It can be understood that the test set includes the types of personnel on the ship to be predicted, as well as the evacuation speed and evacuation time of each type of personnel.

[0049] Aiming at the problem that the prediction accuracy of the personnel evacuation time is greatly reduced when there is a lack of personnel information data and parameter mismatch, the present invention improves the vector weighted average algorithm by using reverse learning and the dynamic candidate mechanism, and proposes an evacuation time prediction method based on MINFO-DELM. DELM is a deep learning network model stacked by multiple extreme learning machines, which can better realize the mapping of data features, and MINFO provides the optimal parameters for DELM. The personnel information data set and MINFO-DELM are used for evacuation time prediction. Experiments show that MINFO-DELM is more accurate in predicting the evacuation time and shows good prediction effects. For the DELM without considering parameter optimization, there will be errors of different degrees in the prediction results. During the prediction process of the evacuation time test set, the highest determination coefficient of MINFO-DELM is 0.9999.

[0050] The method provided by the present invention can be implemented based on software running on a terminal or a server. The terminal can be a computer terminal, a mobile phone terminal, or an electronic terminal installed on a ship. The server can be an edge server or a cloud server. Whether it is a terminal or a server, both include a memory and a processor in terms of hardware structure. An application for executing the above method is stored on the memory. The processor loads the application on the memory and executes the program to implement the above method.

[0051] MINFO Based on Reverse Learning and Dynamic Candidate Mechanism: In the research on predicting the evacuation time of personnel using the DELM model, parameter selection is particularly crucial. The present invention proposes the MINFO-DELM method, which optimizes the weights of the DELM model through MINFO (vector weighted average algorithm). The vector weighted average algorithm performs excellently in terms of convergence speed. However, when facing more complex problems, this algorithm is prone to falling into local optimal solutions. Therefore, the present invention improves the vector weighted average algorithm by using reverse learning and dynamic candidate mechanism, aiming to expand the search space and enhance population diversity, thereby optimizing the accuracy and reliability of evacuation time prediction.

[0052] The principle of the DELM model is as follows: ELM has the advantages of fast calculation speed and low complexity. Suppose there are N samples , where the input data is and the expected output is , and g(x) is the activation function. The output function of the extreme learning machine can be expressed as:

[0053] Among them, is the system bias; is the weight of the connection between the i th hidden layer neuron and the input layer neuron; is the weight of the connection between the i-th hidden layer neuron and the output layer neuron.

[0054] Select ω and b to make the error between the ELM output and the expected output close to 0, which can be expressed as:

[0055] The ELM output function of N random samples can be expressed as:

[0056] Rewrite the above formula into matrix form, which can be expressed as:

[0057]

[0058]

[0059]

[0060] Among them, H is the output matrix of the hidden layer in ELM; T is the expected output vector.

[0061] To further improve the generalization ability and stability of ELM, the kernel function is introduced into ELM. The output function of the Kernel Extreme Learning Machine (KELM) is expressed as:

[0062]

[0063] Among them, is the selected kernel function; C is the regularization coefficient; I is the identity matrix; where is the output of the i th hidden layer neuron.

[0064] DELM is a deep learning network stacked by multiple ELMs, which can better realize the mapping of data features. It can not only improve the prediction accuracy of the model, but also enhance the generalization ability of the model. ELM-AE is the basic unit of DELM. The input sample x is used as the input of the first ELM-AE, and the target output approximates the input x and obtains the output weight . Subsequently, the output of the first ELM-AE hidden layer is used as the input of the second ELM-AE, and the target output approximates . And so on, finally obtaining the output weight matrix of the last layer, thus completing the training process of DELM.

[0065] The modeling process of MINFO-DELM: Due to the unreasonable setting of its weights, DELM is prone to unstable network output. To solve this problem, in this section, MINFO is used to optimize the weights of DELM to overcome the deficiencies of DELM. The flow chart of MINFO-DELM is as Figure 2As shown in the figure. There are 1,593 groups of data in the evacuation experiment. The ratio of the training samples to the test samples is set to 13:2. 1,300 groups of training samples are randomly selected, and 200 groups are selected from the remaining samples as the test samples. The personnel information dataset has been obtained previously. Therefore, six input variables are determined to be location, personnel category, speed, shoulder width, evacuation distance, and evacuation time, and the model output is the evacuation time of each person. The parameter settings of MINFO-DELM are as follows: the number of hidden layers is 18, the number of hidden layer nodes is 36, the weight dimension is 108, the upper bound of the weight is -1, the lower bound of the weight is 1, the size of the MINFO population is 20, and the maximum number of iterations is 100. First, MINFO-DELM is trained with 1,300 groups of training sets; then, the fitness value is evaluated to select the optimal model parameters; finally, the trained MINFO-DELM is used to predict the test set and output the prediction results.

[0066] The evaluation indicators used in this embodiment are the coefficient of determination (R2), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), which can be expressed as:

[0067]

[0068]

[0069]

[0070] Among them, n is the number of test samples; is the actual evacuation time; is the i th average value of the actual evacuation time; is the evacuation time predicted by the model.

[0071] In some embodiments, the training set is used as the input of the vector weighted average algorithm for iterative solution, including: Using the training set, weight boundaries, maximum number of iterations, weight dimension, and fitness function as the input of the vector weighted average algorithm for iterative solution; Among them, the fitness function is the sum of the mean square errors of the evacuation times of each type of person in the training set and the mean square errors of the evacuation times of each type of person in the test set, and the test set is used to test the trained DELM model.

[0072] It can be understood that the principle of the vector weighted average algorithm is as follows: The vector weighted average algorithm has the characteristics of fast convergence speed and strong global search ability. The vector weighted average algorithm mainly includes three steps: the update rule stage, the vector merging stage, and the local search stage.

[0073] To solve the problems that the vector weighted average algorithm (MINFO) is prone to falling into local optimal solutions and has a slow convergence speed, this paper improves the vector weighted average algorithm by using reverse learning and a dynamic candidate mechanism, and then uses the optimized MINFO algorithm to optimize the weights of DELM. The size of the vector set (population) of MINFO is 20, the maximum number of iterations is 100, the upper bound of the weight is 1, and the lower bound of the weight is -1. The inputs of MINFO are the size of the vector set (population), the maximum number of iterations, the upper bound of the weight, the lower bound of the weight, the weight dimension, and the fitness function, and the output is the weight. The fitness function of MINFO is the sum of the mean square errors of the evacuation time training samples and the test samples.

[0074] In some embodiments, when iterating, a reverse learning method is used to generate a vector set, including: When iterating, a reverse vector is generated based on the current vector in the vector set of the current iteration, and the vector closer to the global optimal solution is determined from the current vector and the reverse vector. Based on the vector closer to the global optimal solution, the current vector in the vector set of the current iteration is replaced to obtain the vector set of the next iteration; Among them, the initial vector set is the training set.

[0075] It can be understood that assuming that the vector weighted average algorithm has Np vectors in the D-dimensional search space, the search domain is ( ={ , ,…, }, l = 1,2,…, Np).

[0076] ① Update rule stage. The vector weighted average algorithm uses the mean rule to update the vector position, and a convergence acceleration part is added to the update rule operator to improve the global search ability. The mean rule is applied to MeanRule and can be calculated as:

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086] wherein, is the objective function; , is a function in MeanRule; , and are integers randomly selected from [1, NP]; , , are independent variables; takes a value of 0.02; the value range of rand is [0, 1]; is a function; is the optimal solution vector in the g-th generation population; is the sub-optimal solution vector in the g-th generation population; is the worst solution vector in the g-th generation population; is a random number with a range of [0, 0.5]; , , , , , are 6 different weighting functions respectively; exp is the natural exponential function.

[0087] Scaling factor and can be calculated as:

[0088]

[0089] wherein, g represents the number of iterations; rand is a random number with a value range of [0, 1]; Maxg is the maximum number of iterations.

[0090] The new solution vector can be expressed as:

[0091]

[0092] Among them, CA is the convergence acceleration part.

[0093] If rand < 0.5, the update rule can be described as:

[0094]

[0095] If rand ≥ 0.5, the update rule can be described as:

[0096]

[0097] Among them, and are the new solution vectors in the g-th generation population.

[0098] At this time, the scaling rate of the vector and can be calculated as:

[0099]

[0100] ② Vector merging stage. Combine the vectors calculated previously with rand the vector when < 0.5 to generate a new vector which can be described as:

[0101] Among them, is the new vector obtained by merging the vectors in the g-th generation.

[0102] ③ Local search stage. To prevent the vector weighted average algorithm from falling into the local optimal situation, the local search operator of the vector weighted average algorithm searches for the global optimal solution by considering the global optimal solution and MeanRule to achieve the purpose of converging to the global optimum. According to the local search operator, if r < 0.5, a new vector can be generated and expressed as:

[0103]

[0104]

[0105]

[0106]

[0107] Among them, p is a random number within the range of (0, 1); and is a random number.

[0108] In some embodiments, determining a better vector from the current vector and the reverse vector includes: Evaluating the fitness function values corresponding to the current vector and the reverse vector, and determining a better vector from the current vector and the reverse vector based on the fitness function values.

[0109] It can be understood that the fitness function is the sum of the mean square error of the evacuation time training samples and the mean square error of the test samples.

[0110] In some embodiments, generating a reverse vector based on the current vector in the vector set of the current iteration includes: Adding the maximum value and the minimum value of the current vector in the vector set of the current iteration, and subtracting the current vector to obtain the reverse vector.

[0111] It can be understood that reverse learning determines whether an alternative solution is acceptable by evaluating the fitness function value of the current solution. Reverse learning will generate a reverse vector based on the current vector to approach the global optimal solution. Referring to the following reverse learning formula, assume that the solution in MINF0 is , then the reverse solution is calculated as:

[0112] Among them, is the initial solution set of MINF0; is the minimum value of the current vector; is the maximum value of the current vector; comparing the fitness values of and , saving the better vector and deleting the worse vector.

[0113] In some embodiments, the weights of the DELM model based on the vectors at the updated positions include: Merging the vectors at the updated positions to generate a new vector; Determining a local search operator, and determining the optimal fitness of the new vector based on the local search operator; Updating the global optimal fitness based on the optimal fitness of the new vector in two adjacent iterations; When it is determined that the latest global optimal fitness meets the preset iteration condition, optimizing the weights of the DELM model based on the last generated new vector.

[0114] It can be understood that Figure 3The overall process of MINFO is given. The basic steps of MINFO are as follows: (1) Initialize the parameters, set the value of the parameter Np and the maximum number of iterations Maxg , and generate a vector set (population) using the above reverse learning formula.

[0115] (2) Enter the update rule stage, calculate MeanRule, and calculate the vectors and .

[0116] (3) Enter the vector merging stage to generate a new vector .

[0117] (4) Start local search and calculate the local search operator.

[0118] (5) Calculate the fitness of the updated vector.

[0119] (6) Compare the fitness obtained in each iteration with the optimal fitness obtained in the previous iteration, and update the global optimal fitness.

[0120] (7) Determine whether the iteration condition is satisfied. If it is satisfied, output the global optimal solution; otherwise, return to step (2).

[0121] In this embodiment, the principle of ELM (Extreme Learning Machine) is deeply analyzed, and on this basis, the DELM model is constructed. Further, in view of the complexity and challenges of evacuation time prediction, the modeling process and method of MINFO-DELM are elaborated in detail.

[0122] In some embodiments, the calculation formula for the candidate value is as follows:

[0123]

[0124] where is the candidate value for the i th iteration, Maxg represents the maximum number of iterations, i is the current iteration number, is the candidate value for the i +1th iteration.

[0125] It can be understood that in each iteration of MINFO, the dynamic candidate mechanism can be represented by the above calculation formula for the candidate value, so as to dynamically adjust the MeanRule in each iteration process.

[0126] In some embodiments, the calculation formula for the MeanRule value is as follows:

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133]

[0134]

[0135]

[0136]

[0137] Among them, is the objective function; and are functions in MeanRule ; , and are randomly selected integers from [1, N P ; , and are all independent variables; takes the value of 0.02; rand has a value range of [0, 1]; is the optimal solution vector in the g-th generation population; is the sub-optimal solution vector in the g-th generation population; is the worst solution vector in the g-th generation population; is a random number in [0, 0.5]; , , , , and are 6 different weighting functions respectively; exp is the natural exponential function.

[0138] It can be understood that in this embodiment, the dynamic candidate mechanism is adopted to rewrite MeanRule as:

[0139] Compared with the existing MeanRule calculation formula, in this embodiment, candidate values are added for weighted calculation to optimize the weights of the DELM model and improve the certainty of the model prediction results.

[0140] In some embodiments, the method for predicting the passenger evacuation time in the case of ship inclination provided by the present invention includes the following steps: 1. The ship evacuation software is used to simulate the personnel evacuation process in the case of ship inclination, and the personnel evacuation speed is obtained.

[0141] Next, taking a large cruise ship as the research object, a real-time personnel information dataset required for this study is constructed. The CAD drawings of the cruise ship are imported into the Pathfinder software to establish a personnel evacuation simulation model of the cruise ship, and then set according to the personnel response time specified by the IMO to simulate different evacuation scenarios when the cruise ship is inclined.

[0142] In this embodiment, the evacuation process of personnel in the simulated cruise ship inclination scenario is carried out, and a total of 1593 people participated in the evacuation. Among them, adult males accounted for 32.8%, adult females accounted for 29.4%, the elderly accounted for 35.6%, and children accounted for 2.2%. Different personnel position distributions are set. The personnel evacuation speed has been determined previously. In the personnel evacuation simulation model of the large cruise ship, the relevant simulation parameters are set as shown in Table 1.

[0143] Table 1: Simulation parameters

[0144] The cruise ship personnel evacuation simulation model automatically records the data during the personnel evacuation process. The main data are: position, personnel category, speed, shoulder width, evacuation distance, and evacuation time. The original dataset in the Pathfinder software lacks some personnel information affecting the evacuation time, such as personnel type, evacuation speed, and passenger shoulder width, and this information needs to be manually added and recorded during the simulation. A (partial) personnel information dataset related to the evacuation time is created as shown in Table 2.

[0145] Table 2: Personnel information dataset (partial)

[0146] Table 2 records the data related to the personnel evacuation time. Only part of the data is shown due to space limitations, and there are a total of 1593 groups of data.

[0147] 2. Improvement of the vector weighted average algorithm To solve the problems that the vector weighted average algorithm is prone to falling into local optimal solutions and has a slow convergence rate, this paper improves the vector weighted average algorithm by using opposition-based learning and a dynamic candidate mechanism, and then uses MINFO to optimize the weights of DELM. The size of the vector set (population) of MINFO is 20, the maximum number of iterations is 100, the upper bound of the weight is 1, and the lower bound of the weight is -1. The inputs of MINFO are the size of the vector set (population), the maximum number of iterations, the upper bound of the weight, the lower bound of the weight, the weight dimension, and the fitness function, and the output is the weight. The fitness function of MINFO is the sum of the mean square error of the evacuation time training samples and the mean square error of the test samples.

[0148] Opposition-based learning determines whether an alternative is desirable by evaluating the fitness function value of the current solution. Opposition-based learning generates an opposite vector based on the current vector to approach the global optimal solution. Let the solution in MINFO be , then the opposite solution is calculated as:

[0149] where, is the initial solution of MINFO; is the minimum value of the current vector; is the maximum value of the current vector; Compare the fitness values of and , save the better vector, and delete the worse vector.

[0150] During each iteration of MINFO, the dynamic candidate mechanism can be described as:

[0151]

[0152] where, is the candidate value; Maxg represents the maximum number of iterations; i is the current iteration number; is i +1 generation candidate value.

[0153] After being rewritten using the dynamic candidate mechanism, MeanRule is:

[0154] Figure 3 Figure 50 shows the overall process of MINFO. The basic steps of MINFO are as follows: (1) Initialize the parameters, set the values of the parameters Np and the maximum number of iterations Maxg , and use the opposition-based learning formula (48) to generate the vector set (population).

[0155] (2) Enter the update rule stage, calculate MeanRule, and calculate the vectors and .

[0156] (3) Enter the vector merging stage to generate a new vector .

[0157] (4) Start local search and calculate the local search operator.

[0158] (5) Calculate the fitness of the updated vector.

[0159] (6) Compare the fitness obtained in each iteration with the optimal fitness obtained in the previous iteration and update the global optimal fitness.

[0160] (7) Determine whether the iteration condition is satisfied. If it is satisfied, output the global optimal solution; otherwise, return to step (2).

[0161] 3. Experimental Results and Analysis of Evacuation Time Prediction Prediction Results of Evacuation Time: The equipment used in the experiment is a desktop computer with an Intel(R) Core(TM) i9-13900K CPU and an NVIDIA GeForce RTX 4080 GPU, with a main frequency of 3.00 GHz, 128 GB of memory, and Windows 10 Professional Edition. To reduce the influence of the weights in the DELM network on the prediction accuracy of the DELM network, MINFO is used to search for the optimal parameters of the DELM. To avoid the influence of the differences in position, category, speed, shoulder width, and the numerical values of the actual evacuation time and the dimension units, the created personnel information dataset is normalized. The evacuation time of personnel is predicted using MINFO-DELM, and the prediction results of the evacuation time of personnel are obtained, as shown in Table 3.

[0162] Table 3: Prediction Results of Evacuation Time (Partial)

[0163] As shown in Table 3, MINFO-DELM can accurately predict the evacuation time of personnel of different genders and ages, and the maximum difference between the actual evacuation time and the predicted evacuation time is 6 s. Since ASET is given in the design stage, the time predicted in this paper is RSET. The RSET of the evacuated personnel in Table 3 is less than ASET, indicating that this part of the personnel can safely reach the assembly station.

[0164] Comparison of Evacuation Time Prediction Results of Different Algorithms: The number of hidden layer neurons in MINFO-DELM is selected by trial and error. During the test, the number of hidden layers is 18, the number of hidden layer nodes is 36, the maximum number of iterations of MINFO is 100, and the population size is 20. MINFO-DELM is used to predict the evacuation time, and the prediction results of the evacuation time are as Figure 4 shown, and the fitness curve of MINFO-DELM is as Figure 5 shown.

[0165] Figure 4 It shows that when using MINFO-DELM for prediction, the RMSE of the predicted evacuation time value is 2.91. Figure 5 It shows that MINFO-DELM reaches convergence after 40 iterations.

[0166] As Figure 6 shown, the present invention also provides a device 600 for predicting the evacuation time of passengers in the case of ship tilt, including: A reverse learning module 601, which is used to take the training set as the input of the vector weighted average algorithm, perform iterative solution, and generate a vector set in the reverse learning manner during iteration; the training set includes the personnel categories on the ship, as well as the evacuation speed and evacuation time of each category of personnel; A dynamic candidate module 602, which is used to determine the candidate value for the next iteration based on the product of the candidate value of the current iteration and a preset coefficient, and determine the MeanRule value of the vector weighted average algorithm based on the candidate value for the next iteration; A model training module 603, which is used to update the vector positions in the vector set based on the MeanRule value, determine the weights of the vector DELM model based on the updated positions, and train the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model; A prediction module 604, which is used to input the test set into the passenger evacuation time prediction model to obtain the total time required for the personnel on the ship to evacuate.

[0167] The device for predicting the evacuation time of passengers in the case of ship tilt provided in the above embodiment can implement the technical solutions described in the above embodiment of the method for predicting the evacuation time of passengers in the case of ship tilt. For the specific implementation principles of the above modules or units, reference can be made to the corresponding content in the above embodiment of the method for predicting the evacuation time of passengers in the case of ship tilt, which will not be elaborated here.

[0168] As Figure 7 shown, the present invention also correspondingly provides an electronic device 700. The electronic device 700 includes a processor 701, a memory 702, and a display 703. Figure 7 Only some components of the electronic device 700 are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented.

[0169] In some embodiments, the memory 702 may be an internal storage unit of the electronic device 700, such as the hard disk or memory of the electronic device 700. In other embodiments, the memory 702 may also be an external storage device of the electronic device 700, such as a plug-in hard disk equipped on the electronic device 700, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0170] Furthermore, the memory 702 may also include both the internal storage unit and the external storage device of the electronic device 700. The memory 702 is used to store the application software installed in the electronic device 700 and various types of data.

[0171] In some embodiments, the processor 701 may be a central processing unit (CPU), a microprocessor, or other data processing chips, and is used to run the program code stored in the memory 702 or process data, such as the method for predicting the passenger evacuation time in the case of ship inclination in the present invention.

[0172] In some embodiments, the display 703 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. The display 703 is used to display the information in the electronic device 700 and to display a visual user interface. The components 701-703 of the electronic device 700 communicate with each other through a system bus.

[0173] In some embodiments of the present invention, when the processor 701 executes the program for predicting the passenger evacuation time in the case of ship inclination in the memory 702, the following steps may be implemented: Taking the training set as the input of the vector weighted average algorithm, performing iterative solution, and generating a vector set in a reverse learning manner during iteration; the training set includes the personnel categories on the ship, as well as the evacuation speed and evacuation time of each category of personnel; Determining the candidate value for the next iteration based on the product of the candidate value of the current iteration and a preset coefficient, and determining the MeanRule value based on the candidate value of the next iteration; Updating the vector positions in the vector set based on the MeanRule value, determining the weights of the DELM model based on the vectors at the updated positions, and training the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model; Inputting the test set into the passenger evacuation time prediction model to obtain the total time required for the personnel on the ship to evacuate.

[0174] It should be understood that when the processor 701 executes the passenger evacuation time prediction program in the memory 702 in the case of ship inclination, in addition to the above functions, other functions can also be realized. For specific details, reference can be made to the description of the corresponding method embodiments above.

[0175] Furthermore, the type of the electronic device 700 mentioned in the embodiments of the present invention is not specifically limited. The electronic device 700 can be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of the portable electronic device include, but are not limited to, portable electronic devices equipped with IOS, android, microsoft or other operating systems. The above portable electronic devices can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (such as a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 700 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (such as a touch panel).

[0176] On the other hand, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to execute the passenger evacuation time prediction method provided by the above various methods. The method includes: Taking the training set as the input of the vector weighted average algorithm, performing iterative solution, and generating a vector set in the way of reverse learning during iteration; the training set includes the personnel categories on the ship, as well as the evacuation speed and evacuation time of each category of personnel; Determining the candidate value for the next iteration based on the product of the candidate value of the current iteration and a preset coefficient, and determining the MeanRule value based on the candidate value of the next iteration; Updating the vector positions in the vector set based on the MeanRule value, determining the weights of the DELM model based on the vectors at the updated positions, and training the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model; Inputting the test set into the passenger evacuation time prediction model to obtain the total time required for the personnel on the ship to evacuate.

[0177] Those skilled in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a disk, an optical disk, a read-only memory or a random access memory, etc.

[0178] The above has introduced in detail the method and device for predicting the passenger evacuation time in the case of ship inclination. In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for predicting passenger evacuation time when a ship tilts, characterized in that: include: The training set is used as the input of the vector weighted average algorithm for iterative solution, and the vector set is generated by reverse learning during iteration; The training set includes the categories of personnel on the ship, and the evacuation speed and evacuation time of each category of personnel; Determine the candidate value for the next iteration based on the product of the candidate value for this iteration and the preset coefficient, and determine the MeanRule value based on the candidate value for the next iteration; The vector position in the vector set is updated based on the MeanRule value, the weight of the DELM model of the vector at the updated position is obtained, and the optimized DELM model is trained based on the training set to obtain a passenger evacuation time prediction model; The test set is input into the passenger evacuation time prediction model to obtain the total time required for evacuation of personnel on the ship.

2. The method for predicting passenger evacuation time in the event of ship tilting according to claim 1, characterized in that: The training set is used as the input of the vector weighted average algorithm for iterative solution, including: The training set, weight boundary, maximum number of iterations, weight dimension and fitness function are used as inputs of the vector weighted average algorithm for iterative solution; The fitness function is the sum of the mean square error of the evacuation time of each type of personnel in the training set and the mean square error of the evacuation time of each type of personnel in the test set, and the test set is used to test the trained DELM model.

3. The method for predicting passenger evacuation time in the event of ship tilting according to claim 2, characterized in that: During iteration, a reverse learning method is used to generate a vector set, including: During iteration, a reverse vector is generated based on the current vector in the vector set of this iteration, and a vector close to the global optimal solution is determined from the current vector and the reverse vector. The current vector in the vector set of this iteration is replaced based on the vector close to the global optimal solution to obtain the vector set of the next iteration; The initial vector set is the training set.

4. The method for predicting passenger evacuation time in the event of ship tilting according to claim 3, characterized in that: Determine a better vector from the current vector and the reverse vector, including: The fitness function values ​​corresponding to the current vector and the reverse vector are evaluated, and a better vector is determined from the current vector and the reverse vector based on the fitness function values.

5. The method for predicting passenger evacuation time in the event of ship tilting according to claim 3, characterized in that: Generate a reverse vector based on the current vector in the vector set of this iteration, including: The reverse vector is obtained by subtracting the current vector from the sum of the maximum and minimum values ​​of the current vector in the vector set of this iteration.

6. The method for predicting passenger evacuation time in the event of ship tilting according to claim 1, characterized in that: The weights of the DELM model based on the vector of updated positions include: Merge the vectors of the updated positions to generate a new vector; Determining a local search operator, and determining the optimal fitness of the new vector based on the local search operator; Based on the optimal fitness of the new vector in two adjacent iterations, the global optimal fitness is updated; When it is determined that the latest global optimal fitness satisfies the preset iteration condition, the weight of the DELM model is optimized based on the latest generated new vector.

7. The method for predicting passenger evacuation time in the event of ship tilting according to claim 1, characterized in that: The calculation formula for candidate values ​​is as follows: in, For the i The candidate value of the iteration, Maxg represents the maximum number of iterations, i is the current iteration number, For the i +1 candidate value for iteration.

8. The method for predicting passenger evacuation time in the event of ship tilting according to any one of claims 1 to 7, characterized in that: The calculation formula of MeanRule value is as follows: in, is the objective function; and for MeanRule Functions in , and is from [1, N P ]; , and All are independent variables; rand The value range of is [0, 1]; is the optimal solution vector in the g-th generation population; is the suboptimal solution vector in the g-th generation population; is the worst solution vector in the g-th generation population; is a random number in [0, 0.5]; , , , , and There are 6 different weighting functions respectively; exp is the natural exponential function.

9. A device for predicting passenger evacuation time in the event of a ship tilting, characterized in that: include: A reverse learning module is used to use the training set as the input of the vector weighted average algorithm to perform iterative solution, and to generate a vector set by reverse learning during iteration; the training set includes the categories of personnel on the ship, and the evacuation speed and evacuation time of each category of personnel; A dynamic candidate module, used to determine the candidate value of the next iteration based on the product of the candidate value of the current iteration and the preset coefficient, and determine the MeanRule value based on the candidate value of the next iteration; A model training module, used for updating the vector position in the vector set based on the MeanRule value, weighting the vector DELM model based on the updated position, and training the optimized DELM model based on the training set to obtain a passenger evacuation time prediction model; The prediction module is used to input the test set into the passenger evacuation time prediction model to obtain the total time required for the evacuation of personnel on the ship.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the method for predicting passenger evacuation time in the event of a ship tilting as described in any one of claims 1 to 8.

Images