Submersible unmanned ship resistance forecast approximate model optimization method considering fuzziness

Abstract

The invention provides a submerged unmanned ship resistance forecast approximation model optimization method considering fuzziness, and the method comprises the steps: taking a submerged unmanned ship as a target ship type, and generating a multi-source data set through a ship model test, a numerical simulation test and a cloud computing technology; establishing an approximation model based on a feedforward neural network model, including establishment of an input layer, a hidden layer and an output layer; comparing the approximation model with original data, finding out data which is greatly different from the original data, and finding out six parameters which have great influence on resistance through sensitivity analysis; establishing a triangular membership degree of the six parameters, defining a rule base of the triangular membership degree, and finally obtaining a final resistance value according to defuzzification. According to the method, the approximate model established by the neural network is optimized through fuzzy optimization, the uncertainty caused by multi-source data is effectively reduced, the precision and reliability of resistance optimization of the submerging unmanned ship are improved, the navigation efficiency of the submerging unmanned ship is better improved, the energy consumption is reduced, and the adaptive capacity of the submerging unmanned ship in a complex marine environment is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of data fusion, and in particular to an optimization method for an approximate model of resistance prediction of a submersible and floating unmanned vessel taking fuzziness into consideration. Background Art

[0002] As an advanced marine vehicle, the submersible unmanned vessel combines many advantages of unmanned boats and can navigate at a certain depth below the water surface, with only part of the hull structure exposed above the water surface, while the majority of the hull is underwater. Compared with traditional surface vehicles, this unique navigation method enables it to effectively avoid the impact of waves and other marine environmental factors, significantly reducing the wave resistance of the hull, while avoiding problems such as structural damage that waves may cause to the hull. With the development of the marine field, the application prospects of submersible unmanned vessels are becoming increasingly broad, and they have important potential value in many fields such as marine monitoring and marine resource development. Resistance optimization, as a key link in improving the performance of submersible unmanned vessels, plays a vital role in improving their navigation efficiency, reducing energy consumption, and enhancing their adaptability in complex marine environments.

[0003] Neural network technology has already found some application in the field of ship design. The primary purpose of using neural networks to build approximate models is to improve the efficiency of design, control, and optimization, reducing reliance on complex physical models. Building approximate models through neural networks can significantly enhance the predictive accuracy of surrogate models and can be applied to situations with uncertainty in multiple variables, providing more reliable data support for the design and optimization of submersible and floating unmanned vessels. Furthermore, fuzzy optimization, a method that combines fuzzy logic and optimization algorithms, is particularly well-suited for addressing the uncertainty and complexity of ship design and control. Its core concept is to utilize fuzzy systems to handle uncertain variables and adjust input parameters through optimization algorithms to improve ship performance. Constructing triangular membership fuzzy numbers is a relatively intuitive method for determining membership, effectively addressing the influence of uncertainty. While the application of fuzzy optimization technology in ship design is relatively limited, its potential for addressing complex problems holds great promise for future development.

[0004] Although existing neural network and fuzzy optimization techniques have found some application in ship design, they still face challenges in optimizing resistance for submersible and floating unmanned vessels. Current research on resistance optimization for submersible and floating unmanned vessels often focuses on the application of a single technique, lacking comprehensive consideration of the uncertainty inherent in multi-source data. Multi-source datasets generated through actual ship model tests, numerical simulations, and cloud computing technologies contain numerous uncertainties, which significantly impact the development and optimization of neural network approximate models. Existing methods are often ineffective in addressing these uncertainties, resulting in discrepancies between the optimized models and the actual situation. Furthermore, existing techniques for optimizing neural network models using fuzzy optimization lack a systematic approach and rule base, hindering the full utilization of the advantages of fuzzy optimization technology. This, in turn, limits the accuracy and reliability of resistance optimization for submersible and floating unmanned vessels. Therefore, effectively reducing the uncertainty inherent in multi-source data and optimizing neural network approximate models using fuzzy optimization techniques to improve the accuracy and reliability of resistance optimization for submersible and floating unmanned vessels remains a pressing technical challenge in this field. Summary of the Invention

[0005] To address the aforementioned technical issues, a method for optimizing an approximate model for predicting resistance for submersible and floating unmanned vessels, taking into account fuzziness, is provided. This method optimizes the approximate model established by a neural network through fuzzy optimization, effectively reducing the uncertainty introduced by multi-source data and improving the accuracy and reliability of resistance optimization for submersible and floating unmanned vessels. This improves their navigation efficiency, reduces energy consumption, and enhances their adaptability in complex marine environments.

[0006] The technical means adopted in the present invention are as follows:

[0007] A method for optimizing the approximate model of resistance prediction for a submersible and floating unmanned vessel considering fuzziness is proposed, comprising:

[0008] S1. Taking submersible unmanned vessels as the target ship type, a multi-source data set was generated through ship model tests, numerical simulation tests, and cloud computing technology.

[0009] S2. Based on the feedforward neural network model, establish an approximate model, including the establishment of the input layer, hidden layer and output layer;

[0010] S3. Compare the approximate model with the original data to find the data with the largest difference from the original data, and find the six parameters with the largest impact on the resistance through sensitivity analysis;

[0011] S4. Establish the triangular membership of the six parameters and define its rule base, and finally obtain the final resistance value based on defuzzification.

[0012] Furthermore, in step S1, the ship model test and numerical simulation test are performed with the submersible unmanned ship as the target ship type, only the resistance characteristics of the main hull of the ship are considered, and the influence of the appendages on the resistance is ignored to obtain experimental data and simulation data.

[0013] Furthermore, in step S1, the cloud computing technology uses a reverse cloud generator to calculate six digital features, uses a forward cloud generator to expand data, sets the expansion multiple to 100 times, and obtains insufficient information supplement on the speed, diving depth, and resistance of the submersible unmanned ship.

[0014] Furthermore, in step S2, the feedforward neural network is used as the regression model. In order to quantify the fitting performance of the neural network, the mean square error (MSE) is used as an evaluation index. The calculation formula of the mean square error (MSE) is:

[0015]

[0016] Among them, MSE represents the number of samples, y i Indicates the actual value, Represents the predicted value.

[0017] Furthermore, in step S2, combined with the uncertainty of the mean square error, it also includes repeatedly processing each data point in the simulation data set and the experimental data set to increase the frequency of occurrence of samples.

[0018] Furthermore, step S3 specifically includes:

[0019] S31, import input variables and output variables;

[0020] S32. Calculate the disturbance of each input variable using the following formula:

[0021] ΔX i =0.1×(max(X i )-min(X i ))

[0022] Where ΔX i represents the disturbance amplitude of each input variable, X i represents the i-th input variable;

[0023] S33. Perform sensitivity analysis on each input parameter, establish a linear regression model to fit the original data, and predict the input variables after disturbance, and calculate the output change. The calculation formula is as follows:

[0024]

[0025] Among them, S i Indicates the sensitivity coefficient of the i-th input parameter, ΔYj Indicates the change in output, ΔX ij Indicates the change in the i-th input parameter, Y j and X ij Represent the original values of output and input respectively;

[0026] S34. Normalize the sensitivity coefficients of all parameters to obtain relative weights. The formula is as follows:

[0027]

[0028] Among them, W i represents the normalized sensitivity weight, ∑S represents the sum of all sensitivity coefficients;

[0029] S35. Use the least squares method to perform linear regression fitting. The formula is as follows:

[0030]

[0031] Among them, β0 represents the intercept term, β i represents the regression coefficient, ∈ represents the error term;

[0032] S36. Repeat the above steps until all samples in the input variables are input.

[0033] Furthermore, step S4 specifically includes:

[0034] S41, loading the neural network model and performing neural network prediction;

[0035] S42. Create a fuzzy system and define input variables and output variables;

[0036] S43. Define a fuzzy rule base;

[0037] S44, perform defuzzification;

[0038] S45. Perform neural network prediction. If there is a large difference with the original data, perform fuzzy optimization.

[0039] Furthermore, in step S43, a fuzzy rule base is defined, specifically including:

[0040] S431. Define the triangular membership of the input variables and the output variables. Set the input variables to three fuzzy ranges: large, medium, and small. Define five ranges for the output resistance.

[0041] S432. Set the permutations and combinations of the six variables to obtain their respective resistance values, and assign corresponding rule weights.

[0042] Furthermore, in step S44, the centroid method is used as a key step of defuzzification, which specifically includes:

[0043] The core idea of the centroid method is to calculate the centroid of the fuzzy output, that is, the weighted average of the output membership function. The specific formula is as follows:

[0044]

[0045] Among them, y represents the value of the output variable, μ(y) represents the output membership function, for each y value, it represents its membership (that is, the degree to which the value belongs to a certain fuzzy set), a and b both represent the range of output values, y defuzz Represents the clear value after defuzzification, that is, the final output value.

[0046] Compared with the prior art, the present invention has the following advantages:

[0047] 1. The present invention optimizes the approximate model established by the neural network through fuzzy optimization, effectively reducing the uncertainty brought by multi-source data and improving the accuracy of the model in predicting the resistance of submersible unmanned vessels in complex marine environments.

[0048] 2. The present invention provides an optimization method for the approximate model of resistance prediction of submersible and floating unmanned vessels taking into account fuzziness. First, an approximate model is established based on a neural network to find data with a large gap from the original data. The triangle membership is established through a fuzzy system, and a rule base is established for specific situations. The data with a large gap is optimized, and the data with a small gap is retained to establish an approximate model based on the neural network, so that the approximate model is more in line with the actual situation.

[0049] Based on the above reasons, the present invention can be widely promoted in fields such as data fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0051] Figure 1 Flow chart of the method of the present invention.

[0052] Figure 2 This is the effect diagram of the approximate model established by the neural network of the present invention.

[0053] Figure 3 This is a schematic diagram of sensitivity analysis of the present invention.

[0054] Figure 3The parameters of the horizontal axis are ship length, ship width, draft, bulbous bow length, bulbous bow width, parallel midbody, bilge radius, and arc top radius.

[0055] Figure 4 A parameter triangle membership diagram is established for the present invention.

[0056] Figure 5 This is a general comparison diagram before and after the optimization of the approximate model of the present invention. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0058] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0059] like Figure 1 As shown, the present invention provides an optimization method for an approximate model of resistance prediction of a submersible unmanned vessel considering fuzziness, comprising:

[0060] S1. Taking submersible unmanned vessels as the target ship type, a multi-source data set was generated through ship model tests, numerical simulation tests, and cloud computing technology.

[0061] S2. Based on the feedforward neural network model, establish an approximate model, including the establishment of the input layer, hidden layer and output layer;

[0062] S3. Compare the approximate model with the original data to find the data with the largest difference from the original data, and find the six parameters with the largest impact on the resistance through sensitivity analysis;

[0063] S4. Establish the triangular membership of the six parameters and define its rule base, and finally obtain the final resistance value based on defuzzification.

[0064] During specific implementation, as a preferred embodiment of the present invention, in step S1, the ship model test and numerical simulation test are based on a submersible unmanned ship as the target ship type, only considering the resistance characteristics of the main hull of the ship, ignoring the influence of the appendages on the resistance, and obtaining experimental data and simulation data.

[0065] In this example, the resistance of a submersible unmanned vessel was determined through three methods: ship model testing, numerical simulation, and cloud computing expansion. To ensure the accuracy and reliability of the research data, the data collection process through ship model resistance tests only considered the resistance characteristics of the main hull of this type of submersible unmanned vessel, ignoring the effects of the appendages on resistance. The dimensions of the water tank were 160.0m × 7.0m × 3.7m (length × width × water depth). With the submersible unmanned vessel as the target ship type, ship model tests and numerical simulation tests were conducted, with six different diving depth conditions set from 0# to 5#, corresponding to diving depths of 0m, 0.054m, 0.32m, 0.48m, 0.64m, and 0.96m, respectively. The test tests the hydrostatic resistance at multiple speed points from 0.4m / s to 1.7m / s, which are 0.4m / s, 0.6m / s, 0.8m / s, 1.0m / s, 1.1m / s, 1.2m / s, 1.3m / s, 1.4m / s, 1.5m / s, 1.6m / s and 1.7m / s, respectively. The corresponding Froude number (Fr) ranges from 0.1010 to 0.4291, and the Reynolds number (Re) ranges from 0.5284×106 to 2.2459×106. The comparison of the resistance of the ship model test and the numerical simulation under each working condition is carried out by Figure 2 and Figure 3 To express.

[0066] In specific implementation, as a preferred embodiment of the present invention, in step S1, the cloud computing technology uses the reverse cloud generator to calculate six digital features, uses the forward cloud generator to expand the data, sets the expansion multiple to 100 times, and obtains the insufficient information supplement of the speed, diving depth and resistance of the submersible unmanned ship. Figure 4 The figure shows the comparison of cloud expansion models when the diving depth is 0m.

[0067] In this embodiment, the fused dataset is used as the target sample for data augmentation. To prevent the augmented data from deviating too much from the original real data, constraint conditions are set based on the extreme values of V and D in the original dataset and allowing them to float by 10%. The specific value range is: 0.2000 < V < 1.800 (m / s), 0.0000 < D < 1.000 (m). Any augmented samples outside the following range will be excluded to ensure the accuracy and reliability of the augmented data. First, taking the 55 groups of "speed - resistance" data in the fused sample set as the target event, its six digital features are calculated using the reverse cloud generator. Subsequently, the forward cloud generator is used for data augmentation to solve the problems of small initial data volume and sparse distribution. The augmentation multiple is set to 100 times, obtaining the insufficient information supplementation of the speed, diving depth, and resistance of the submersible unmanned ship. After augmenting the data to 100 times, 5500 augmented data samples of the resistance performance of the submersible unmanned ship are obtained. Compared with the original fused sample set, the data samples in the speed range of 0.8 - 1 m / s and the diving depth range of 0 - 0.5 m in the augmented sample set are significantly enhanced, making the overall data distribution more reasonable and optimized.

[0068] During specific implementation, as a preferred implementation manner of the present invention, in step S2, the feedforward neural network is used as the regression model (FNN). To quantify the fitting performance of the neural network, the mean square error (MSE) is used as the evaluation index. The calculation formula of the mean square error (MSE) is:

[0069]

[0070] where MSE represents the number of samples, y i represents the actual value, represents the predicted value.

[0071] During specific implementation, as a preferred implementation manner of the present invention, in step S2, in combination with the uncertainty of the mean square error, it also includes repeating each data point in the simulation dataset and the experimental dataset to increase the occurrence frequency of the samples.

[0072] In this example, by comparing when the number of neurons is set to 10, when selecting repeated data points, it mainly depends on their key importance in model training, especially those samples that show greater uncertainty under specific conditions. Through this strategy, the diversity and representativeness of the dataset can be increased, thereby enhancing the generalization ability of the model, reducing the risk of overfitting, and ensuring that the model has robust prediction ability in a wider range of scenarios.

[0073] During specific implementation, as a preferred implementation manner of the present invention, step S3 specifically includes:

[0074] S31. Import the input variables and output variables;

[0075] S32. Calculate the disturbance of each input variable (change 10%) using the following formula:

[0076] ΔX i =0.1×(max(X i )-min(X i ))

[0077] Where ΔX i represents the disturbance amplitude of each input variable, X i represents the i-th input variable; in this embodiment, sensitivity is used to store the sensitivity coefficient of each input parameter, which is initialized to zero;

[0078] S33. Perform sensitivity analysis on each input parameter, establish a linear regression model to fit the original data, and predict the input variables after disturbance, and calculate the output change. The calculation formula is as follows:

[0079]

[0080] Among them, S i Indicates the sensitivity coefficient of the i-th input parameter, ΔY j Indicates the change in output, ΔX ij Indicates the change in the i-th input parameter, Y j and X ij Represent the original values of output and input respectively;

[0081] S34. Normalize the sensitivity coefficients of all parameters to obtain relative weights. The formula is as follows:

[0082]

[0083] Among them, W i represents the normalized sensitivity weight, ∑S represents the sum of all sensitivity coefficients;

[0084] S35. Use the least squares method to perform linear regression fitting. The formula is as follows:

[0085]

[0086] Among them, β0 represents the intercept term, β i represents the regression coefficient, ∈ represents the error term;

[0087] S36. Repeat the above steps until all samples in the input variables are input.

[0088] In this embodiment, the ship length, ship width, draft, parallel midbody, bilge radius, and arc top radius are considered as input variables for fuzzy optimization.

[0089] In specific implementation, as a preferred embodiment of the present invention, step S4 specifically includes:

[0090] S41, loading the neural network model and performing neural network prediction;

[0091] S42. Create a fuzzy system and define input variables and output variables;

[0092] S43. Define a fuzzy rule base; in this embodiment, such as "L==Medium&B==High&D==Medium&bl==Low&rb==Medium&rd==Low=>Resistance=Low(0.6)";

[0093] S44, perform defuzzification;

[0094] S45. Perform neural network prediction. If there is a large difference with the original data, perform fuzzy optimization.

[0095] In specific implementation, as a preferred embodiment of the present invention, in step S43, a fuzzy rule base is defined, which specifically includes:

[0096] S431. Define the triangular membership of the input variables and the output variables. Set the input variables to three fuzzy ranges: large, medium, and small. Define five ranges for the output resistance.

[0097] S432. Set the permutations and combinations of the six variables to obtain their respective resistance values, and assign corresponding rule weights.

[0098] In specific implementation, as a preferred embodiment of the present invention, in step S44, the centroid method is used as a key step of defuzzification, which specifically includes:

[0099] Calculate the center of gravity of the fuzzy output, that is, the weighted average of the output membership function. The specific formula is as follows:

[0100]

[0101] Among them, y represents the value of the output variable, μ(y) represents the output membership function, for each y value, it represents its membership (that is, the degree to which the value belongs to a certain fuzzy set), a and b both represent the range of output values, y defuzz Represents the clear value after defuzzification, that is, the final output value. Figure 5 As shown, it is a general comparison diagram before and after the optimization of the approximate model of the present invention.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the approximate model of resistance prediction for submersible and floating unmanned vessels considering fuzziness, characterized in that: include: S1. Taking submersible unmanned vessels as the target ship type, a multi-source data set was generated through ship model tests, numerical simulation tests, and cloud computing technology. S2. Based on the feedforward neural network model, establish an approximate model, including the establishment of the input layer, hidden layer and output layer; S3. Compare the approximate model with the original data to find the data with the largest difference from the original data, and find the six parameters with the largest impact on the resistance through sensitivity analysis; S4. Establish the triangular membership of the six parameters and define its rule base, and finally obtain the final resistance value based on defuzzification.

2. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1 is characterized in that: In step S1, the ship model test and numerical simulation test are performed with the submersible unmanned ship as the target ship type, only the resistance characteristics of the main hull of the ship are considered, and the influence of the appendages on the resistance is ignored to obtain experimental data and simulation data.

3. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1, characterized in that: In step S1, the cloud computing technology uses a reverse cloud generator to calculate six digital features, uses a forward cloud generator to expand data, sets the expansion multiple to 100 times, and obtains insufficient information supplement on the speed, diving depth, and resistance of the submersible unmanned ship.

4. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1, characterized in that: In step S2, the feedforward neural network is used as the regression model. In order to quantify the fitting performance of the neural network, the mean square error is used as the evaluation index. The calculation formula of the mean square error is: Among them, MSE represents the number of samples, y i Indicates the actual value, Represents the predicted value.

5. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1 is characterized in that: In step S2, combined with the uncertainty of the mean square error, each data point in the simulation data set and the experimental data set is repeatedly processed to increase the frequency of occurrence of samples.

6. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1, characterized in that: Step S3 specifically includes: S31, import input variables and output variables; S32. Calculate the disturbance of each input variable using the following formula: ΔX i =0.1×(max(X i )-min(X i )) Where ΔX i represents the disturbance amplitude of each input variable, X i represents the i-th input variable; S33. Perform sensitivity analysis on each input parameter, establish a linear regression model to fit the original data, and predict the input variables after disturbance, and calculate the output change. The calculation formula is as follows: Among them, S i Indicates the sensitivity coefficient of the i-th input parameter, ΔY j Indicates the change in output, ΔX ij Indicates the change in the i-th input parameter, Y j and X ij Represent the original values of output and input respectively; S34. Normalize the sensitivity coefficients of all parameters to obtain relative weights. The formula is as follows: Among them, W i represents the normalized sensitivity weight, ∑S represents the sum of all sensitivity coefficients; S35. Use the least squares method to perform linear regression fitting. The formula is as follows: Among them, β0 represents the intercept term, β i represents the regression coefficient, ∈ represents the error term; S36. Repeat the above steps until all samples in the input variables are input.

7. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1, characterized in that: Step S4 specifically includes: S41, loading the neural network model and performing neural network prediction; S42. Create a fuzzy system and define input variables and output variables; S43. Define a fuzzy rule base; S44, perform defuzzification; S45. Perform neural network prediction. If there is a large difference with the original data, perform fuzzy optimization.

8. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1, characterized in that: In step S43, a fuzzy rule base is defined, specifically including: S431. Define the triangular membership of the input variables and the output variables. Set the input variables to three fuzzy ranges: large, medium, and small. Define five ranges for the output resistance. S432. Set the permutations and combinations of the six variables to obtain their respective resistance values, and assign corresponding rule weights.

9. The method for optimizing the resistance prediction approximate model of a submersible unmanned vessel considering fuzziness according to claim 1, characterized in that: In step S44, the centroid method is used as a key step of defuzzification, which specifically includes: Calculate the center of gravity of the fuzzy output, that is, the weighted average of the output membership function. The specific formula is as follows: Among them, y represents the value of the output variable, μ(y) represents the output membership function, for each y value, represents its membership, a and b both represent the output value range, y defuzz Represents the clear value after defuzzification, that is, the final output value.

Images