Prediction of unpowered diving and floating for large-depth manned submersible based on geometric similarity Authors: Zhonghui Hu, Shuai Liu, Wenxin Qu, Lei Jiang, Cong Ye Published: 2026 ABSTRACT Large-depth manned submersibles generally conduct diving and floating operations in an unpowered mode. On the premise of guaranteeing satisfactory diving and floating speed, this mode can significantly reduce energy consumption and prolong underwater operating time, which serves as one of the key foundations for the underwater operational performance of manned submersibles. Therefore, the study on prediction methods for unpowered diving and floating motion is of great engineering significance. At present, existing prediction methods are mainly established on the standard motion model, whose results are extremely sensitive to the handling of hydrodynamic terms in the model, thus limiting the prediction accuracy. In this paper, a prediction method for unpowered diving and floating motion of manned submersibles is proposed based on geometric similarity theory, by deriving the relationship between the submersible’s diving and floating speed and its net weight in water. Taking the “Jiao Long” manned submersible as a research object, the established model is employed to predict the 7000 m unpowered diving and floating motion using the 5000 m deep-sea trial data. Comparative analysis indicates that the proposed method achieves higher prediction accuracy than the conventional standard motion model, which preliminarily verifies its feasibility and effectiveness and its general applicability remains to be further examined in follow-up investigations. Meanwhile, many assumptions are adopted in the development of the proposed method. Therefore, this method is only applicable to the quasi-steady segments of the same manned submersible, and is not suitable for transient phases such as water entry, ballast release transitions, or motions involving significant attitude coupling. Furthermore, the method is only intended for operational application phases, rather than the demonstration and design phases of manned submersibles. Introduction Manned submersibles are specialized deep-sea submersible systems capable of transporting scientific and engineering personnel, together with various specialized equipment, to conduct rapid, accurate and sustainable round-trip voyages to multiple seabed locations for high-efficiency detection, sampling and operational missions1. It fully leverages the subjective initiative and creativity of on-site professionals, serving as one of the critical pieces of equipment for marine access, exploration, development, and conservation2. During the 1980 s, multiple manned submersibles, such as the American “Alvin”3, the Russian “MIR-1” and “MIR-2”4,5, the Japanese “Shinkai 6500”6, and the French “Nautile”7, were commissioned. With operating depths mainly concentrated between 2,000 m and 6,500 m, they exerted a significant influence on deep-sea applications such as scientific research, environmental investigation, resource prospecting, and salvage operations. The American “Alvin” conducted detailed investigations on the seabed in the Galápagos Rift region, confirmed the existence of deep-sea hydrothermal vents. The Russian Mir submersible carried Russian and American scientists and jointly discovered the largest known hydrothermal ore deposit to date in the Atlantic Ocean, which was named after the Mir submersible. Japanese scientists aboard the “Shinkai 6500” identified faults on the slope of the Japan Trench at a depth of 6200 m and carried out comprehensive geological surveys. France and other partners implemented the large international “KM3NeT” project, in which the “Nautile” was used to maintain the detector array of the neutrino telescope undersea observatory, advancing frontier interdisciplinary research such as particle astrophysics. Since the 21 st century, the development of manned submersibles, especially full-ocean-depth manned submersibles with an operating depth of 11,000 m, has sparked a new round of scientific and technological competition. In 2012, the manned submersible “Deepsea Challenger”, developed by an American team led by James Cameron, descended to a depth of 10,898 m in the Mariana Trench, establishing a new depth record for solo dives8. On November 10, 2020, China’s indigenously developed full-ocean-depth manned submersible “Fen Dou Zhe” successfully descended to a depth of 10,909 m in the Mariana Trench, establishing a new record for China’s manned deep-sea diving operations9. Diving and floating are essential stages in the operation of large-depth manned submersibles, with unpowered diving and floating motion as the primary method. Diving and floating ballast consists of load counterweights arranged on both sides of the submersible and serves as a critical component for enabling the manned submersible to perform unpowered diving and floating motion at a predetermined speed. Specifically, the diving ballast is employed to achieve the unpowered diving motion of the manned submersible, and the floating ballast to realize its unpowered floating motion. The detailed procedure is as follows. Prior to the commencement of operational missions, unpowered diving and floating motion prediction shall be performed to verify whether the weight of the onboard diving and floating ballast is sufficient to enable the manned submersible to achieve the intended diving and floating speed. On the deck of the mother ship, the validated ballast is installed at the mounting interface of the ballast release device. Deploy the manned submersible into the sea. The manned submersible began its dive from the water surface, entering the unpowered diving phase. Upon reaching the target depth, the operator manipulates the ballast release device via the control system to jettison the ballast, bringing the manned submersible into a force-balanced state. When the manned submersible has completed underwater operations, the operator uses the control system to actuate the ballast release device and jettison the ballast, allowing the submersible to enter the unpowered floating phase. Under energy-constrained conditions, this locomotion mode can significantly reduce the energy consumption of the manned submersible, reserving more energy for effective operations and ensuring sufficient operational endurance10–12. Prediction of unpowered diving and floating motion constitutes an essential approach for realizing the accurate configuration of diving and floating ballast, and directly affects the control of motion velocity. Regarding the investigation of prediction methodologies, Liu Zhengyuan et al. developed a mathematical model for diving and floating motion of submersibles grounded on the standard motion model, thereby realizing the prediction of unpowered diving and floating maneuvers for manned submersibles13. Building on the foundation laid in literature13, Pan Binbin et al. introduced the concept of variable volume shrinkage ratio to further refine the kinematic mathematical model, and developed an unpowered diving and floating motion analysis system for manned submersibles14. All the aforementioned methods are formulated on the basis of the standard motion model. The hydrodynamic terms of the motion model are predominantly characterized by polynomial expressions or derived from the results of hydrodynamic scaled model tests. Due to the large angle of attack and strong nonlinearity inherent in the unpowered diving and floating motion of manned submersibles, the conventional polynomial expressions fall outside the valid range of polynomial expansion. This leads to increased computational errors in the hydrodynamic terms, resulting in significant deviations in the prediction of the submersible’s unpowered diving and floating motion. If the hydrodynamic terms are represented based on hydrodynamic model test results, the scale effects between the full-scale submersible and the scaled model will impose stringent requirements on the parameter configuration of model tests and the processing of experimental data. To reduce the adverse impact of hydrodynamic term processing strategies in the standard motion model on the prediction accuracy of unpowered diving and floating motion, this study establishes a novel prediction method for the unpowered diving and floating motion of manned submersibles. Based on the force analysis of manned submersibles, the geometric similarity theory is introduced to derive the quantitative relationship between diving and floating velocity and the underwater gravity of the submersible. In this paper, the “Jiao Long” manned submersible is taken as the research object. Based on the 5000 m sea trial data, a prediction model for the unpowered diving and floating motion of the “Jiao Long” manned submersible is obtained through training. Subsequently, the established unpowered diving and floating motion prediction method is used to predict the 7000 m unpowered diving and floating motion of the “Jiao Long” manned submersible. The feasibility and effectiveness of the unpowered diving and floating motion prediction method based on geometric similarity theory are validated through the comparison of prediction results. This provides a possible technical means for the accurate prediction of unpowered diving and floating motions in practical applications. Prediction modeling of unpowered diving and floating motion based on geometric similarity Hypothesis The research work is carried out in the context of engineering applications. The hypothesis for simplified calculation is given in literature15, as follows: During the unpowered diving and floating motion of the submersible, only the variation of the velocity in the vertical direction is considered.The unpowered diving and floating motion of a manned submersible is approximated as constant velocity motion, whereby the acceleration of the unpowered diving and floating motion is approximated to be zero.In the calculation process, only the smooth dive phase and the smooth float phase of the manned submersible are considered. During the unpowered diving and floating motion of the submersible, only the variation of the velocity in the vertical direction is considered. During the unpowered diving and floating motion of the submersible, only the variation of the velocity in the vertical direction is considered. The unpowered diving and floating motion of a manned submersible is approximated as constant velocity motion, whereby the acceleration of the unpowered diving and floating motion is approximated to be zero. The unpowered diving and floating motion of a manned submersible is approximated as constant velocity motion, whereby the acceleration of the unpowered diving and floating motion is approximated to be zero. In the calculation process, only the smooth dive phase and the smooth float phase of the manned submersible are considered. In the calculation process, only the smooth dive phase and the smooth float phase of the manned submersible are considered. Force analysis The authors have comprehensively analyzed the forces acting on the manned submersible during unpowered diving and floating motion, together with the corresponding calculation methods, in literature16. For brevity, these details are not repeated here, and the results are presented directly as follows. The forces acting on the manned submersible during unpowered diving and floating motion are illustrated in Fig. 1. Fig. 1Force analysis. Force analysis. As illustrated in Fig. 1, the forces exerted on the manned submersible during unpowered diving and floating motions comprise gravity, buoyancy, and hydrodynamic forces. The relationships among these forces are as follows: 1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\vec{G} + \vec{B} + \vec{F}_{d} = M\vec{a}$$\end{document} In the equation, M denotes the weight of the manned submersible, and a represents its acceleration during unpowered diving and floating motion. In light of the assumptions in Sect. 2.1, Eq. (1) can be further simplified as: 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} \Delta \vec{G} + \vec{F}_{d} = 0 \hfill \\ \Delta \vec{G} = \vec{G} + \vec{B} \hfill \\ \end{gathered} \right.$$\end{document} In the equation, ΔG represents the underwater gravity of the manned submersible. Gravity G The weight M of the manned submersible is composed of the fixed weight of the submersible, the weight of pilots, the weight of consumables inside the manned tank, the weight of tools, the weight of diving and floating ballast, and the weight of other temporarily loaded equipment and payloads. The gravitational force G on the manned submersible is expressed as: 3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G = \left( {M_{0} + M_{p} + M_{c} + M_{t} + M_{d} + M_{f} + M_{{other}} } \right) \cdot g$$\end{document} In the equation, M0 denotes the fixed weight of the submersible, Mp the weight of pilots, Mc the weight of consumables inside the manned tank, Mt the weight of tools, Md the weight of diving ballast, Mf the weight of floating ballast, and Mother the weight of other temporarily mounted payloads. Buoyancy B During unpowered diving and floating, seawater density and hydrostatic pressure vary with depth. Changes in hydrostatic pressure induce variations in the displacement volume of the manned submersible. Consequently, the buoyancy force B acting on the manned submersible is expressed as: 4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} B = \rho g\left( {1 - k} \right)V_{0} \hfill \\ k = k_{1} + k_{2} + k_{3} \hfill \\ \end{gathered} \right.$$\end{document} In the equation, ρ denotes the density of seawater. g denotes the gravitational acceleration. V0 denotes the displaced volume when the manned submersible is completely submerged at the start of the dive at the sea surface. k denotes the coefficient of displaced volume loss, which is defined as the ratio of the displaced volume lost by the submersible to V0. k1 denotes the displaced volume loss coefficient of the pressure-resistant structure, which is defined as the ratio of the displaced volume lost by the pressure-resistant structure of submersible to V0. k2 denotes the displaced volume loss coefficient of the buoyant material, which is defined as the ratio of the displaced volume lost by the buoyant material of submersible to V0. k3 denotes the displaced volume loss coefficient of the pressure-compensating component, which is defined as the ratio of the displaced volume lost by the pressure-compensating component of submersible to V0. As documented in the literature17, the density of seawater at depths calculated via the EOS80 seawater equation of state. The coefficient of displaced volume loss for pressure-resistant structures The displaced volume loss coefficient of the pressure-resistant structure k1 is expressed as: 5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k_{1} = k_{{11}} + k_{{12}}$$\end{document} In the equation, k11 denotes the displaced volume loss coefficient of the spherical pressure-resistant structure, k12 the displaced volume loss coefficient of the cylindrical pressure-resistant structure. The displaced volume loss coefficient of the spherical pressure-resistant structure k11 is expressed as: 6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} k_{{11}} = \frac{{\sum {\frac{{4\pi }}{3}\left[ {(r_{{Si}} + t_{{Si}} + \Delta r_{{Si1}} )^{3} - (r_{{Si}} + t_{{Si}} + \Delta r_{{Si0}} )^{3} } \right]} }}{{V_{0} }} \hfill \\ \Delta r_{{Si0}} = \frac{{P_{{Si}} \cdot (r_{{Si}} + t_{{Si}} /2)^{2} }}{{2E_{{Si}} \cdot t_{i} }} \cdot (1 - \mu _{{Si}} ) \hfill \\ \Delta r_{{Si1}} = \frac{{(P_{{Si}} - P) \cdot r_{{Si}}^{2} }}{{2E_{{Si}} \cdot t_{{Si}} }} \cdot (1 - \mu _{{Si}} ) \hfill \\ P = \rho \cdot g \cdot h \hfill \\ \end{gathered} \right.$$\end{document} In the equation, rS denotes the inner radius of the spherical shell, tS the thickness, ΔrS0 the radial deformation at the onset of diving, ΔrS1 the radial deformation at depth h, ES the Young’s modulus of the spherical shell material, µS the Poisson’s ratio of the spherical shell material, PS the internal pressure acting on the spherical pressure-resistant structure, P the seawater pressure at depth h, and h the operating depth of the manned submersible. The displaced volume loss coefficient of the cylindrical pressure-resistant structure k12 is expressed as:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} {k_{21}}=\frac{{\sum {\pi \left[ {{{({r_{Ci1}}+{t_{Ci}}+\Delta {r_{Ci1}})}^2}({l_{Ci}}+\Delta {l_{Ci1}}) - {{({r_{Ci1}}+{t_{Ci}}+\Delta {r_{Ci0}})}^2}({l_{Ci}}+\Delta {l_{Ci0}})} \right]} }}{{{V_0}}} \hfill \\ \Delta {r_{Ci0}}=\frac{{2 - {\mu _{Ci}}}}{{{E_{Ci}}}} \times \frac{{{P_{Ci}} \cdot r_{{Ci1}}^{2} \cdot {r_{Ci2}}}}{{r_{{Ci2}}^{2} - r_{{Ci1}}^{2}}} \hfill \\ \Delta {l_{Ci0}}=\frac{{1 - 2{\mu _{Ci}}}}{{{E_{Ci}}}} \times \frac{{{P_{Ci}} \cdot r_{{Ci1}}^{2} \cdot {l_{Ci}}}}{{r_{{Ci2}}^{2} - r_{{Ci1}}^{2}}} \hfill \\ \Delta {r_{Ci1}}=\frac{{1 - 2{\mu _{Ci}}}}{{{E_{Ci}}}} \times \frac{{{P_{Ci}} \cdot r_{{Ci1}}^{2} - P \cdot r_{{Ci2}}^{2}}}{{r_{{Ci2}}^{2} - r_{{Ci1}}^{2}}} \cdot {r_{Ci2}}+\frac{{1+{\mu _{Ci}}}}{{{E_{Ci}}}} \times \frac{{{P_{Ci}} - P}}{{r_{{Ci2}}^{2} - r_{{Ci1}}^{2}}}r_{{Ci1}}^{2} \cdot {r_{Ci2}} \hfill \\ \Delta {l_{Ci1}}=\frac{{1 - 2{\mu _{Ci}}}}{{{E_{Ci}}}} \times \frac{{{P_{Ci1}} \cdot r_{{Ci1}}^{2} - P \cdot r_{{Ci2}}^{2}}}{{r_{{Ci2}}^{2} - r_{{Ci1}}^{2}}} \cdot {l_{Ci}} \hfill \\ P=\rho \cdot g \cdot h \hfill \\ \end{gathered} \right.$$\end{document} In the equation, rC1 denotes the inner radius of the cylindrical shell, rC2 the outer radius, lC the axial length, tC the thickness, ΔrC0 the radial deformation at the onset of diving, ΔlC0 the axial deformation at the onset of diving, ΔrC1 the radial deformation at diving depth h, ΔlC1 the axial deformation at depth h, EC the Young’s modulus of the cylindrical shell material, µC the Poisson’s ratio of the cylindrical shell material, and PC the internal pressure of the cylindrical pressure-resistant structure. The coefficient of displaced volume loss for buoyant materials The displaced volume loss coefficient of the buoyant material k2 is expressed as: 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k_{2} = k_{{21}} + k_{{22}}$$\end{document} In this context, k21 represents the displaced volume loss coefficient caused by the absorption of water of the buoyant material, whereas k22 represents the displaced volume loss coefficient caused by the volume contraction of the buoyant material. The volume loss of buoyant material due to water absorption can be expressed as the coefficient of buoyancy loss, designated k21, which can be expressed as follows: 9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k_{{21}} = \frac{{\rho _{{bm}} \cdot V_{{bm0}} \cdot \eta _{{bm}} }}{{\rho \cdot V_{0} }}$$\end{document} In this context, ρbm denotes the density of the buoyant material, and Vbm0 denotes the initial displaced volume of the buoyant material, ηbm represents the water absorption fraction of the buoyant material. The volume loss of buoyant material due to volume shrinkage can be expressed as the buoyancy loss coefficient, designated k22, which can be expressed as follows: 10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} k_{{22}} = - \frac{P}{{K_{{bm}} }} \cdot \frac{{V_{{bm0}} }}{{V_{0} }} \hfill \\ P = \rho \cdot g \cdot h \hfill \\ K_{{bm}} = \frac{{E_{{bm}} }}{{3\left( {1 - 2\mu _{{bm}} } \right)}} \hfill \\ \end{gathered} \right.$$\end{document} In this context, Kbm represents the volumetric elastic modulus, Ebm denotes the Young’s modulus of the buoyant material, µbm the Poisson’s ratio of the buoyant material. The coefficient of displaced volume loss for the pressure-compensating component The displaced volume loss coefficient of the pressure-compensating component is k3 expressed as:11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k_3}={\eta _{hf}} \cdot \frac{{{V_{hf0}}}}{{{V_0}}}$$\end{document} In this context, ηhf represents the compression fraction of a hydraulic fluid, Vhf0 represents the volume of hydraulic fluid at atmospheric pressure. Motion prediction model The hydrodynamic forces acting on a manned submersible can be expressed as follows:12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F_d}=\frac{1}{2}\rho {v^2}S \cdot {C_Z}$$\end{document} In this context, ρ represents the density of seawater, S denotes the wetted surface area of the deep-sea manned submersible, v denotes the velocity in the dive and float directions, and Cz denotes the drag coefficients in the diving and floating directions. During the unpowered diving and floating motion of the manned submersible, its velocity changes slightly, resulting in a relatively small variation of the Reynolds number. Accordingly, the drag coefficients in the diving and floating directions Cz is assumed to be constant in this study. Based on the geometric similarity theory presented in literature18, the relationships between the wetted surface area S, displaced volume V, and characteristic length l of the manned submersible are respectively given by: 13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S \propto l^{2}$$\end{document} 14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V \propto l^{3}$$\end{document} According to Eqs. (13) and (14), the relationship between the wetted surface area S and displaced volume V of the manned submersible is given by:15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S \propto {l^2} \propto {\left( {{V^{1/3}}} \right)^2} \propto {V^{2/3}}$$\end{document} It is assumed that the variation in seawater density and the change in the displacement volume of the manned submersible with depth are small quantities. Meanwhile, for the same manned submersible, the weight variation is approximately equal to the equivalent displaced mass. According to the scaling law in geometric similarity and the above assumptions, the following relation holds:16\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta G \propto V \propto {l^3}$$\end{document} The relationship between the wetted surface area S of the manned submersible and its underwater gravity ΔG can be obtained as:17\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S \propto {l^2} \propto {\left( {\Delta {G^{1/3}}} \right)^2} \propto \Delta {G^{2/3}}$$\end{document} Substituting Eq. (12) into Eq. (2) yields:18\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta G=\frac{1}{2}\rho {v^2}S \cdot {C_Z}$$\end{document} Substituting Eq. (17) into Eq. (18) yields:19\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta G=\frac{1}{2}\rho {v^2}S \cdot {C_Z} \propto \frac{1}{2}\rho {v^2} \cdot {C_Z} \cdot \Delta {G^{2/3}} \Rightarrow {v^2} \propto \frac{2}{{\rho \cdot {C_Z}}} \cdot \Delta {G^{1/3}}$$\end{document} Together with the aforementioned assumptions that the change in seawater density ρ is a small quantity and the drag coefficients in the diving and floating directions Cz remains constant, we obtain:20\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${v^2} \propto \Delta {G^{1/3}}$$\end{document} Taking the square root of both sides gives:21\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$v \propto \Delta {G^{1/6}}$$\end{document} According to Eq. (21), for the same manned submersible, the diving and floating velocity v of the manned submersible is proportional to the 1/6 power of its underwater gravity ΔG. Based on this, the prediction model for the diving and floating velocity v can be established as: 22\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} \vec{v} = b \cdot \Delta \vec{G}^{{1/6}} + c \hfill \\ \Delta \vec{G} = \vec{G} + \vec{B} \hfill \\ G = \left( {M_{0} + M_{p} + M_{c} + M_{t} + M_{{ad}} + M_{{other}} } \right) \cdot g \hfill \\ B = \rho \cdot g \cdot \left( {1 - k} \right) \cdot V_{0} \hfill \\ \end{gathered} \right.$$\end{document} In this equation, the gravitational force G acting on the manned submersible is calculated using the method presented in Sect. 2.2.1. The buoyant force B exerted on the manned submersible is determined by the approach described in Sect. 2.2.2. Parameters b and c are the coefficients of the prediction model, which are identified from the training dataset. It can be seen from Eq. (22) that: Parameter b can be interpreted as the velocity gain coefficient, which reflects the sensitivity of the velocity in the diving and floating directions to the net underwater gravity and represents the response intensity of diving and floating speed to net gravity. It is jointly determined by the drag coefficients in the diving and floating directions Cz, seawater density ρ, and wetted surface area S. A larger b indicates a more significant influence of net gravity on the diving and floating velocity.Parameter c can be regarded as the velocity correction term. Since the proposed method is based on the smooth phase of diving and floating motion, an initial velocity value exists accordingly. Parameter b can be interpreted as the velocity gain coefficient, which reflects the sensitivity of the velocity in the diving and floating directions to the net underwater gravity and represents the response intensity of diving and floating speed to net gravity. It is jointly determined by the drag coefficients in the diving and floating directions Cz, seawater density ρ, and wetted surface area S. A larger b indicates a more significant influence of net gravity on the diving and floating velocity. Parameter b can be interpreted as the velocity gain coefficient, which reflects the sensitivity of the velocity in the diving and floating directions to the net underwater gravity and represents the response intensity of diving and floating speed to net gravity. It is jointly determined by the drag coefficients in the diving and floating directions Cz, seawater density ρ, and wetted surface area S. A larger b indicates a more significant influence of net gravity on the diving and floating velocity. Parameter c can be regarded as the velocity correction term. Since the proposed method is based on the smooth phase of diving and floating motion, an initial velocity value exists accordingly. Parameter c can be regarded as the velocity correction term. Since the proposed method is based on the smooth phase of diving and floating motion, an initial velocity value exists accordingly. Parameter identification method for motion prediction model In this paper, the least squares method is adopted to identify the parameters b and c in Eq. (22). Let: 23\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} x = \Delta \vec{G}^{{1/6}} \hfill \\ y = \vec{v} \hfill \\ \end{gathered} \right.$$\end{document} Substituting Eq. (23) into Eq. (22) yields: 24y = b · x + c Let the number of training samples be n, then for the given n samples \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ {\left( {{x_1},{y_1}} \right),\left( {{x_2},{y_2}} \right),...,\left( {{x_i},{y_i}} \right)} \right\}$$\end{document},i=1,2, … ,n, the sum of squared errors R can be expressed as: 25\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R = \sum\limits_{{i = 1}}^{n} {\left( {y_{i} - b \cdot x_{i} - c} \right)^{2} }$$\end{document} Based on the least squares criterion, it follows that: 26\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} \sum\limits_{{i = 1}}^{n} {\left( {y_{i} - b \cdot x_{i} - c} \right)x_{i} } = 0 \hfill \\ \sum\limits_{{i = 1}}^{n} {\left( {y_{i} - b \cdot x_{i} - c} \right)} = 0 \hfill \\ \end{gathered} \right.$$\end{document} Equation (24) yields: 27\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{ \begin{gathered} b = \frac{{n\sum {x_{i} y_{i} } - \sum {x_{i} } \sum {y_{i} } }}{{n\sum {x_{i}^{2} - \left( {\sum {x_{i} } } \right)^{2} } }} \hfill \\ c = \frac{{\sum {x_{i}^{2} \sum {y_{i} } } - \sum {x_{i} y_{i} } \sum {x_{i} } }}{{n\sum {x_{i}^{2} - \left( {\sum {x_{i} } } \right)^{2} } }} \hfill \\ \end{gathered} \right.$$\end{document} Prediction method for unpowered diving and floating motion The prediction method for the unpowered diving and floating motion of the manned submersible is illustrated in Fig. 2. First, the parameters b and c in the prediction model are identified using Eq. (27) based on the training dataset constructed from historical data, thereby establishing the unpowered diving and floating motion prediction model. Then, the diving and floating motion is calculated according to the parameters of the manned submersible and the desired diving/floating velocity. The calculation includes the diving phase and the floating phase, both of which use the same algorithm. The calculation procedure is illustrated below taking the diving phase as an example. The calculation is performed iteratively with depth as the iterative variable. At the start of the calculation, the initial depth is set to zero, which is implemented by the depth increment module. Based on the submersible parameters and the current depth, the gravitational force is computed by the gravity calculation module, and the buoyant force is obtained by the buoyancy calculation module. The calculated gravitational and buoyant forces are then input into the established unpowered diving and floating motion prediction model to obtain the velocity at the current depth. Afterwards, the depth is increased by a predefined step, and the above procedure is repeated to obtain velocities at different depths until the target depth is reached. Fig. 2Overall calculation flow. Overall calculation flow. The methodology for the gravity calculation module in Fig. 2 is shown in Fig. 3. Using the weight parameters of the manned submersible, the gravitational force is evaluated via Eq. (3). Fig. 3Gravity calculation module. Gravity calculation module. Key parameters for this calculation are summarized in Table 1. Table 1The parameters necessary to calculate the gravity force.ParametersParameter usageThe fixed weight of the submersible M0These parameters are utilised for the purpose of calculating the variable gravity G of the submersible.The weight of pilots MpThe weight of consumables in the manned tank McThe weight of tools MtThe weight of the diving and floating ballast MdThe weight of the diving and floating ballast MfOther weights Mother Table 1The parameters necessary to calculate the gravity force.ParametersParameter usageThe fixed weight of the submersible M0These parameters are utilised for the purpose of calculating the variable gravity G of the submersible.The weight of pilots MpThe weight of consumables in the manned tank McThe weight of tools MtThe weight of the diving and floating ballast MdThe weight of the diving and floating ballast MfOther weights Mother The parameters necessary to calculate the gravity force. The buoyancy calculation module in Fig. 2 is detailed in Fig. 4. Specifically, the volume loss coefficients are first derived using Eqs. (4)–(11), based on the properties of the spherical and cylindrical pressure-resistant structures, buoyancy material, and hydraulic oil. Subsequently, the seawater density at the current operating depth is calculated using the EOS80 equation of state. The buoyant force is then determined according to Eq. (4). Fig. 4Buoyancy calculation module. Buoyancy calculation module. Parameters required for buoyancy estimation are presented in Table 2. Table 2The parameters necessary to calculate the buoyancy force.ParametersParameter usageThe displaced volume when the manned submersible is completely submerged at the start of the dive at the sea surface V0This parameter is utilised for the purpose of calculating the displaced volume loss coefficient k.The current depth of the submersible hThis parameter is utilised for the purpose of calculating process parameters such as seawater pressure and seawater density, derived from the current depth.The primary parameters of the spherical shell pressure-resistant bodiesRadius rSThese parameters are utilised for the purpose of calculating the displaced volume loss coefficient k11 of the spherical shell pressure-resistant bodies.Wall thickness tSYoung’s modulus ESPoisson’s ratio µSThe primary parameters of the cylindrical shell pressure-resistant bodiesInner radius rC1These parameters are utilised for the purpose of calculating the displaced volume loss coefficient k12 of the cylindrical shell pressure-resistant bodies.Outer radius rC2Wall thickness tCLength lCYoung’s modulus ECPoisson’s ratio µCThe primary parameters of the buoyant materialVolume at atmospheric pressure Vbm0These parameters are utilised for the purpose of calculating the displaced volume loss coefficient k21 caused by the absorption of water of the buoyant material and the displaced volume loss coefficient k22 caused by the volume contraction of the buoyant material.Density ρbmWater absorption fraction ηbmThese parameters are utilised for the purpose of calculating the displaced volume loss coefficient k21 caused by the absorption of water of the buoyant material.Young’s modulus EbmThese parameters are utilised for the purpose of calculating the displaced volume loss coefficient k22 caused by the volume contraction of the buoyant material.Poisson’s ratio µbmThe primary parameters of the pressure-compensating componentVolume of hydraulic fluid at atmospheric pressure Vhf0These parameters are utilised for the purpose of calculating the displaced volume loss coefficient of the pressure-compensating component k3.Compression fraction of hydraulic fluid ηhf Table 2The parameters necessary to calculate the buoyancy force.ParametersParameter usageThe displaced volume when the manned submersible is completely submerged at the start of the dive at the sea surface V0This parameter is utilised for the purpose of calculating the displaced volume loss coefficient k.The current depth of the submersible hThis parameter is utilised for the purpose of calculating process parameters such as seawater pressure and seawater density, derived from the current depth.The primary parameters of the spherical shell pressure-resistant bodiesRadius rSThese parameters are utilised for the purpose of calculating the displaced volume loss coefficient k11 of the spherical shell pressure-resistant bodies.Wall thickness tSYoung’s modulus ESPoisson’s ratio µSThe primary parameters of the cylindrical shell pressure-resistant bodiesInner radius rC1These parameters are utilised for the purpose of calculating the displaced volume loss coefficient k12 of the cylindrical shell pressure-resistant bodies.Outer radius rC2Wall thickness tCLength lCYoung’s modulus ECPoisson’s ratio µCThe primary parameters of the buoyant materialVolume at atmospheric pressure Vbm0These parameters are utilised for the purpose of calculating the displaced volume loss coefficient k21 caused by the absorption of water of the buoyant material and the displaced volume loss coefficient k22 caused by the volume contraction of the buoyant material.Density ρbmWater absorption fraction ηbmThese parameters are utilised for the purpose of calculating the displaced volume loss coefficient k21 caused by the absorption of water of the buoyant material.Young’s modulus EbmThese parameters are utilised for the purpose of calculating the displaced volume loss coefficient k22 caused by the volume contraction of the buoyant material.Poisson’s ratio µbmThe primary parameters of the pressure-compensating componentVolume of hydraulic fluid at atmospheric pressure Vhf0These parameters are utilised for the purpose of calculating the displaced volume loss coefficient of the pressure-compensating component k3.Compression fraction of hydraulic fluid ηhf The parameters necessary to calculate the buoyancy force. Computational validation In this paper, the “Jiao Long” manned submersible is taken as the research object. The unpowered diving and floating motion is predicted using the prediction model proposed in this study, and the predicted results are compared with the sea trial data and the prediction results from literature16, which is based on the standard motion model. Specifically, the 5000 m diving test data of the D44 dive of the “Jiao Long” submersible are utilized as the training samples, whereas the 7000 m diving test data of the D55 dive and the results from literature16 are employed as the validation samples. Main parameters of the “Jiao Long” manned submersible The key parameters of the “Jiao Long” manned submersible provided by literature15 are summarized in Table 3. Table 3The primary parameters of the “Jiao Long” manned submersible.Parameter termValueThe fixed weight M020,414 kgThe fixed displaced volume V020.8047 m3Displaced volume loss coefficient of the pressure-resistant structure k13.65 × 10− 7hDisplaced volume loss coefficient of the buoyant material k22.23 × 10− 6hDisplaced volume loss coefficient of the pressure-compensating component k32.67 × 10− 7h Table 3The primary parameters of the “Jiao Long” manned submersible.Parameter termValueThe fixed weight M020,414 kgThe fixed displaced volume V020.8047 m3Displaced volume loss coefficient of the pressure-resistant structure k13.65 × 10− 7hDisplaced volume loss coefficient of the buoyant material k22.23 × 10− 6hDisplaced volume loss coefficient of the pressure-compensating component k32.67 × 10− 7h The primary parameters of the “Jiao Long” manned submersible. Identification of prediction model parameters In this paper, the parameters of the prediction model are identified based on the 5000 m diving test data of the D44 dive of the “Jiao Long” manned submersible presented in literature16. During the 5000 m diving test of the D44 dive, the variation of the in-water gravity of the “Jiao Long” submersible with depth during the unpowered diving and floating motion is illustrated in Fig. 5. Fig. 5△G curve of “Jiao Long” with depth during 5000 m diving test. △G curve of “Jiao Long” with depth during 5000 m diving test. The parameters in Eq. (22) are identified using Eq. (27), and the identification results are listed in Table 4. Table 4Prediction model parameter identification results.Motion phaseParameter termValueDiving b 0.3171 c −0.6210Floating b 0.3842 c −0.9442 Table 4Prediction model parameter identification results.Motion phaseParameter termValueDiving b 0.3171 c −0.6210Floating b 0.3842 c −0.9442 Prediction model parameter identification results. Substituting the identified parameters into Eq. (22) yields the unpowered diving and floating motion prediction model of the “Jiao Long” manned submersible as follows: 28\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$v = \left\{ \begin{gathered} 0.3171 \cdot \left( {G - B} \right)^{{1/6}} - 0.621,diving \hfill \\ 0.3842 \cdot \left( {G - B} \right)^{{1/6}} - 0.9442,floating \hfill \\ \end{gathered} \right.$$\end{document} 7000 m unpowered diving and floating motion prediction The 7000 m unpowered diving and floating motion of the “Jiao Long” manned submersible is predicted using Eq. (27). Payload conditions for the submersible during the dive 50 are given in Table 5. Table 5The payload parameters of dive 50.Payload termWeights (kg)Displaced volume (m3)Consumables in the manned tank350Pilots2240Tools350.0588Diving ballast2460.0315Floating ballast6000.0769OtherBallast water1760Lead blocks for equalization2030.0174 Table 5The payload parameters of dive 50.Payload termWeights (kg)Displaced volume (m3)Consumables in the manned tank350Pilots2240Tools350.0588Diving ballast2460.0315Floating ballast6000.0769OtherBallast water1760Lead blocks for equalization2030.0174 The payload parameters of dive 50. The corresponding prediction results are illustrated in Fig. 6. Fig. 6Predicted velocity curves of 7000 m unpowered diving and floating motion. Predicted velocity curves of 7000 m unpowered diving and floating motion. It can be observed from Fig. 6 that the predicted results are highly consistent with the sea trial data in terms of variation trend. Compared with the prediction results in literature16, the proposed method yields better agreement with the experimental measurements than the standard motion model. Error comparison and analysis The error comparison of the 7000 m unpowered diving and floating motion prediction is displayed in Fig. 7. Fig. 7Comparison of error in prediction of diving and floating speed. Comparison of error in prediction of diving and floating speed. The error comparison of the prediction results is summarized in Table 6. The error metrics adopted include root mean square error (RMSE), maximum absolute error, mean absolute error (MAE), maximum relative error, and average relative error. Table 6Comparison of error of prediction between two models.Types of errorPrediction method based on geometric similarityPrediction method based on standard dynamic modelRMSEDiving speed0.73 m/min1.71 m/minFloating speed0.41 m/min0.58 m/minMaximum absolute errorDiving speed2.99 m/min4.21 m/minFloating speed1.41 m/min1.47 m/minMAEDiving speed0.56 m/min1.41 m/minFloating speed0.32 m/min0.48 m/minMaximum relative errorDiving speed10.36%10.88%Floating speed4.89%5.09%Average relative errorDiving speed1.58%3.98%Floating speed0.94%1.38% Table 6Comparison of error of prediction between two models.Types of errorPrediction method based on geometric similarityPrediction method based on standard dynamic modelRMSEDiving speed0.73 m/min1.71 m/minFloating speed0.41 m/min0.58 m/minMaximum absolute errorDiving speed2.99 m/min4.21 m/minFloating speed1.41 m/min1.47 m/minMAEDiving speed0.56 m/min1.41 m/minFloating speed0.32 m/min0.48 m/minMaximum relative errorDiving speed10.36%10.88%Floating speed4.89%5.09%Average relative errorDiving speed1.58%3.98%Floating speed0.94%1.38% Comparison of error of prediction between two models. In accordance with the error comparison results presented in Fig. 7; Table 6, a comparative error analysis is carried out between the prediction method based on geometric similarity and that based on standard dynamic model. As illustrated in Fig. 7, during the diving phase, the prediction error associated with the standard dynamic model rises monotonically with increasing depth, whereas that of the geometric similarity–based method remains considerably steadier with noticeably smaller fluctuations. In the floating phase, the prediction errors of both methods display mild, stable variations with a narrow performance gap, and the geometric similarity–based method still yields a marginal yet consistent advantage. As revealed in Table 6, the RMSE, MAE, and average relative error of the geometric similarity–based prediction method are all lower than those of the standard dynamic model–based method, indicating that its prediction results are generally closer to the measured data. In terms of root-mean-square error, the reduction reaches 57.3% for diving speed and 29.3% for floating speed, demonstrating that the geometric similarity–based method can significantly suppress large errors and achieve superior overall stability. With respect to maximum absolute error and maximum relative error, the maximum absolute error for diving speed is reduced by 28.9%, while that for floating speed remains nearly equivalent with a slight advantage; the maximum relative errors of the two methods are close and both controlled at a low level, implying better control over extreme deviations and higher stability for the proposed method. Regarding mean absolute error and average relative error, the mean absolute error is reduced by 60.2% for diving speed and 33.3% for floating speed; the average relative error is reduced by 60.3% for diving speed and 31.8% for floating speed, confirming that the geometric similarity–based method provides higher overall accuracy and better agreement with actual conditions. The errors of the geometric similarity–based prediction method mainly arise from: The method only considers quasi-steady segments and ignores transient phases such as initial water entry, ballast release transitions, and motions involving significant attitude coupling.When introducing the geometric similarity theory, it is assumed that the variation in seawater density and the change in the displacement volume of the manned submersible with depth are small quantities, and the drag coefficient is approximated as a constant. The method only considers quasi-steady segments and ignores transient phases such as initial water entry, ballast release transitions, and motions involving significant attitude coupling. The method only considers quasi-steady segments and ignores transient phases such as initial water entry, ballast release transitions, and motions involving significant attitude coupling. When introducing the geometric similarity theory, it is assumed that the variation in seawater density and the change in the displacement volume of the manned submersible with depth are small quantities, and the drag coefficient is approximated as a constant. When introducing the geometric similarity theory, it is assumed that the variation in seawater density and the change in the displacement volume of the manned submersible with depth are small quantities, and the drag coefficient is approximated as a constant. Conclusion Based on the force analysis of manned submersibles, this paper introduces the geometric similarity theory to establish a prediction model and develop a corresponding method for the unpowered diving and floating motion of manned submersibles. Taking the “Jiao Long” manned submersible as the research object, the parameters in the prediction model are identified using the least squares method based on 5,000 m sea trial data and the unpowered diving and floating motion at 7000 m is predicted, which preliminarily validates the feasibility and effectiveness of the proposed method. Nevertheless, its general applicability remains to be further examined in follow-up investigations. Based on the aforementioned validation results, a comprehensive comparison is conducted between the proposed method and existing representative approaches from four perspectives: model assumptions, hydrodynamic dependence, model complexity, and applicable scope. The detailed comparison results are presented in Table 7. Table 7Comparison of prediction methods for unpowered diving and floating motion.MethodAssumptionsHydrodynamic dependenceModel complexityApplicable scopeLiterature136-DOF motion polynomial fittingHigh (sensitive to hydrodynamic term processing)HighFull motion phase (including transient)Literature15 Vertical motion Quasi-steady segmentsMedium (depends on model test data)MediumThe smooth dive and float phaseProposed method Vertical motion Quasi-steady segmentsLow (weakened by geometric scaling)LowThe smooth dive and float phase Table 7Comparison of prediction methods for unpowered diving and floating motion.MethodAssumptionsHydrodynamic dependenceModel complexityApplicable scopeLiterature136-DOF motion polynomial fittingHigh (sensitive to hydrodynamic term processing)HighFull motion phase (including transient)Literature15 Vertical motion Quasi-steady segmentsMedium (depends on model test data)MediumThe smooth dive and float phaseProposed method Vertical motion Quasi-steady segmentsLow (weakened by geometric scaling)LowThe smooth dive and float phase Comparison of prediction methods for unpowered diving and floating motion. 6-DOF motion polynomial fitting Vertical motion Quasi-steady segments Vertical motion Quasi-steady segments It can be seen from Table 7 that after introducing the geometric similarity theory: The influence of hydrodynamic forces in the prediction model is weakened, and the dependence of prediction accuracy on the processing methods of hydrodynamic terms is reduced.The complexity of the model is lowered, making the prediction more concise and convenient.The applicable scope of the method is limited. The method is only suitable for quasi-steady segments and not for transient phases such as initial entry, ballast release transitions, or motions involving significant attitude coupling. The influence of hydrodynamic forces in the prediction model is weakened, and the dependence of prediction accuracy on the processing methods of hydrodynamic terms is reduced. The influence of hydrodynamic forces in the prediction model is weakened, and the dependence of prediction accuracy on the processing methods of hydrodynamic terms is reduced. The complexity of the model is lowered, making the prediction more concise and convenient. The complexity of the model is lowered, making the prediction more concise and convenient. The applicable scope of the method is limited. The method is only suitable for quasi-steady segments and not for transient phases such as initial entry, ballast release transitions, or motions involving significant attitude coupling. The applicable scope of the method is limited. The method is only suitable for quasi-steady segments and not for transient phases such as initial entry, ballast release transitions, or motions involving significant attitude coupling. Meanwhile, many assumptions are adopted in the development of the proposed method, and its validation is only preliminary due to the limitation of test samples. Therefore, the method still has the following limitations: The prediction method for unpowered diving and floating motion of manned submersibles based on geometric similarity theory is only suitable for the application phase, rather than the demonstration and design phases of manned submersibles.In the prediction model, the transferability of parameters is limited. Parameters can only be transferred to the same manned submersible but cannot be directly transferred to manned submersibles with different shapes and main dimensions, for which re-identification is required.Although the method demonstrates higher prediction accuracy in the case study, its general applicability remains to be further examined in follow-up investigations. The prediction method for unpowered diving and floating motion of manned submersibles based on geometric similarity theory is only suitable for the application phase, rather than the demonstration and design phases of manned submersibles. The prediction method for unpowered diving and floating motion of manned submersibles based on geometric similarity theory is only suitable for the application phase, rather than the demonstration and design phases of manned submersibles. In the prediction model, the transferability of parameters is limited. Parameters can only be transferred to the same manned submersible but cannot be directly transferred to manned submersibles with different shapes and main dimensions, for which re-identification is required. In the prediction model, the transferability of parameters is limited. Parameters can only be transferred to the same manned submersible but cannot be directly transferred to manned submersibles with different shapes and main dimensions, for which re-identification is required. Although the method demonstrates higher prediction accuracy in the case study, its general applicability remains to be further examined in follow-up investigations. Although the method demonstrates higher prediction accuracy in the case study, its general applicability remains to be further examined in follow-up investigations.