Ship Intelligent Engine Room Signal Acquisition and Analysis System

Abstract

This invention relates to the technical field of ship systems, specifically to a ship intelligent engine room signal acquisition and analysis system, including a signal acquisition module, a reconstruction and fusion module, a safety domain calculation module, a safety margin prediction module, and a graded response module. The signal acquisition module is used to synchronously acquire multi-source time-series signals of the ship's power system; the reconstruction and fusion module constructs a joint state vector; the safety domain calculation module calculates the safety domain volume of the ship's power system at the current moment; the safety margin prediction module outputs the remaining safety margin of the system; and the graded response module is used to execute graded early warning or control actions based on the remaining safety margin of the system. In the deeply coupled control scenario of a ship's intelligent engine room, this invention achieves real-time, quantitative, and forward-looking assessment and proactive safety protection of the stability state of a high-dimensional nonlinear system, laying a crucial safety foundation for intelligent ship navigation.

Description

Technical Field

[0001] This invention relates to the technical field of ship systems, and more particularly to a ship intelligent engine room signal acquisition and analysis system. Background Technology

[0002] With the global shipping industry's increasing demand for energy conservation, emission reduction, and intelligent operation, modern ships are increasingly integrating advanced intelligent control systems. These systems aim to achieve optimal global instantaneous energy efficiency by coordinating the control of multiple power and auxiliary equipment, such as main engines, generators, steering gears, and anti-roll fins. This trend is driving the evolution of traditional monitoring and control strategies based on isolated equipment or local subsystems towards a ship-wide, deeply coupled collaborative optimization paradigm. Current mainstream energy-saving control strategies and related stability monitoring methods face fundamental challenges in dealing with this new and complex system. First, most optimization algorithms rely on linear or quasi-linearized system models, describing equipment dynamics by linearizing them near specific operating points. While these models are effective under stable operating conditions, their accuracy drops sharply when the system deviates significantly from design conditions in pursuit of maximum energy efficiency and strong coupling control is introduced. Second, common frequency domain analysis methods (such as vibration monitoring based on Fourier transform) excel at identifying periodic fault characteristics, but their ability to distinguish and warn of chaotic or low-frequency oscillation modes with wide bandwidth and time-varying characteristics generated by multivariable nonlinear coupling is insufficient. Third, the widely used alarm mechanisms based on threshold values ​​of physical quantities of a single device (such as temperature and pressure limits) are essentially local and reactive protections that cannot detect or assess system-level dynamic instability risks caused by the interaction of multiple subsystems.

[0003] When the main engine, generator, steering gear, anti-roll fins, and other actuators are deeply coupled and controlled by a higher-level intelligent system, the entire ship's propulsion system is essentially driven and operates within a high-dimensional, highly nonlinear dynamic range. Within this range, the energy and information exchange between subsystems is exceptionally close, resulting in complex phase trajectory morphology. At this point, minute disturbances can be amplified through coupling, causing low-frequency oscillations or even chaotic states. Traditional linear stability criteria are difficult to identify in a timely manner, posing safety risks. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes a ship intelligent engine room signal acquisition and analysis system, the specific solution of which is as follows: The ship's intelligent engine room signal acquisition and analysis system includes: The signal acquisition module is used to synchronously acquire multi-source timing signals of the ship's power system, including main engine vibration signals, generator electrical signals, steering gear and anti-roll fin action signals, and hull motion signals. The reconstruction and fusion module is used to perform phase space reconstruction based on the multi-source time-series signals to generate multiple corresponding reconstructed state vectors; and to merge the multiple reconstructed state vectors to construct a joint state vector X(t); The safety domain calculation module is used to perform topological data analysis based on the joint state vector X(t) and extract at least one topological invariant; based on the at least one topological invariant, it calculates the safety domain volume V(t) of the ship's propulsion system at the current time t. The safety margin prediction module is used to perform safety boundary approximation judgment and prediction based on the safety domain volume V(t) and its historical sequence calculated by the safety domain calculation module, and output the remaining safety margin R of the system. The graded response module is used to perform graded early warning or control actions based on the remaining safety margin of the system.

[0005] Furthermore, the main engine vibration signal is a three-axis vibration acceleration signal at the engine foot; the generator electrical signal includes the three-phase voltage and three-phase current signals of the stator winding; and the hull motion signal includes the roll angle, pitch angle, and heave acceleration measured by the inertial measurement unit.

[0006] Furthermore, in the reconstruction and fusion module, phase space reconstruction is performed based on the multi-source time-series signals to generate multiple corresponding reconstructed state vectors, as follows: Obtain one time series signal from a multi-source time series signal, and obtain the embedding dimension d and delay time τ of the time series signal; Using the delay time τ as an interval, d data points are sequentially selected from the time-series signal to form a d-dimensional reconstructed state vector; The embedding dimension d and the delay time τ are determined by the false nearest neighbor method and the mutual information method, respectively.

[0007] Furthermore, in the security domain calculation module, topological data analysis is performed based on the joint state vector X(t) to extract at least one topological invariant, as follows: Point cloud data is generated based on the joint state vector X(t), and persistent cohomology calculation is performed on the point cloud data; From the results of the persistent homology calculation, extract the set of persistent barcodes corresponding to the one-dimensional homotopy group, and calculate the average length of all persistent barcodes in the persistent barcode set as the first topological invariant L1; Extract the set of persistent barcodes corresponding to the zero-dimensional homotopy group, and count the number of persistent barcodes in the set, which is used as the second topological invariant N0.

[0008] Furthermore, in the safety domain calculation module, the calculation of the ship's propulsion system safety domain volume V(t) at the current time t is as follows: Based on the point cloud data of the joint state vector X(t), a safe domain geometric model is constructed, wherein the construction radius α is adaptively determined according to the point cloud density to capture the key topological features identified by the persistent cohomology calculation; Calculate the volume of the high-dimensional geometric body enclosed by the security domain geometric model, generate a volume value, and use the volume value as the security domain volume V(t).

[0009] Furthermore, the safety margin prediction module specifically includes: Calculate the rate of change ΔV of the safe domain volume V(t) relative to the previous time step V(t-Δt); Determine whether the current system state meets the preset warning conditions, which are set based on one or more of the following: the security domain volume V(t), the rate of change ΔV, and the topological invariants obtained from the security domain calculation module. If the aforementioned warning conditions are met, then based on the safety domain volume V(t) and the rate of change ΔV, the change in the safety domain volume within the future Tp period is extrapolated and predicted, and the minimum value in the prediction results is used as the basis for calculating the remaining safety margin R of the system. The warning conditions include at least one of the following: The current safe domain volume V(t) is less than or equal to one-third of the baseline safe volume V0; The instantaneous year-on-year growth rate of the first topological invariant L1 obtained from the security domain computation module reaches or exceeds 50%; The number of second topological invariants N0 obtained from the security domain computation module increases to twice or more of the value of the previous time step.

[0010] Furthermore, the extrapolation prediction is achieved through a Kalman predictor, and the system's remaining safety margin R is determined by the following formula: , in, This represents the predicted safe zone volume at time i, obtained by extrapolation using the Kalman predictor. This is the preset prediction time window length; The rate of change ΔV is calculated as follows: , where Δt is the preset sampling and calculation period.

[0011] Furthermore, in the tiered response module, the execution of tiered early warning or control actions is as follows: Let the baseline safe volume be V0; If the remaining safety margin R of the system is greater than 1 / 3V0, then a first-level warning message is generated and sent to the driver's console display terminal; If 1 / 3V0≥R>1 / 6V0, a ​​second-level warning message is generated and sent to the dashboard display terminal. At the same time, a pause command is sent to the energy efficiency optimization controller to suspend further energy efficiency optimization operations. If R≤1 / 6V0, a ​​third-level warning message is generated and sent to the driver's console display terminal. At the same time, a decoupling control command set is sent to each actuator of the power system. The decoupling control command set includes: a command to reduce the generator excitation angle, a command to reduce the rudder angle of the servo motor and the amplitude of the anti-roll fin angle, and a command to restore the main engine fuel injection advance angle to the reference value.

[0012] Furthermore, it also includes a sample storage learning module: Automatically store all the multi-source time-series signals from the moment before the warning condition is triggered to a preset time period after the trigger, as well as the corresponding joint state vector X(t), topological invariants, and safe domain volume V(t), to form a sample of an instability warning event.

[0013] Compared with the prior art, the present invention can achieve at least the following beneficial effects: 1. This invention innovatively applies topological data analysis methods by synchronously acquiring multi-source signals and constructing a joint state vector. Starting from the essential geometric structure of high-dimensional nonlinear systems, it calculates the volume of the safety domain in real time and predicts its changing trend. This enables early, quantitative assessment and graded active protection of the stability decay of ship power systems under deeply coupled energy efficiency optimization conditions. It fundamentally solves the technical problem that traditional linear models and single-device threshold methods cannot predict the risk of systemic instability, and provides core technical support for the safe and efficient operation of intelligent ships.

[0014] 2. This invention constructs a novel technical path through phase space reconstruction and topological data analysis. By directly extracting features from the geometry and topology of the high-dimensional joint state vector, it can fundamentally characterize and quantify the nonlinear dynamic behavior of the entire coupled system, solving the problem of the failure of traditional linear stability criteria. By calculating the safety domain volume V(t) and making predictions based on its changing trend, it achieves dynamic assessment of the overall stability margin of the system, can identify early signs before micro-disturbances are amplified, and buys critical time for proactive intervention. By establishing a real-time safety monitoring closed loop, it can not only assess risks but also link with the upper-level energy efficiency optimization system through graded early warning or control actions. This achieves dynamic constraints of the safety boundary on energy efficiency optimization at the algorithm level, ensuring that while ships enjoy the energy efficiency benefits brought by intelligence, the operational safety of their power system is always kept within a quantifiable and monitorable protection boundary. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a system principle block diagram of the present invention. Detailed Implementation

[0016] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0017] Please refer to Figure 1 This invention provides a ship intelligent engine room signal acquisition and analysis system, including a signal acquisition module, a reconstruction and fusion module, a safety domain calculation module, a safety margin prediction module, and a graded response module; The signal acquisition module is used to synchronously acquire multi-source timing signals of the ship's power system, including main engine vibration signals, generator electrical signals, steering gear and anti-roll fin action signals, and hull motion signals. The main engine vibration signal is a three-axis vibration acceleration signal at the engine foot; the generator electrical signal includes the three-phase voltage and three-phase current signals of the stator winding; the hull motion signal includes the roll angle, pitch angle and heave acceleration measured by the inertial measurement unit.

[0018] The reconstruction and fusion module is used to perform phase space reconstruction based on the multi-source time-series signals to generate multiple corresponding reconstructed state vectors; and to merge the multiple reconstructed state vectors to construct a joint state vector X(t); In the reconstruction and fusion module, phase space reconstruction is performed based on the multi-source time-series signals to generate multiple corresponding reconstructed state vectors, as follows: Obtain one time series signal from a multi-source time series signal, and obtain the embedding dimension d and delay time τ of the time series signal; It should be noted that the embedding dimension *d* and the delay time *τ* are two core parameters for phase space reconstruction. The delay time *τ* determines the time interval for extracting data points from the original sequence to form the state vector. If *τ* is too small, the components of the reconstructed vector are highly correlated, resulting in information redundancy; if *τ* is too large, the components are almost independent, failing to reflect the dynamic correlation of the system. In this invention, the mutual information method is preferably used to determine the *τ* value of each signal. Specifically, this involves calculating the mutual information function between the original time series and its own sequence after a delay of *τ*, and selecting the *τ* value corresponding to the first local minimum of this function. This balances the independence and correlation between the components of the state vector, ensuring that the reconstructed state space retains the dynamic information of the original system to the greatest extent possible.

[0019] Using the delay time τ as an interval, d data points are sequentially selected from the time-series signal to form a d-dimensional reconstructed state vector; The embedding dimension d and the delay time τ are determined by the false nearest neighbor method and the mutual information method, respectively.

[0020] It should be noted that the spurious nearest neighbor method is preferably used to determine the embedding dimension d for each signal. The basic principle of this method is that in a low-dimensional embedding space, due to projection compression, state points that are not originally adjacent may appear to be nearest neighbors. As the embedding dimension d increases, these spurious nearest neighbors gradually separate. When increasing the dimension no longer significantly reduces the proportion of spurious nearest neighbors, the d value is considered sufficiently large. By independently determining the optimal d and τ for each source signal, it is ensured that each reconstructed state vector can most effectively represent the local dynamics of its corresponding subsystem.

[0021] It should be noted that after obtaining the reconstructed state vectors corresponding to multiple subsystems such as the main engine, generator, steering gear, anti-roll fins, and hull motion, they are merged to construct a joint state vector X(t). This merging operation is not a simple vector concatenation; its key lies in strict time synchronization. At any time t, X(t) contains the reconstructed state information of all key subsystems at the same physical moment, thus mathematically forming a high-dimensional comprehensive state point that can describe the global coupled dynamics of the entire ship's propulsion system. This X(t) is the direct data basis for subsequent topology data analysis to evaluate the global stability of the system.

[0022] It should be noted that combining the spurious nearest neighbor method with the mutual information method to determine parameters is a preferred approach in this embodiment. Its advantage lies in the fact that both methods are based on the characteristics of the data itself, requiring no prior assumptions about the system's mathematical model, making them particularly suitable for complex, time-varying, and difficult-to-accurately model real-world industrial systems such as intelligent ship engine rooms. This data-driven parameter determination method enhances the adaptability and engineering applicability of the technical solution of this invention.

[0023] It should be noted that the purpose of performing phase space reconstruction on multi-source time series signals in the reconstruction and fusion module is to map each one-dimensional observation time series into a higher-dimensional state space through mathematical methods, so as to recover and reveal the complete information of the underlying dynamic system driving the signal.

[0024] The safety domain calculation module is used to perform topological data analysis based on the joint state vector X(t) and extract at least one topological invariant; based on the at least one topological invariant, it calculates the safety domain volume V(t) of the ship's propulsion system at the current time t. It should be noted that the core purpose of topological data analysis of the joint state vector X(t) in the safety domain calculation module is to bypass traditional amplitude or frequency-based indicators and directly extract essential characteristics that profoundly reflect the overall dynamic stability of the system from the geometry and connectivity of the system's global state space (i.e., point cloud). Under nonlinear coupling conditions that pursue ultimate energy efficiency, the operating state of a ship's propulsion system may reside on a complex high-dimensional attractor. Traditional linear methods struggle to characterize its stability margin, while topological methods can effectively describe the robustness of its spatial structure.

[0025] In the security domain calculation module, topological data analysis is performed based on the joint state vector X(t) to extract at least one topological invariant, as follows: The specific topological data analysis is as follows: Point cloud data is generated based on the joint state vector X(t), and persistent cohomology calculation is performed on the point cloud data; From the results of the persistent homology calculation, extract the set of persistent barcodes corresponding to the one-dimensional homotopy group, and calculate the average length of all persistent barcodes in the persistent barcode set as the first topological invariant L1; Extract the set of persistent barcodes corresponding to the zero-dimensional homotopy group, and count the number of persistent barcodes in the set, which is used as the second topological invariant N0.

[0026] It's important to note that persistent cohomology computation is a core method in topological data analysis. It quantifies the birth and disappearance of topological features (such as connected components and holes) in point cloud data at different scales. The basic process is as follows: Imagine a sphere with radius ε constantly increasing, centered on each data point. As ε increases, the spheres begin to intersect and merge, and the topological structure of the point cloud changes accordingly. Persistence refers to the length of the interval ε spanned by a certain topological feature (such as a one-dimensional hole). The longer this interval, the more significant the feature, and the more likely it is to be a true reflection of the underlying data structure, rather than a random phenomenon caused by noise. Performing this computation on the X(t) point cloud yields a topological fingerprint describing the state-space structure of the system at the current moment.

[0027] It should be noted that the first topological invariant L1 (the average length of a persistent barcode) extracted from the persistent cohomology calculation results has a physical meaning related to the dissipated or oscillatory energy of the system state. In the state space of a ship's propulsion system, a significant and persistent one-dimensional hole (corresponding to a long barcode) often suggests the existence of a continuous, stable periodic or cyclic dynamic mode in the system. An increase in the value of L1 can be interpreted as an enhancement in the energy or stability of this cyclic mode. This may correspond to an undesirable and strong nonlinear resonant coupling between different subsystems (such as main engine torsional vibration and shafting undulation), and is an important topological signal that the system is approaching the instability boundary.

[0028] It should be noted that the extracted second topological invariant N0 (number of zero-durability barcodes) has a physical meaning related to the degree of fragmentation or splitting of the system state. Zero-dimensional features describe connected components. Under ideal stable conditions, system state points should be tightly clustered in a main connected region. If N0 increases significantly, it means that the state point cloud exhibits multiple discrete, separate clusters in space. In real systems, this may correspond to the operating state jumping between different attractors, or abrupt changes or bifurcation in the system's dynamic behavior, disrupting overall consistency and serving as a key indicator of severe stability degradation.

[0029] It should be noted that by combining the two topological invariants, L1 and N0, with the safety domain volume V(t) constructed based on the same point cloud, a multi-dimensional and complementary stability assessment system is formed. L1, from the perspective of mode energy, N0, from the perspective of structural integrity, and V(t), from the perspective of the size of the stability space, jointly characterize the nonlinear stability of the ship's propulsion system. This fusion assessment method based on topological data can detect the systemic instability risk caused by deep coupling control earlier and more sensitively than methods that rely solely on a single physical quantity threshold or linear model.

[0030] In the safety domain calculation module, the calculation of the ship's propulsion system safety domain volume V(t) at the current time t is as follows: Based on the point cloud data of the joint state vector X(t), a safe domain geometric model is constructed, wherein the construction radius α is adaptively determined according to the point cloud density to capture the key topological features identified by the persistent cohomology calculation; Calculate the volume of the high-dimensional geometric body enclosed by the security domain geometric model, generate a volume value, and use the volume value as the security domain volume V(t).

[0031] The construction radius α of the security domain geometric model is adapted to the point cloud density of the joint state vector X(t), so that the constructed security domain geometric model can capture the key topological features identified by the persistent cohomology calculation.

[0032] It should be noted that by calculating the volume of the high-dimensional geometry enclosed by this safety domain geometric model, we can quantitatively measure the size of the state space region occupied by the current state point cloud. This region intuitively reflects the range of the set of all possible and interconnected stable operating states of the system under the current dynamic mode. Therefore, its volume is defined as the safety domain volume V(t).

[0033] It should be noted that the adaptive determination logic of the construction radius α is the key to this embodiment. The value of α is essentially a threshold of a spatial scale: when the distance between two points is less than α, a connection can be established between them (forming edges, triangles, and other high-dimensional simplexes). If α is fixed to a constant value, it will be impossible to consistently capture the inherent structure in dynamic systems with uneven point cloud density distribution or those that change over time. Therefore, this invention adaptively calculates α based on the point cloud density of the current joint state vector X(t). A preferred implementation is as follows: first, calculate the statistical distribution of the distances between all points in the point cloud and their nearest neighbors, such as the median or the 75th percentile, and then set α to a multiple of this statistical distance, such as 1.5 to 2 times. In this way, when the state points are densely clustered (high density), α will automatically decrease to finely characterize the local structure; when the state points are relatively dispersed (low density), α will automatically increase to maintain the connectivity of the structure, thereby ensuring that the model is robust to dynamically changing data.

[0034] It is important to note that capturing key topological features identified by persistent cohomology computation implies that the construction of the security domain geometric model needs to be coordinated with the structural information revealed by the topological invariants (L1, N0) extracted in the security domain computation module. Specifically, in persistent cohomology analysis, each topological feature (such as a one-dimensional hole) has a disappearance scale (i.e., the radius value at which the feature disappears). To faithfully preserve those persistent (i.e., longer barcode-like) features considered to be the true structure of the data in the geometric model, while ignoring transient features that may be noise-induced, an adaptively determined α value can be correlated with the disappearance scale of the key one-dimensional holes. For example, α can be set slightly larger than the birth scale of these key holes but smaller than their disappearance scale, thus ensuring that these important toroidal structures are constructed in the security domain geometric model and ultimately contribute to the volume calculation. This makes the security domain volume V(t) not only a measure of spatial size, but also a topological geometric measure containing stability information of key dynamic modes of the system.

[0035] The safety margin prediction module is used to perform safety boundary approximation judgment and prediction based on the safety domain volume V(t) and its historical sequence calculated by the safety domain calculation module, and output the remaining safety margin R of the system. The safety margin prediction module specifically includes: Calculate the rate of change ΔV of the safe domain volume V(t) relative to the previous time step V(t-Δt); Determine whether the current system state meets the preset warning conditions, which are set based on one or more of the following: the security domain volume V(t), the rate of change ΔV, and the topological invariants obtained from the security domain calculation module. If the aforementioned warning conditions are met, then based on the safe domain volume V(t) and the rate of change ΔV, the future... The change in the safety domain volume within a time period is extrapolated and predicted, and the minimum value in the prediction results is used as the basis for calculating the remaining safety margin R of the system. The warning conditions include at least one of the following: The current safe domain volume V(t) is less than or equal to one-third of the baseline safe volume V0; The instantaneous year-on-year growth rate of the first topological invariant L1 obtained from the security domain computation module reaches or exceeds 50%; The number of second topological invariants N0 obtained from the security domain computation module increases to twice or more of the value of the previous time step.

[0036] It should be noted that the baseline safety volume V0 is a key calibration parameter. It defines the typical value of the safety domain volume obtained from historical data or initial trial operation statistics under recognized, well-considered steady-state operating conditions, such as straight navigation in calm seas at the design speed. Establishing V0 provides a static and reliable comparison benchmark for the dynamically changing V(t). Using V(t) ≤ V0 / 3 as the primary warning condition has the physical meaning that when the system's current stable operating state space shrinks to one-third of its baseline size, it indicates that the system's dynamic behavior has significantly deviated from its design steady state, the stability buffer space has been largely consumed, and it has entered a risk zone requiring high vigilance.

[0037] It should be noted that the warning conditions based on topological invariants provide a deeper insight into the nature of stability. The first condition, that the instantaneous year-on-year growth rate of the topological invariant L1 is ≥50%, aims to capture abrupt changes in the system's dynamics. A rapid surge in L1 means that the key toroidal structure characterizing the intensity of the system's nonlinear coupled oscillations suddenly becomes unusually significant and persistent in the state space. This is usually direct topological evidence of the excitation of nonlinear resonances or strong energy exchange between different oscillation modes, and is a potential precursor to instability preceding large-amplitude mechanical vibrations. The second condition, that the topological invariant N0 increases to twice or more of its value at the previous moment, aims to capture the structural collapse of the system's state. Doubling N0 means that the state point cloud rapidly splits from a relatively coherent mass into multiple discrete clusters. This strongly suggests that the system may be undergoing dynamic bifurcation, its operating state may be jumping between different attractors, and the overall coordination and consistency of its behavior is being lost, which is a critical signal of systemic instability.

[0038] The extrapolation prediction is achieved through a Kalman predictor, and the system's remaining safety margin R is determined by the following formula: , in, This represents the predicted safe zone volume at time i, obtained by extrapolation using the Kalman predictor. This is the preset prediction time window length; The rate of change ΔV is calculated as follows: ΔV = V(t) - V(t - Δt), where Δt is the preset sampling and calculation period.

[0039] It should be noted that the core of the Kalman predictor lies in its internal state transition model. In the specific application of this invention, the dynamic change process of the safety region volume V(t) can be modeled as a state space, for example, the system state can be defined as... This means that it simultaneously includes the volume value and its changing trend. The predictor updates the time step (prediction step) by using the posterior estimate of the previous time step and the equation describing the state evolution, such as V(t+1) = V(t) + ΔV(t) * Δt + process noise, and then updates the measurement step (correction step) by combining the actual observed value V(t) at the current time step. This recursive prediction-correction structure enables it to effectively filter out random fluctuations (observation noise) that may exist in the volume calculation, and based on the inherent dynamic law of the system, it makes a more robust and accurate estimate of the future state than simple linear extrapolation, thus obtaining a reliable prediction sequence. .

[0040] It should be noted that the prediction time window length It is a key design parameter that directly determines the lead time for forecasting. The configuration needs to comprehensively consider the dynamic response inertia of the ship's propulsion system and the necessary safety decision-making time. If If the forecast and early warning are too short, the forecast and early warning may not allow enough time for subsequent control and intervention; if If the time frame is too long, the uncertainty of the prediction will increase significantly, reducing the reliability of the warning. A preferred method is... It should be greater than or equal to the total time required from triggering the warning to executing decoupled control (such as a three-level response in a tiered response module) and causing a clear trend change in the system state. For example, if the effective time of the entire control loop is approximately 20 seconds, then... It can be set to 25-30 seconds to provide a reasonable safety buffer.

[0041] It should be noted that the formula This embodies the worst-case criterion for safety assessment in this invention. Instead of using the average or final value over the forecast period, it uses the minimum of all future forecast values ​​as the residual safety margin R. The conservative safety philosophy of this design lies in the fact that as long as the future... Within a given timeframe, there exists a moment when the predicted safe zone volume may fall to a dangerous level. The system should then assess the overall risk and take action based on this weakest point. This allows the early warning mechanism to prevent the most dangerous instantaneous conditions, rather than average safety, greatly enhancing the robustness of ships in dealing with sudden stability degradation under complex sea conditions and intelligent control coupling.

[0042] It is important to note that the selection of Δt (sampling and calculation period) in the rate of change ΔV = V(t) - V(t - Δt) is crucial, requiring a balance between reflecting dynamic details and ensuring computational stability. Δt must be less than the oscillation period of the system's critical instability modes; for example, the low-frequency oscillation period of the host-shaft-generator coupling might be several seconds, in order to capture the trend of volume change. At the same time, Δt should not be too small to avoid excessive computational load and noise amplification. In practice, Δt can be determined based on the control period of the main controlled equipment (such as the host speed controller) and the throughput capacity of the signal acquisition system, typically maintaining consistency with or being an integer multiple of the signal sampling period of the step signal acquisition module to ensure data link synchronization and coordination. Through ΔV, the Kalman predictor can quantify the instantaneous rate of system stability decay, which is indispensable dynamic information for accurate extrapolation.

[0043] It should be noted that the core of the safety margin prediction module lies in establishing a dynamic and forward-looking stability decay early warning and quantitative prediction mechanism. It does not issue an alarm after the system becomes unstable, but rather actively determines whether the system is heading towards a dangerous boundary by utilizing the real-time changing trends of the safety domain volume V(t) and its topological associated indices (L1, N0), and quantitatively predicts its remaining range, i.e., the system's remaining safety margin R.

[0044] The graded response module is used to perform graded early warning or control actions based on the remaining safety margin of the system.

[0045] In the tiered response module, the execution of tiered early warning or control actions is as follows: Let the baseline safe volume be V0; If the remaining safety margin R of the system is greater than 1 / 3V0, then a first-level warning message is generated and sent to the driver's console display terminal; If 1 / 3V0≥R>1 / 6V0, a ​​second-level warning message is generated and sent to the dashboard display terminal. At the same time, a pause command is sent to the energy efficiency optimization controller to suspend further energy efficiency optimization operations. If R≤1 / 6V0, a ​​third-level warning message is generated and sent to the driver's console display terminal. At the same time, a decoupling control command set is sent to each actuator of the power system. The decoupling control command set includes: a command to reduce the generator excitation angle, a command to reduce the rudder angle of the servo motor and the amplitude of the anti-roll fin angle, and a command to restore the main engine fuel injection advance angle to the reference value.

[0046] It should be noted that the first-level response (R>V0 / 3) is designed as a purely suggestive warning. When the system is still within a relatively ample safety margin (volume greater than one-third of the baseline), but has triggered the warning conditions in the safety margin prediction module, such as abnormal topology indicators, this level only generates warning information on the dashboard display terminal, such as text, icons, or sound prompts, informing the driver that the system is entering a monitoring operating range. Its purpose is to enhance the crew's situational awareness, transforming the implicit judgments of the intelligent system into explicit human knowledge, preparing for possible subsequent manual or automatic system intervention, without directly interfering with the current automated control process.

[0047] It should be noted that the second-level response (V0 / 3≥R>V0 / 6) marks the system's transition from the attention phase to the active defense phase. In this phase, besides escalating warning information to the bridge, the core action is sending a pause command to the energy efficiency optimization controller. This forces the upper-level intelligent control system to stop executing energy efficiency optimization algorithms that could further deepen coupling and become nonlinear (such as attempting a more aggressive speed-main engine-generator coordination point), and locks the current control command or switches to a conservative steady-state control mode. Its fundamental purpose is to freeze the control decision-making process that could push the system into a more dangerous state, providing the system with a window of opportunity to stabilize itself or recover margin through routine adjustments—a crucial automatic decision to prevent the situation from escalating.

[0048] It should be noted that the third-level response (R≤V0 / 6) is the highest level of emergency stabilization intervention. At this point, the system is considered extremely close to the instability boundary, and control must be applied immediately and proactively to forcibly reduce the dynamic coupling strength within the system. To this end, the system sends a predefined set of decoupling control instructions to each dynamic subsystem: Reduce generator excitation angle command: This command aims to rapidly reduce the internal electromotive force of the synchronous generator, thereby weakening the strong electromagnetic coupling between the generator and the ship's power grid, as well as between the power grid and other power equipment, and suppressing potential electromechanical oscillations.

[0049] Commands to reduce the amplitude of rudder angle and anti-roll fin angle: These commands aim to reduce the strong and rapid intervention of the hull motion control subsystem in the ship's attitude. Large rudder fin movements themselves generate strong hydrodynamic loads and inertial coupling; reducing their amplitude can quickly cut off strong external excitations from the heading / attitude control loop, allowing the hull-main engine system to return to a smoother dynamic.

[0050] The command to restore the engine fuel injection advance angle to the reference value aims to reverse unconventional, critical combustion control strategies that may be used in pursuit of ultimate thermal efficiency, so that the engine operating point returns from a nonlinear range that may induce combustion instability or drastic torque fluctuations to a well-proven reference condition with smoother dynamic characteristics.

[0051] The overall strategy of this set of combined instructions is strategic retreat. By temporarily sacrificing some performance (efficiency, maneuverability), it simultaneously and rapidly reduces the overall nonlinear coupling energy and interaction intensity of the entire system from multiple dimensions, creating the necessary conditions for restoring global stability.

[0052] It should be noted that the graded response module establishes a graded response system of perception, early warning, decision-making, and control that is linked to the precise quantification of predictive margin. Its core design principle is: based on the degree of decay of the system's remaining safety margin R, measures with increasing intensity and depth of intervention are adopted to maximize the preservation of the intelligent system's energy efficiency optimization function under the absolute premise of ensuring navigation safety, thereby achieving a dynamic balance between safety and efficiency.

[0053] The sample storage learning module automatically stores all the multi-source time-series signals from the moment before the warning condition is triggered to a preset time period after the trigger, as well as the corresponding joint state vector X(t), topological invariants, and safe domain volume V(t), forming a sample of an instability warning event for subsequent adaptive training of the prediction model.

[0054] It should be noted that the automatic storage and learning mechanism for instability warning event samples established by the sample storage and learning module is the core of this intelligent system's ability to continuously evolve. It stipulates that once a warning condition is triggered (regardless of whether it ultimately escalates to control intervention), the system will automatically save a full-dimensional data snapshot for a period before and after the event (e.g., 2 minutes before triggering to 5 minutes after triggering). This sample library not only contains the original sensor signals, but more importantly, it includes the intermediate products (X(t)) and high-level features (topological invariants, V(t)) of the entire analysis chain, forming a digital specimen for in-depth analysis. These specimens are used for offline adaptive training and calibration of prediction models (such as the state transition model of a Kalman predictor) or warning thresholds, enabling the system to continuously learn from boundary conditions encountered in actual navigation, optimize its prediction accuracy, and potentially discover new, unpredictable instability precursor patterns, thereby achieving increasingly intelligent safety protection capabilities with continued use.

[0055] In summary, this patent application, in the context of deep coupling control of intelligent ship engine rooms, achieves real-time, quantitative, and forward-looking assessment of the stability state of high-dimensional nonlinear systems and proactive safety protection, laying a crucial safety foundation for intelligent ship navigation.

[0056] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0057] In the embodiments provided by this invention, it should be understood that the disclosed system or method can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0058] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0059] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0060] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the basic characteristics of the present invention.

[0061] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A ship intelligent engine room signal acquisition and analysis system, characterized in that, include: The signal acquisition module is used to synchronously acquire multi-source timing signals of the ship's power system, including main engine vibration signals, generator electrical signals, steering gear and anti-roll fin action signals, and hull motion signals. The reconstruction and fusion module is used to perform phase space reconstruction based on the multi-source time-series signals to generate multiple corresponding reconstructed state vectors; and to merge the multiple reconstructed state vectors to construct a joint state vector. The safety domain calculation module is used to perform topological data analysis based on the joint state vector, extract at least one topological invariant, and calculate the safety domain volume of the ship's propulsion system at the current moment based on the at least one topological invariant. The safety margin prediction module is used to perform safety boundary approximation judgment and prediction based on the safety domain volume and its historical sequence calculated by the safety domain calculation module, and output the remaining safety margin of the system. The graded response module is used to perform graded early warning or control actions based on the remaining safety margin of the system.

2. The ship intelligent engine room signal acquisition and analysis system as described in claim 1, characterized in that: In the signal acquisition module, the main engine vibration signal is the triaxial vibration acceleration signal at the engine foot; the generator electrical signal includes the three-phase voltage and three-phase current signals of the stator winding; and the hull motion signal includes the roll angle, pitch angle, and heave acceleration measured by the inertial measurement unit.

3. The ship intelligent engine room signal acquisition and analysis system as described in claim 1, characterized in that: In the reconstruction and fusion module, phase space reconstruction is performed based on the multi-source time-series signals to generate multiple corresponding reconstructed state vectors, as follows: Obtain one time series signal from a multi-source time series signal, and obtain the embedding dimension and delay time of the time series signal; Using the delay time as an interval, d data points are sequentially selected from the time-series signal to form a d-dimensional reconstructed state vector; The embedding dimension and the delay time are determined by the false nearest neighbor method and the mutual information method, respectively.

4. The ship intelligent engine room signal acquisition and analysis system as described in claim 1, characterized in that: In the security domain calculation module, topological data analysis is performed based on the joint state vector to extract at least one topological invariant, as follows: Point cloud data is generated based on the joint state vector, and persistent cohomology calculation is performed on the point cloud data; From the results of the persistent homology calculation, extract the set of persistent barcodes corresponding to the one-dimensional homotopy group, and calculate the average length of all persistent barcodes in the persistent barcode set as the first topological invariant; Extract the set of persistent barcodes corresponding to the zero-dimensional homotopy group, and count the number of persistent barcodes in the set as the second topological invariant.

5. The ship intelligent engine room signal acquisition and analysis system as described in claim 4, characterized in that: In the safety domain calculation module, the calculation of the current moment's ship propulsion system safety domain volume is as follows: Based on the point cloud data of the joint state vector, a safe domain geometric model is constructed, wherein the construction radius is adaptively determined according to the point cloud density to capture key topological features identified by the persistent cohomology calculation; Calculate the volume of the high-dimensional geometric body enclosed by the security domain geometric model, generate a volume value, and use the volume value as the security domain volume.

6. The ship intelligent engine room signal acquisition and analysis system as described in claim 1, characterized in that: The safety margin prediction module specifically includes: Calculate the rate of change of the safety domain volume relative to the previous time step; Determine whether the current system state meets the preset warning conditions, which are set based on one or more of the following: security domain volume, rate of change, and topological invariants obtained from the security domain calculation module. If the aforementioned warning conditions are met, then based on the safety domain volume and rate of change, the change in the safety domain volume in the future time period is extrapolated and predicted, and the minimum value in the prediction results is used as the basis for calculating the remaining safety margin of the system. The warning conditions include at least one of the following: The current security domain volume is less than or equal to one-third of the baseline security volume; The instantaneous year-on-year growth rate of the first topological invariant obtained from the security domain computing module reaches or exceeds 50%; The number of second topological invariants obtained from the security domain computation module increases to twice or more the value of the previous time step.

7. The ship intelligent engine room signal acquisition and analysis system as described in claim 6, characterized in that: The extrapolation prediction is achieved through a Kalman predictor, and the system's remaining safety margin R is determined by the following formula: ,in, This represents the predicted safe zone volume at time i, obtained by extrapolation using the Kalman predictor. This is the preset prediction time window length; The rate of change ΔV is calculated as follows: , where Δt is the preset sampling and calculation period.

8. The ship intelligent engine room signal acquisition and analysis system as described in claim 1, characterized in that: In the tiered response module, the execution of tiered early warning or control actions is as follows: Let the baseline safe volume be V0; If the remaining safety margin R of the system is greater than 1 / 3V0, then a first-level warning message is generated and sent to the driver's console display terminal; If 1 / 3V0≥R>1 / 6V0, a ​​second-level warning message is generated and sent to the dashboard display terminal. At the same time, a pause command is sent to the energy efficiency optimization controller to suspend further energy efficiency optimization operations. If R≤1 / 6V0, a ​​third-level warning message is generated and sent to the driver's console display terminal. At the same time, a decoupling control command set is sent to each actuator of the power system. The decoupling control command set includes: a command to reduce the generator excitation angle, a command to reduce the rudder angle of the servo motor and the amplitude of the anti-roll fin angle, and a command to restore the main engine fuel injection advance angle to the reference value.

9. The ship intelligent engine room signal acquisition and analysis system as described in claim 1, characterized in that: It also includes a sample storage learning module: The system automatically stores all the multi-source time-series signals from the moment before the warning condition is triggered to a preset time period after the trigger, as well as the corresponding joint state vector, topological invariant, and safe domain volume, to form a sample of an instability warning event.

Images