Fault identification and fault-tolerant control device for cable allowance control of optical cable laying ship

Abstract

The invention relates to a fault identification and fault-tolerant control device for controlling the cable allowance of an optical cable laying ship. The device comprises a fast fault identification module, a slow fault confirmation module and a fault-tolerant control instruction output unit. The rapid fault identification module predicts and generates candidate fault states based on the cable tension, the second derivative of the water depth, the model residual abrupt change and the tension trend; the low-speed fault confirmation module further performs stability confirmation on candidate fault states through filtering trend judgment, combined information entropy analysis and a multi-cycle consistency voting mechanism; the fault-tolerant control instruction output unit comprises a multi-model control subunit, a sliding mode control subunit and a finite time domain prediction subunit, and can fuse three types of control strategies to generate a fault-tolerant adjustment instruction after a fault is confirmed, thereby achieving the dynamic optimization control of the speed of the cable winch. According to the method, the cable allowance can be accurately controlled under complex sea conditions and topographic conditions, and the method has the capability of rapidly identifying abnormal states and dynamic fault tolerance.

Description

Technical Field

[0001] The present application relates to the field of submarine optical cable laying, and in particular to a fault identification and fault-tolerant control device for controlling the cable margin of an optical cable laying vessel. Background Art

[0002] During submarine cable laying, the cable is lowered into the sea using a winch aboard the laying vessel and then lowered to the seabed by gravity. Due to the large variations in seabed topography, particularly in areas with sudden slope changes, improper matching of the winch release speed with the vessel's forward speed, water depth, and other parameters can easily lead to excessive slack or over-tension in the cable underwater. Excessive slack can cause the cable to form loops and accumulate in the water, increasing the risk of mechanical damage. Over-tensioning can easily lead to excessive tension, potentially causing cable breakage.

[0003] To achieve precise control during cable payout, traditional systems often use fixed proportional parameters to adjust the payout speed, or employ simple closed-loop control based on tension feedback signals. However, in actual operations, cable tension is affected by multiple factors, resulting in feedback signal delays. Fixed models cannot effectively respond to dynamic factors such as sudden changes in water depth and vessel speed. Consequently, when sudden disturbances or measurement anomalies occur during system operation, traditional methods are unable to promptly identify and enter fault-tolerant mode, creating operational risks. Summary of the Invention

[0004] In order to accurately identify the abnormal change trend of cable tension during the cable laying process and combine multi-source data to realize early judgment of potential fault status, the present application provides a fault identification and fault-tolerant control device for cable margin control of optical cable laying vessels.

[0005] The present application provides a fault identification and fault-tolerant control device for cable margin control on an optical cable laying vessel, which adopts the following technical solution: A fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel, the device comprising: A fast fault identification module is used to collect real-time navigation parameters of the vessel and real-time operation parameters of the equipment in a preset first time period, and generate candidate fault status signals based on the second-order derivative changes of tension and water depth, the sudden jump of the control model residual error, and the tension change trend prediction; A slow fault confirmation module is configured to confirm the candidate fault state in a preset second time period, the slow fault confirmation module comprising: The filtering trend judgment unit is used to judge the continuous change trend of cable tension or water depth through median filtering and smoothing derivatives. The joint information entropy calculation unit is used to calculate the joint information entropy of real-time navigation parameters and real-time equipment operation parameters, and to judge the degree of deviation of the current system state from the preset normal entropy range. a multi-cycle consistency voting unit, configured to determine the number of occurrences of a candidate fault state in a plurality of said first time periods, and to determine whether a fault state is triggered based on trend determination and entropy deviation results; The fault-tolerant control instruction output unit is used to output a trigger instruction to the margin controller after the fault state is confirmed, so as to switch into the fault-tolerant control mode.

[0006] By adopting this technical solution, a two-level judgment structure combining a fast fault identification module and a slow fault confirmation module enables timely detection and robust confirmation of abnormal conditions in dynamic environments. This is particularly suitable for paving areas with complex seabed topography and frequent hydrological disturbances. The fast fault identification module uses the second-order derivatives of tension and water depth as mutation indicators to quickly capture sudden tension changes caused by sudden topography changes or ship speed fluctuations. The model residual and trend prediction mechanism further provides anomaly judgment based on modeling errors and future evolution trends, making candidate fault conditions sufficiently sensitive.

[0007] However, due to the short-cycle, high-frequency characteristics of sea disturbances, mutations within a single cycle may be false alarms. Therefore, this solution introduces filtered trend consistency judgment and joint information entropy offset detection through a slow confirmation module. Multiple state parameters are comprehensively evaluated in the form of information entropy to determine the current system deviation. A multi-cycle voting mechanism is used to verify stability in the time dimension. Fault confirmation is triggered only when the trend direction remains consistent and the entropy value significantly deviates. This design not only avoids unnecessary control switching caused by occasional disturbances, but also improves the accuracy of identifying trend-related faults, enabling the system to intervene in control adjustments at the early stage of fault signs. In principle, this effectively reduces the risk of cumulative damage to optical cables caused by sudden tension loss.

[0008] Optionally, the fast fault identification module generates candidate fault states based on the following mutation identification conditions: like or It is determined to be a mutation disturbance state; If the residual r T =|T measured -T model |>μ r +2σ r , then it is determined to be an abnormal state of the model residual; If the tension exceeds the set threshold within 5 seconds based on Kalman filtering extrapolation, it is determined to be a trend prediction abnormal state.

[0009] By adopting this technical solution and setting multiple, parallel mutation identification conditions, we can classify and determine different types of anomalies during cable laying, enhancing the pertinence and interpretability of the identification mechanism. In practical applications, abnormal changes in cable tension can result from transient shocks caused by external sea disturbances, as well as from the cumulative trend of decreased control model accuracy or the system state slowly deviating from the desired trajectory.

[0010] Specifically, the second-order derivatives of tension and water depth are used to monitor the acceleration changes of the system state. Their large jumps often reflect the physical impact caused by sudden disturbances, such as sudden changes in winch load caused by sudden waves or sudden changes in terrain. The model residual reflects the structural deviation between the actual measured value and the predicted value of the system. When the residual exceeds the statistical threshold μ r +2σ r When the tension is exceeded, it can effectively identify systemic problems such as sensor anomalies, model mismatch or execution hysteresis; Kalman filter extrapolation is used for tension trend prediction, which can identify the direction and rate of change before the tension exceeds the limit, and in principle gives the system the ability to intervene in control adjustments in advance.

[0011] The parallel deployment of the three types of identification conditions enables the rapid fault identification module to respond to emergencies in a timely manner while capturing long-term cumulative effects or potential trends of loss of control, thereby effectively improving the integrity and timeliness of fault identification and ensuring that candidate fault states have engineering usability and clear judgment basis.

[0012] Optionally, the filtering trend judgment unit includes: The median filter subunit is used to perform sliding window median filtering on the continuously collected cable tension signal or water depth signal to eliminate instantaneous fluctuations or abnormal noise. The real-time parameters of the ship's navigation include ship speed and water depth, and the real-time parameters of the equipment's operation include cable tension and cable deployment speed. a smoothed derivative calculation subunit for calculating the first-order derivative based on the filtered signal and applying a low-pass filter to extract the continuous trend component; A trend consistency judgment subunit is used to judge whether the signs of the tension or water depth derivative are consistent in at least two second time periods in the past. If the signs are consistent and the amplitude change is within a preset range, it is determined that a continuous trend exists; The trend direction output subunit is used to output the trend direction signal and mark it as an upward trend or a downward trend.

[0013] By adopting the above technical solution, median filtering, smoothing derivative and trend consistency judgment are set in the filtering trend judgment unit, which effectively improves the ability to stably identify the trend of tension or water depth changes, especially in strongly disturbed sea conditions, and can reduce the interference of single-cycle mutations on the system judgment logic.

[0014] Optionally, the joint information entropy calculation unit assigns preset weight coefficients w1, w2, and w3 to the sudden disturbance state, the model residual abnormal state, and the trend prediction abnormal state, respectively, where w3>w2>w1; In three consecutive first time periods, the candidate states appearing in each period are multiplied by the corresponding weight value, and the total score value S is accumulated. fault ; Wherein, the preset second time period includes three consecutive first time periods; The slow confirmation module triggers the fault confirmation signal based on the following fuzzy rules: When S fault ≥θ S ; Or the joint information entropy offset exceeds the threshold ΔH>∈ H ; Or the trend judgment subunit judges that the trend in the same direction is abnormal for two consecutive periods; The system is judged to have entered a fault state and a fault-tolerant trigger instruction is output.

[0015] By adopting the above technical solution, weights are set for different types of candidate fault states, and the score value S is accumulated over multiple cycles. fault , making the fault diagnosis process discriminative and memorable. Sudden disturbances are treated as low-weighted references, while trend-based anomalies are given higher weight due to their forward-looking nature, helping the system prioritize responses to future high-risk conditions. Furthermore, combining information entropy shift and trend consistency as complementary criteria avoids the scoring mechanism's reliance on a single metric, fundamentally improving the stability of fault confirmation and its ability to mitigate false positives.

[0016] Optionally, the fault-tolerant control instruction output unit includes: A multi-model control subunit, used to switch between preset control models according to seabed topography characteristics, including flat slope model, upslope model and downslope model; The sliding mode control subunit is used to construct a sliding mode surface for the cable margin error and perform disturbance compensation on the winch control command in real time. The finite time domain prediction subunit is used to predict the future margin error trajectory based on the current state and provide feedforward adjustment to the sliding mode control subunit.

[0017] By employing this technical solution, the terrain is divided into three types: flat, uphill, and downhill, and the corresponding control models are matched to adapt the basic cable release speed to the terrain slope. The sliding mode control subunit uses the cable margin error to construct a sliding mode surface to quickly suppress error drift. The finite time domain prediction subunit makes a forward-looking estimate of future error evolution based on the current state and outputs a feedforward adjustment variable. The synergistic integration of these three subunits can simultaneously address both immediate response to sudden changes in local terrain and proactively adjust for trend deviations, effectively reducing the problems of over-release of cables or sudden tension increases caused by delayed response.

[0018] Optionally, the multi-model control subunit is configured to perform the following steps: Calculate the elevation angle of the seabed terrain corresponding to the current ship position based on the input seabed path data; Select the flat slope model, upslope model or downslope model based on the relative relationship between the seabed elevation angle and the current cable laying angle: load the predefined tension-vessel speed control parameters and margin release rate parameters according to the selected model; Based on the current ship speed v ship (t), water depth d(t) and the control parameters of the selected model, calculate the model control quantity u model (t); Among them, the flat slope model corresponds to the elevation angle γ(t) being less than the first set threshold; The uphill model corresponds to the elevation angle γ(t) being greater than the second set threshold, and the real-time laying angle α(t) being less than the elevation angle γ(t), and is selected; the downhill model corresponds to the elevation angle γ(t) being less than the negative threshold, and the real-time laying angle α(t) being greater than the elevation angle γ(t).

[0019] By implementing this technical solution, the system calculates the seabed elevation angle in real time and selects a matching control model based on the laying angle, enabling cable payout rate adjustments to better adapt to actual terrain variations. Different terrains correspond to different tension-vessel speed parameters and margin release strategies. This prevents problems such as insufficient tension on uphill sections or looping due to excessive margin on downhill sections. This fundamentally improves the control output's adaptability to dynamic terrain characteristics and reduces the risk of error accumulation associated with relying on a single model.

[0020] Optionally, the model control quantity u model (t) is calculated as a linear combination or as a ratio based on the target residual release rate; Linear combination form: u model (t) = k v ·v ship (t)+k d d(t)+k c , where k v ,k d ,k c is the scale parameter corresponding to the model; The ratio form of the release rate based on the target margin is: Among them, L target (t) is the current target cable length, T release (t) is the release period.

[0021] By adopting the above technical solution and introducing two switchable model control variable calculation methods, winch speed control becomes more adaptable. The linear combination method adjusts the cable payout speed in real time based on ship speed and water depth, making it suitable for continuous control in conventional flat terrain. The ratio method directly calculates the required cable payout per unit time based on the target cable payout length and cycle time, making it suitable for short-segment, fine-grained control scenarios with stricter laydown rhythm requirements. Switching between these two methods allows dynamic selection based on the characteristics of the operation section, improving the control strategy's accuracy in matching different terrains and operation phases.

[0022] Optionally, the sliding mode control subunit is configured to perform the following steps: Real-time calculation of the margin error e(t)=L between the actual cable length and the preset target length actual (t)-L target (t); Construct the sliding surface function based on the residual error and its derivative Where λ is the slope parameter of the sliding surface; Output control compensation u based on sliding surface function sm (t) = -K s sign(s(t)), where K s is the sliding mode gain coefficient, sign(·) is the sign function; Perform boundary layer processing on the sign function sign(·) or use a super-helical sliding mode algorithm to alleviate control signal oscillation; The sliding mode control compensation u sm (t) and model control quantity u model (t) Output u after fusion total (t) = α·u model (t)+β·u sm (t)+γ·u pred (t), for the winch speed control system to call, where u pred (t) is the feedforward control quantity output by the finite time domain prediction subunit, α, β, γ are the control quantity fusion weight coefficients, satisfying α+β+γ=1, which are set according to empirical values or adjusted by the system operation status.

[0023] By adopting the above technical solution, by constructing the sliding surface The margin error and its rate of change are uniformly incorporated into the control logic to achieve a rapid response to the dynamic behavior of the error. When s(t) deviates from zero, the system immediately outputs the compensation amount u sm (t) = -K s sign(s(t)) allows for clear-direction adjustments at the initial stage of error amplification. The introduction of boundary layer or super-helical sliding mode technology effectively suppresses the high-frequency oscillations caused by the sign function in traditional sliding mode control. This fundamentally improves the smoothness of the adjustment process and the practicality of the controller in continuous sea conditions.

[0024] Optionally, the limited time domain prediction subunit is configured to perform the following steps: Establishing a prediction margin error trajectory equation based on a state space model or an identification model, wherein the state space model assumes that the system state change has short-term continuity and is suitable for control prediction in a slowly varying tension environment; Set the rolling minimization objective function within the prediction time domain Among them, e(t+k) is the prediction margin error, Δu(t+k) is the winch speed change, and ρ is the control balance coefficient; The optimal first few steps of the predicted future winch speed adjustment {u(t+1), u(t+2), ..., u(t+N)} are used as the feedforward control quantity u pred (t) Output to the sliding mode control subunit; The prediction time domain length N is dynamically adjusted according to the system load status and response characteristics.

[0025] By adopting the above technical solution and introducing a finite time domain prediction subunit, the control system can identify the development trend of the margin error in advance and perform feedforward correction. The prediction model assumes that the system state changes smoothly in a short time based on the state space, which is suitable for capturing the error evolution trajectory under slowly changing tension. By optimizing the weighted sum of the square of the margin error and the square of the control change in the objective function, the drastic fluctuation of the winch speed can be suppressed while ensuring the control accuracy. The output feedforward adjustment u pred (t) can compensate for the error accumulation caused by response lag, and dynamically adjust the prediction time domain length N to ensure the timeliness and adaptability of control calculations under different load conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 FIG2 shows a connection diagram of internal modules of a fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to an embodiment of the present invention.

[0027] Figure 2 FIG. 1 is a diagram illustrating the internal module connections of a slow fault confirmation module according to an embodiment of the present invention.

[0028] Figure 3 FIG. 4 is a diagram illustrating the internal module connections of a filtering trend determination unit according to an embodiment of the present invention.

[0029] Figure 4 FIG. 1 is a diagram illustrating the internal module connections of a fault-tolerant control instruction output unit according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The present application will be further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0031] In the following description, for the purpose of explanation, many specific details are set forth in order to provide a thorough understanding of the inventive concepts. Some of the figures in the drawings of the present disclosure, which are part of this specification, represent structures and devices in block diagram form to avoid making the disclosed principles complicated and obscure. For the sake of clarity, not all features of an actual implementation are necessarily described. In addition, the language used in this disclosure has been selected primarily for readability and instructional purposes and may not have been selected to delineate or limit the subject matter of the invention, thereby resorting to the necessary claims to determine such inventive subject matter. References in this disclosure to "one embodiment" or "an embodiment" mean that the specific features, structures or characteristics described in conjunction with that embodiment are included in at least one embodiment, and multiple references to "one embodiment" or "an embodiment" should not be understood to necessarily all refer to the same embodiment.

[0032] Unless expressly limited, the terms "a", "an" and "the" are not intended to refer to a singular entity, but rather to include a general class of which a specific example may be used for illustration. Thus, the use of the term "a" or "an" may mean any number of at least one, including "one", "one or more", "at least one", and "one or more than one". The term "or" means any of the alternatives and any combination of the alternatives, including all, unless the alternatives are expressly indicated to be mutually exclusive. The phrase "at least one of" when combined with a list of items refers to a single item in the list or any combination of the items in the list. The phrase does not require all of the listed items unless expressly limited to that.

[0033] The present application discloses a fault identification and fault tolerance control device for cable margin control of optical cable laying vessels, referring to Figure 1 ,The device includes a fast fault identification module and a slow fault confirmation module.

[0034] The fast fault identification module and the slow fault confirmation module work together to form the pre-conditioning layer for the cable-laying vessel's margin control system, serving as the starting point for the entire closed-loop control system. This two-layer design aims to effectively identify adverse conditions that could lead to abnormal cable posture, uncontrolled tension, or sudden changes in winch commands, while balancing response speed and accuracy. This allows for timely triggering of fault-tolerance mechanisms to prevent issues such as cable dangling, looping, and excessive tension during actual operations.

[0035] Specifically, the rapid fault identification module is used to collect the real-time navigation parameters of the ship and the real-time operation parameters of the equipment in a preset first time period, and generate candidate fault status signals based on the second-order derivative changes of tension and water depth, the control model residual jump and the tension change trend prediction.

[0036] The main function of the fast fault identification module is high-frequency scanning and preliminary judgment. Its working cycle is usually set to 1 or 2 seconds, which is much shorter than the adjustment cycle of the control system. In this module, by real-time monitoring of key parameters such as ship speed, water depth, cable tension, cable release speed, etc., mathematical derivative analysis, control model fitting residual calculation and trend prediction are used to quickly identify "candidate fault states". For example, if the second-order derivative of cable tension or water depth increases suddenly in a short period of time, it means that the current system may be affected by sudden disturbances (such as underwater sudden slopes or flow field changes). The output generated by this module does not trigger fault tolerance immediately, but packages the suspected state as a candidate and provides it as input to the subsequent slow confirmation module for further verification.

[0037] Specifically, the fast fault identification module generates candidate fault states based on the following mutation identification conditions: like or It is determined to be a mutation disturbance state; If the residual r T =|T measured -T model |>μ r +2σ r , then it is determined to be an abnormal state of the model residual; If the tension exceeds the set threshold within 5 seconds based on Kalman filtering extrapolation, it is determined to be a trend prediction abnormal state.

[0038] The core task of the fast fault identification module is to quickly determine whether the system has a potential abnormal evolution trend based on multi-source real-time data, and output the suspicious state as a "candidate fault state" for further processing by the slow confirmation module. The main data processed by this module include cable tension T(t), water depth D(t), ship speed v ship (t) and cable release speed v cable In the system implementation, these data are obtained through tension sensors, ship-borne depth sounders, GPS positioning systems and cable payout monitors, and are input into the module after unified time stamping.

[0039] The identification logic is based on three complementary discrimination mechanisms, each targeting a different type of system instability precursor. The first is the sudden disturbance identification mechanism. The system calculates the second-order derivative of cable tension and water depth every second, namely: and If any of these triggers or This indicates a sudden change in the variable, possibly due to a sudden slope, surge, or mechanical impact from a winch. For example, during an actual laying operation, when the vessel reached the top of a certain seabed cliff, the water depth rapidly changed from 30 meters to 50 meters in 3 seconds. The rate of change of the second-order derivative reached 10 times the average value during stable operation. The rapid identification module immediately marked this as a "sudden disturbance state" and generated a candidate fault state.

[0040] The second type of mechanism is the model residual jump identification mechanism. This mechanism relies on a set of tension prediction models, which take ship speed, casting speed and water depth as input variables and output tension prediction value T model (t), and calculate the residual difference with the measured tension: r T (t)=|T measured (t)-T model (t)| Then the system is based on the residual mean μ of the last 10 seconds r and standard deviation σ r Calculate the abnormal threshold, when r T (t)>μ r +2σ r This method can effectively capture control modeling errors caused by ocean current changes, winch friction, or sensor offset, and has been proven reliable in multiple long-distance laying missions.

[0041] The third type of mechanism is the trend prediction anomaly recognition mechanism. This mechanism uses the Kalman filter to perform state estimation and trend extrapolation on the tension value, and constructs the following recursive structure: in To estimate the tension state, u(t) is the input control variable, A, B is the system dynamic matrix obtained by identification. After each prediction, the system compares the predicted value with the tension upper limit T max If the predicted result crosses the threshold within the next five seconds, a "trend-based prediction anomaly" is considered present. For example, during a gentle descent on the seabed, although the current tension was within the normal range, Kalman predicted, based on the ship's speed and cable payout rate, that the tension threshold would be breached in the sixth second. The system therefore generated a candidate state in advance.

[0042] These three mechanisms operate in parallel, outputting signals representing three candidate states: sudden disturbance, residual anomaly, and trend-based prediction anomaly. Each state is accompanied by a timestamp and state weight for subsequent fusion judgment by the slow module. This multi-criteria, time-sensitive design enables the fast fault identification module to achieve high responsiveness to complex sea conditions and various instability modes without increasing computing resources.

[0043] Specifically, the slow fault confirmation module is used to confirm the candidate fault state in a preset second time period, and the slow fault confirmation module includes.

[0044] The slow fault confirmation module operates on a longer time scale, with a processing cycle of, for example, 10 seconds. It emphasizes trend stability analysis and multi-cycle consistency judgment, and has the functions of denoising and false positive filtering. Three discrimination mechanisms are designed within this module: the first is to use the filtering trend judgment unit to identify whether there is a continuous trend of increasing tension or water depth to avoid mistaking instantaneous disturbances for faults; the second is to use the joint information entropy calculation unit to analyze the joint probability distribution of multiple navigation and equipment state variables. If the current system state deviates significantly from the state entropy during normal operation, it can be regarded as the system entering an abnormal area; the third is to introduce candidate state memory in the time dimension through the multi-cycle consistency voting mechanism, and vote on the suspected states that frequently jump in a short period of time to avoid triggering system responses due to isolated noise.

[0045] Specifically, the slow fault confirmation module includes a filtering trend judgment unit, a joint information entropy calculation unit, a multi-cycle consistency voting unit and a fault-tolerant control instruction output unit.

[0046] This module's processing revolves around three core logics: trend determination, information entropy deviation analysis, and multi-cycle consensus voting. These three independent yet collaborative judgment logics serve as a basis for triggering the system to switch to fault-tolerant control mode, demonstrating complementary and redundant design principles.

[0047] The filtering trend judgment unit is used to judge the continuous change trend of cable tension or water depth through median filtering and smooth derivative. Specifically, the filtering trend judgment unit includes a median filtering subunit, a smooth derivative calculation subunit, a trend consistency judgment subunit and a trend direction output subunit.

[0048] The median filter subunit is used to perform sliding window median filtering on the continuously collected cable tension signal or water depth signal to eliminate instantaneous fluctuations or abnormal noise; among them, the real-time parameters of ship navigation include ship speed and water depth, and the real-time parameters of equipment operation include cable tension and cable deployment speed.

[0049] As the first signal processing step in the slow fault confirmation module, the median filter subunit is designed to remove transient noise, abnormal spikes, and short-period disturbances from the actual tension signal T(t) and water depth signal d(t), providing a stable and continuous basic data sequence for subsequent derivative calculation and trend analysis. Considering that cable-laying vessels operating in marine environments are affected by waves, currents, and mechanical structure vibrations, the tension and water depth signals often contain a certain degree of high-frequency irregular disturbances. Direct trend analysis of the raw data is prone to misjudgment, so median filtering is necessary for preprocessing.

[0050] The median filter is a nonlinear filter. Its basic idea is to select the values of several sampling points in a symmetrical time window with the current time point as the center, sort these values by size, and take the median as the current output. It has better ability than the average filter in smoothing the signal while retaining the edge information. It is particularly suitable for processing signals containing spikes or pulse interference. Assuming that the median filter window length is N = 2k + 1, then for the current time t, the filter output T med The calculation formula for (t) is: T med (t)=median{T(t-kΔt),T(t-(k-1)Δt),…,T(t),…,T(t+kΔt)} Where Δt is the data sampling interval, which is usually set to 1 second.

[0051] In practical applications, the choice of median filter window length depends on the expected jitter period and the control system response time. In engineering verification, for tension signals, setting a window length of 5 (i.e., taking 2 values before and after) can effectively filter out small oscillations in the 0.2-0.5Hz frequency band. For water depth signals, due to the lower frequency of change, the window length can be appropriately increased to 7 or 9 to improve anti-interference performance. In an actual paving task along a gentle slope, the original tension signal fluctuated from 1800N to 2200N within 3 seconds and then quickly dropped back to 1900N, showing a typical spike shape. However, after median filtering, the output remained around 1900N at this moment, completely suppressing the transient spike and providing the correct background for trend judgment.

[0052] In addition to basic tension and water depth signals, the median filter subunit can also be expanded to filter tension derivatives or water depth change rates, serving as input for the subsequent trend direction determination submodule. To improve computational efficiency, the median filter can be implemented in the system as a sliding window update structure, avoiding repeated sorting operations and meeting real-time requirements on embedded processors.

[0053] The smoothed derivative calculation subunit is used to calculate the first-order derivative based on the filtered signal and apply a low-pass filter to extract the continuous trend component.

[0054] The smoothed derivative calculation subunit calculates the first-order derivative of the median-filtered cable tension or water depth signal and further smoothes it, extracting the main trend characteristics of the signal changes and eliminating high-frequency disturbances caused by sampling errors, mechanical vibration, or environmental fluctuations. This process serves as a bridge in the slow fault confirmation module, inheriting the stable signal output by the median filter subunit and providing directional basis for the trend consistency judgment subunit.

[0055] In actual implementation, the discrete calculation of the first-order derivative usually adopts a simple differential form. Assuming the current time is t and the sampling period is Δt, the first-order derivative of the tension signal T(t) can be expressed as: This differential operation is sensitive to changes in the input signal. If used directly, it will amplify high-frequency noise and drastic signal changes. Therefore, it must be combined with a low-pass filter for smoothing to avoid false trend judgments. A common method is first-order exponential weighted filtering, and the specific formula is as follows: Where α∈[0,1] is a smoothing factor that controls the mixing ratio of the current derivative response speed and the historical derivative value. In engineering experience, α is generally set between 0.3 and 0.5. In relatively stable sea conditions, a higher value can be used to enhance trend responsiveness. In highly volatile environments, the value should be appropriately reduced to improve noise suppression.

[0056] For the water depth signal d(t), the processing process is the same as that of the tension signal, which is first subjected to median filtering and then to difference and weighted smoothing to generate a smoothed derivative. The derivative units are N / s (tension) and m / s (water depth). The system will set a reasonable derivative amplitude judgment threshold (such as 0.3N / s or 0.05m / s) to determine whether the subsequent trend has physical significance and avoid false triggering of trend marks by small value changes.

[0057] For example, during a laying process along a slope descent, the cable tension gradually decreased, but was accompanied by random fluctuations in the middle. The original derivative sequence showed obvious high-frequency jitter. After smoothing, a set of outputs with negative derivative signs and a stable amplitude of -0.4N / s at three consecutive time points was generated. This stable negative derivative just reflects the physical state of the winch system: the winch system is steadily accelerating the cable release and the tension is continuously decreasing.

[0058] The trend consistency judgment subunit is used to judge whether the signs of the tension or water depth derivatives in the past at least two second time periods are consistent. If the signs are consistent and the amplitude changes are within the preset range, it is determined that there is a continuous trend. Specifically, the trend consistency judgment subunit is used to judge whether the change trend of the cable tension or water depth within a continuous period is consistent in direction based on the output result of the smoothed derivative calculation subunit, and use this as the basis for judging whether the current system is in a continuous change state. The core of this unit is to identify "continuous same-direction" trends and set a certain tolerance range to filter out non-structural changes such as short-term reversals and derivative critical value jitter, thereby enhancing the system's ability to identify trend evolution.

[0059] In this system, the trend consistency judgment is based on each second time period (for example, 10 seconds), and the smoothed derivative within this time period is or The sequence is analyzed for sign consistency. The specific process is as follows: within a time window, the derivative values are grouped chronologically, the sign of the derivative at each moment is recorded, and the length and number of consecutive segments with the same sign are counted. For example, if 10 derivative samples are collected within 10 seconds, of which 8 are positive and 2 are negative, the system can determine that the overall trend in this interval is "upward."

[0060] To avoid misjudgments due to critical fluctuations, the system introduces a minimum consistency ratio threshold. For example, the derivative sign consistency rate must reach 70% or more before the trend is considered valid. At the same time, to prevent small fluctuations in the value from affecting the judgment, the absolute value of the derivative must also exceed the set threshold ∈. For example, for the tension signal, ∈ can be set to 0.2N / s. If it is lower than this value, the derivative is considered "invalid" and does not participate in the trend consistency judgment. Combining the above rules, the following trend consistency function can be defined: Among them, n same is the number of samples with the same derivative sign, n total is the total number of samples, θ c is the symbol consistency threshold.

[0061] For example, in an upslope section of the seabed, cable tension has been steadily increasing over the past 30 seconds due to topographical changes. The system recorded derivative data for three consecutive 10-second periods, with over 90% of the derivatives in each period being positive and the average value exceeding 0.5 N / s. Based on this, the trend consistency determination subunit identifies the current state as "trending continuously upward" and transmits this status flag to the trend direction output subunit.

[0062] The trend direction output subunit outputs a trend direction signal and marks it as either an upward or downward trend. This subunit not only transitions the decision logic from Boolean judgment to directional description but also provides a priori basis for subsequent fault-tolerant control model selection (e.g., pre-triggering of an upslope or downslope model).

[0063] The trend direction is determined based on the average value and sign of the smoothed derivatives within the current time window. Once the trend consistency determination subunit confirms the existence of a clear trend within the current period, the system performs a weighted average of all valid derivatives to calculate the trend derivative mean: Where n represents the number of samples with consistent trends, is the derivative value at each valid moment. Subsequently, the system outputs a trend direction signal based on the sign of this mean: like Then output the "upward trend" mark; like Then output the "downward trend" mark; like The output is "no trend" or "stationary" flag.

[0064] Threshold∈ T The setting is used to prevent the situation where the average value of the derivative is too small but the sign is consistent from being misjudged as a trend. In practical applications, for cable tension, ∈ T It can be set to 0.3N / s, and for water depth signals it can be appropriately relaxed to 0.05m / s to take into account both trend sensitivity and judgment stability.

[0065] The trend direction signal is typically represented by a three-valued variable (rising, falling, and stable) and includes a timestamp for synchronization with the results of other slow-speed judgment modules. For example, during a field paving operation, the system determined that the tension derivative consistency was True over the past two second time periods, with an average derivative of +0.6 N / s. The trend direction output subunit outputted an "upward trend." Subsequently, the joint information entropy deviation result also reached a threshold. Ultimately, the slow-speed confirmation module determined the system status to be "trending abnormal tension increase," triggering fault-tolerant control.

[0066] The joint information entropy calculation unit is used to calculate the joint information entropy of real-time navigation parameters and equipment operation parameters, and determine the degree of deviation between the current system state and the preset normal entropy range. Specifically, the joint information entropy calculation unit assigns preset weight coefficients w1, w2, and w3 to the sudden disturbance state, model residual abnormal state, and trend prediction abnormal state, respectively, where w3>w2>w1; In three consecutive first time periods, the candidate states appearing in each period are multiplied by the corresponding weight value, and the total score value S is accumulated.fault ; Wherein, the preset second time period includes three consecutive first time periods; The slow confirmation module triggers the fault confirmation signal based on the following fuzzy rules: When S fault ≥θ S ; Or the joint information entropy offset exceeds the threshold ΔH>∈ H ; Or the trend judgment subunit judges that the trend in the same direction is abnormal for two consecutive periods; The system is judged to have entered a fault state and a fault-tolerant trigger instruction is output.

[0067] First, within each preset first time period (e.g., 1 second), the rapid fault identification module may output one or more candidate fault states. These states are uniformly categorized into three types: sudden disturbance state, model residual abnormal state, and trend-predicted abnormal state. The system assigns preset weight coefficients w1, w2, and w3 to these three states, where w3>w2>w1. For example, a typical setting may be: sudden disturbance state w1=1, model residual abnormal state w2=2, and trend-predicted abnormal state w3=3. The design of the weights reflects the credibility ranking of different candidate states in fault confirmation.

[0068] In a complete second time period (e.g. 10 seconds), the system records the candidate states in three consecutive first periods and executes the following cumulative scoring logic period by period. Whenever a candidate state is identified in the current period, its corresponding weight is added to the current period score. The system calculates the total score value S in a weighted sum manner fault : where n i represents the number of candidate states identified in the i-th cycle, w j The weight value corresponding to each candidate state. For example, if the system identified 1 trend prediction anomaly and 2 residual anomalies within three cycles, the total score is 3+2+2=7.

[0069] This scoring mechanism allows the gradual accumulation of system abnormality evidence through "weak abnormality superposition" when the joint information entropy has not exceeded the limit and the trend judgment is not continuous. As long as the score reaches the set threshold θ S (such as 6 or 7), the fault confirmation can be triggered. The judgment conditions in parallel also include: information entropy offset value ΔH = |H(t)-H ref |Exceeds the threshold ∈ H , or the trend judgment subunit outputs a consistent trend direction signal (up or down) for two consecutive cycles.

[0070] For example, in a certain task, the system identifies mutation disturbances and residual anomalies for two consecutive cycles, but the information entropy value is still within the normal range and the trend has not formed a consistent judgment. If the cumulative score value in these two cycles reaches or exceeds θ S =6, the system will still consider the current state to have sufficient fault probability and trigger fault-tolerant control.

[0071] The multi-cycle consistency voting unit is used to determine the number of times a candidate fault state occurs in multiple first time periods, and to confirm whether the fault state is triggered based on the trend judgment and entropy deviation results. The input of this unit is the candidate state tag stream output by the fast fault identification module every second. Each tag not only contains the state type (such as sudden disturbance, residual anomaly or trend prediction anomaly), but also comes with timestamp information. The system sets a sliding time window of fixed length (usually three consecutive first time periods, such as 1 second × 3), and performs statistical analysis on all candidate states in the window. For each time period t i ,If the system detects the existence of a candidate state, the cycle is considered to be a "valid cycle", otherwise it is an "empty cycle".

[0072] The statistical logic is divided into two layers: the first is the number of candidate states, which is the number of times N the candidate states are detected in three cycles. active ; The second is the candidate state type structure statistics. Each candidate state type is counted separately, and combined with its weight coefficients w1, w2, w3 set in the joint information entropy module, the weighted total score value within the time period is calculated: where n i represents the number of candidate states in the i-th cycle, is the weight of the j-th state in the cycle.

[0073] The system also models a voting mechanism for the "repetitiveness" of candidate states. Specifically, this is determined by whether the same candidate state is identified in at least two of the three cycles, forming a continuous or quasi-continuous pattern in time order. For example, if a trend-based forecast anomaly state appears in both cycles t1 and t3, the state is considered consistent across cycles and counted as a positive vote.

[0074] The voting mechanism is also integrated with the results of the trend direction output sub-unit: if the trend sub-unit outputs the same direction in two consecutive second cycles (e.g., both are upward trends), and three candidate states appear at the same time, the system automatically determines it as "fault consistency enhancement under trend guidance" and increases the current score value S fault Multiply by a correction gain factor (e.g. 1.2) to enter the fault-tolerant trigger logic with higher confidence.

[0075] For example, in a water depth fluctuation condition, the system continuously detects "model residual anomaly", "trend prediction anomaly" and "trend prediction anomaly" within three periods of 1 second, 2 seconds and 3 seconds, with corresponding weights of 2, 3 and 3 respectively, and the total score is 8. In addition, the trend judgment subunit has confirmed a continuous upward trend in the previous round, so the system will add 8×1.2=9.6 to the threshold θ S =6, and determines that the fault confirmation condition is met, and outputs a fault-tolerant trigger signal.

[0076] The fault-tolerant control command output unit is used to output a trigger command to the margin controller to switch to fault-tolerant control mode after a fault condition is confirmed. Once the slow fault confirmation module confirms that the system has entered a risky situation through entropy offset, trend consistency, or a scoring threshold, the unit immediately initiates the fault-tolerant process and outputs a control command to adjust the winch payout rate to ensure that the cable maintains moderate tension, bottoms out naturally, and avoids dangerous phenomena such as dragging or cable piling even in unsteady conditions.

[0077] In terms of control logic structure, the fault-tolerant control instruction output unit is composed of a multi-model control subunit, a sliding mode control subunit, and a finite time domain prediction subunit. Each of them constructs a control response strategy for different working conditions, and finally forms a fault-tolerant control instruction u in a fusion form. total (t), which is received and executed by the winch system.

[0078] First, the multi-model control subunit calculates the seabed elevation angle γ(t) of its location in real time based on the mapping relationship between the current ship position and the seabed path data, and compares it with the current cable laying angle α(t), and then selects the flat slope, uphill or downhill control model. The switching of the model is based on the following logic: if the seabed elevation angle is less than a certain positive threshold, it is considered a flat slope; if the elevation angle is large and the current laying angle is insufficient, it means that the cable may be suspended in the air, and an uphill model needs to be adopted; if the elevation angle is negative and the cable angle is too large, it means that there is a risk of cable stacking, and a downhill control model needs to be adopted. In each model, the system uses variables such as tension, water depth, and ship speed to calculate the basic control output u model (t),u model (t) refers to the basic control instruction output by the currently selected control model based on the current ship speed, water depth and terrain conditions. It is used to determine the release speed or tension target of the winch and is the most basic reference control quantity in the entire fault-tolerant control system. model (t) is generally in the form of a linear combination or margin-time ratio, reflecting the prior response of the ship to tension adjustment under different terrains.

[0079] Then, the sliding mode control subunit monitors the error between the actual cable length and the theoretical target length in real time, e(t) = L actual (t)-L target (t), and construct the sliding surface function Where λ is the slope coefficient of the surface, which indicates the sensitivity of the control system to error feedback. The controller generates the sliding mode compensation according to the sign of the sliding surface: u sm (t) = -K s ·sign(s(t)) To prevent control oscillations caused by the sign function, the system uses boundary layer methods (such as tanh approximation) or super-helical sliding mode to suppress chattering, making the output smoother and more stable. The main function of this part is to provide rapid compensation for margin deviation.

[0080] At the same time, the finite-time prediction subunit predicts the margin error trajectory e(t+1), e(t+2), … within the next N steps based on state-space modeling or adaptive identification model, and solves the control variable using the following rolling optimization objective function: Among them, p is the penalty coefficient used to balance the control accuracy and energy consumption, and Δu is the rate of change of the control quantity. The system uses the output of the first few steps in the optimal adjustment sequence obtained as the feedforward quantity u pred (t) is transmitted to the sliding mode control subunit for fusion. This structure ensures that the system can adjust the control instructions in advance under trend tension changes to avoid response lag.

[0081] Finally, the three control outputs are fused according to the set weights α, β, and γ to form the total control output in fault-tolerant mode: u total (t) = α·u model (t)+β·u sm (t)+γ·u pred (t) Here, α + β + γ = 1. The weights can be preset to fixed values (e.g., 0.4:0.4:0.2) or dynamically adjusted based on parameters such as the current environmental fluctuation amplitude or trend slope. For example, the sliding mode weight can be increased when sudden disturbances dominate, while the predictive control weight can be increased when trending anomalies dominate, to maximize adaptability to complex marine engineering environments.

[0082] Specifically, the multi-model control subunit is used to execute the following steps S101-S104.

[0083] S101. Calculate the elevation angle of the current ship position corresponding to the seabed terrain based on the input seabed path data; S102. Select a flat slope model, an upslope model, or a downslope model based on the relative relationship between the seabed elevation angle and the current cable laying angle: S103. Load predefined tension-vehicle speed control parameters and margin release rate parameters according to the selected model; S104. Based on the current ship speed v ship (t), water depth d(t) and the control parameters of the selected model, calculate the model control quantity u model (t); wherein, the flat slope model corresponds to an elevation angle γ(t) less than a first set threshold; The uphill model corresponds to the elevation angle γ(t) being greater than the second set threshold, and the real-time laying angle α(t) being less than the elevation angle γ(t), and is selected; The downhill model corresponds to the elevation angle γ(t) being less than the negative threshold, and the real-time laying angle α(t) being greater than the elevation angle γ(t).

[0084] The unit first relies on the pre-processed seabed path data. The path data comes from the seabed DEM (digital elevation model) constructed by multi-beam bathymetry, side-scan sonar or ROV mapping before the laying operation, and has been refined into a sequence of track points in the planning stage {P i Each point contains elements such as longitude and latitude, depth, elevation angle, and path tangent direction. During the laying process, the system uses the real-time ship position P(t) to correspond to the nearest point on the path and obtains the seabed elevation angle γ(t) at that point.

[0085] The elevation angle γ(t) is defined as the average slope between the current path point and several path points before and after it: Where d(t) is the water depth, and Δs is the differential distance between the front and rear paths. This angle reflects the seabed trend before and after the current ship position: positive values indicate an upslope, negative values indicate a downslope, and values close to zero indicate a flat or gradual slope.

[0086] Next, the system combines the elevation angle with the cable laying angle α(t). This angle, α(t), refers to the average slope of the cable from the winch outlet to the water surface or bottom, and is typically estimated by the tension solver or the visual / inertial tracking module. By comparing the relative magnitudes of α(t) and γ(t), the system can determine whether the cable is at risk of "hanging in the air," "looping," or "excessively touching the bottom," and then select the appropriate model: If |γ(t)|<θ1, that is, the slope is less than the set threshold (such as 2°), the flat slope model is selected; If γ(t)>θ2 and α(t)<γ(t), that is, the cable laying angle is not enough to touch the bottom, the uphill model is selected; If γ(t)<-θ3 and α(t)>γ(t), there is a risk of over-release of the cable, and the downslope model is selected.

[0087] Each model contains a set of corresponding control parameters, such as tension-speed adjustment coefficient, margin release rate function, etc. The model output is usually in the form of: u model (t) = k v ·vship (t)+k d d(t)+k c or: Where: v ship (t) is the current ship speed; d(t) is the current water depth; k v ,k d ,k c is the empirical coefficient under the current model.

[0088] Taking an actual laying task as an example, when entering a continuous downhill section, the system judged that the elevation angle reached -6.5°, while the laying angle was only -3.0°, which obviously caused an over-release risk. Therefore, it automatically switched to the downhill model and the system dynamically adjusted k d A positive value increases the control sensitivity to water depth changes, thereby compressing excessive cable release and effectively avoiding underwater loops.

[0089] Optionally, the model control quantity u model (t) is calculated as a linear combination or as a ratio based on the target residual release rate; Linear combination form: u model (t) = k v ·v ship (t)+k d d(t)+k c , where k v ,k d ,k c is the scale parameter corresponding to the model; The ratio form of the release rate based on the target margin is: Among them, L target (t) is the current target cable length, T release (t) is the release period.

[0090] Specifically, the sliding mode control subunit is used to execute the following steps S201-S205.

[0091] S201. Real-time calculation of the margin error between the actual cable length and the preset target length e(t) = L actual (t)-L target (t); S202. Constructing sliding surface function based on residual error and its derivative Where λ is the slope parameter of the sliding surface; S203. Output the control compensation u based on the sliding surface function. sm (t) = -K s sign(s(t)), where K sis the sliding mode gain coefficient, sign(·) is the sign function; S204. Perform boundary layer processing on the sign function sign(·) or use a super-helical sliding mode algorithm to alleviate control signal oscillation; S205. Set the sliding mode control compensation u sm (t) and model control quantity u model (t) Output u after fusion total (t) = α·u model (t)+β·u sm (t)+γ·u pred (t), for the winch speed control system to call, where u pred (t) is the feedforward control quantity output by the finite time domain prediction subunit, α, β, γ are the control quantity fusion weight coefficients, satisfying α+β+γ=1, which are set according to empirical values or adjusted by the system operation status.

[0092] The sliding mode control subunit first obtains the following two core inputs from the system measurement module: 1. The actual cable length L actual (t), obtained by a counter or encoder; 2. Current target cable length L target (t), provided by the control planning or path planning module.

[0093] The system calculates the current margin error based on this: e(t)=L actual (t)-L target (t) And estimate its derivative through the first-order difference form: To construct the sliding surface function, a slope factor λ>0 is introduced and the sliding surface is defined as: The physical meaning of the sliding surface s(t) is to combine the margin error and its changing trend to form a "comprehensive deviation measure" of the cable status.

[0094] Next, the control compensation term is calculated based on the sliding surface value: u sm (t) = -K s ·sign(s(t)) Among them, K s >0 is the sliding mode gain coefficient, and sign(·) is the sign function, which indicates the direction of error correction. Because the sign function is discontinuous at zero, using it directly can lead to chattering, i.e., high-frequency switching of control commands. In practical applications, one of the following two methods is usually used for smoothing: Boundary layer method: replace the sign function with the hyperbolic tangent function tanh(s / ε), where ε is the boundary layer thickness; Super spiral sliding mode method: Introducing a continuous but high-order convergent sliding mode law, such as u sm (t) = -K s ·|s(t)| 1 / 2 ·sign(s(t)).

[0095] This type of processing can significantly reduce control oscillation without sacrificing control response speed, extend the service life of the mechanical actuator structure, and improve system stability.

[0096] Finally, the sliding mode control compensation is weighted and fused with the multi-model control output and the finite time domain prediction feedforward to form the final control instruction: u total (t) = α·u model (t)+β·u sm (t)+γ·u pred (t) The fusion coefficient, α+β+γ=1, can be statically set or dynamically adjusted based on actual operating conditions (such as the intensity of external disturbances and the significance of trend characteristics). For example, when operating in areas with severe waves, the weight of β can be increased to make sliding mode control dominant and enhance system response speed; while in environments with slowly changing trend tension, γ can be increased to enable predictive control to play a greater role.

[0097] Specifically, the limited time domain prediction subunit is used to perform the following steps S301-S304.

[0098] S301. Establishing a prediction margin error trajectory equation based on a state space model or an identification model, wherein the state space model assumes that the system state change has short-term continuity and is suitable for control prediction in a slowly varying tension environment; S302. Set the rolling minimization objective function within the prediction time domain Among them, e(t+k) is the prediction margin error, Δu(t+k) is the winch speed change, and ρ is the control balance coefficient; S303. The optimal first few steps of the predicted future winch speed adjustment {u(t+1), u(t+2), ..., u(t+N)} are used as the feedforward control value u pred (t) Output to the sliding mode control subunit; S304. Dynamically adjust the prediction time domain length N according to the system load status and response characteristics.

[0099] The input signals of the finite time domain prediction subunit include: The current system state vector x(t) usually contains the margin error e(t), its first-order derivative and influencing variables such as ship speed, tension, and water depth; Current winch speed control value u(t-1); The state-control history over a period of time is used for state prediction modeling.

[0100] First, the unit uses state space modeling or identification model to establish the prediction margin error trajectory equation. Taking the linear discrete state space as an example, the model form is: x(t+1)=Ax(t)+Bu(t) e(t)=Cx(t) Where A, B, and C are the state transition matrix, control matrix, and observation matrix. These are updated offline or online using recursive least squares (RLS) or subspace identification based on historical system operating data. If the system exhibits significant nonlinearities, a nonlinear ARX (Auto-Regressive with eXogenous Input) model or an extended state estimator (EKF) based on a Kalman filter can also be used.

[0101] Next, the system sets the rolling horizon optimization control objective function based on the above model. Assuming that the prediction horizon length is N, the objective function is defined as: in: e(t+k) is the residual error of the predicted kth step; Δu(t+k)=u(t+k)-u(t+k-1); ρ>0 is the control energy consumption suppression weight, which adjusts the smoothness of the control command change; u(t+k) is the future winch speed adjustment to be optimized.

[0102] In this objective function, quadratic optimization is used for both the control change term and the error term, ensuring that the overall objective function maintains convexity. The optimization goal is to minimize the prediction margin error while avoiding overly aggressive control changes. This objective function can be solved using quadratic programming (QP). If computing resources are limited, gradient descent can also be used for an approximate solution.

[0103] In the obtained predictive control sequence {u(t+1),u(t+2),…,u(t+N)}, the first M steps (such as 1 to 3 steps) are usually taken to form the feedforward control quantity u pred In the actual system, this value acts as a trend guide for sliding mode control. That is, before the system has obvious errors, it can predict that the tension is about to be too high or too low, so the winch speed is slightly increased or decreased in advance to prevent the system from entering the error diffusion state.

[0104] For example, when a cable laying vessel is about to enter a downhill area, although the current tension is still at a normal value, the state prediction model, based on the continuous downhill trend, water depth changes and the linkage evolution of the current ship speed, determines that the tension will drop below the lower limit in 5 seconds. The system outputs a control value u in advance to slightly reduce the speed. pred (t)<0, the model control and sliding mode feedback are combined to smoothly transition the slope section and avoid cable looping caused by a sudden drop in tension.

[0105] Finally, the prediction subunit dynamically adjusts the prediction time domain length N, adapting the calculated load and control accuracy based on indicators such as system fluctuation characteristics, response delay, and the current load change rate. In highly disturbed environments, N is shortened to speed up response; in slowly changing trends, N is extended to improve foresight.

[0106] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A fault identification and fault-tolerant control device for cable margin control on an optical cable laying vessel, characterized in that: The device includes: A fast fault identification module is used to collect real-time navigation parameters of the vessel and real-time operation parameters of the equipment in a preset first time period, and generate candidate fault status signals based on the second-order derivative changes of tension and water depth, the sudden jump of the control model residual error, and the tension change trend prediction; A slow fault confirmation module is configured to confirm the candidate fault state in a preset second time period, the slow fault confirmation module comprising: Filter trend judgment unit, used to judge the continuous change trend of cable tension or water depth through median filtering and smooth derivative; A joint information entropy calculation unit is used to calculate the joint information entropy of real-time navigation parameters and real-time equipment operation parameters, and to determine the degree of deviation between the current system state and the preset normal entropy range; a multi-cycle consistency voting unit, configured to determine the number of occurrences of a candidate fault state in a plurality of said first time periods, and to determine whether a fault state is triggered based on trend determination and entropy deviation results; The fault-tolerant control instruction output unit is used to output a trigger instruction to the margin controller after the fault state is confirmed, so as to switch into the fault-tolerant control mode.

2. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 1, characterized in that: The fast fault identification module generates candidate fault states based on the following mutation identification conditions: like or It is determined to be a sudden disturbance state; where T(t) is the cable tension, D(t) is the water depth, δ T and δ D is the threshold; If the residual r T =|T measured -T model |>μ r +2σ r , then it is determined to be an abnormal state of the model residual; where T measured (t) is the tension measurement value, T model (t) is the predicted tension value; μ r is the residual mean, σ r is the standard deviation; If the tension exceeds the set threshold within 5 seconds based on Kalman filtering extrapolation, it is determined to be a trend prediction abnormal state.

3. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 2, characterized in that: The filtering trend judgment unit includes: The median filter subunit is used to perform sliding window median filtering on the continuously collected cable tension signal or water depth signal to eliminate instantaneous fluctuations or abnormal noise. The real-time parameters of the ship's navigation include ship speed and water depth, and the real-time parameters of the equipment's operation include cable tension and cable deployment speed. a smoothed derivative calculation subunit for calculating the first-order derivative based on the filtered signal and applying a low-pass filter to extract the continuous trend component; A trend consistency judgment subunit is used to judge whether the signs of the tension or water depth derivative are consistent in at least two second time periods in the past. If the signs are consistent and the amplitude change is within a preset range, it is determined that a continuous trend exists; The trend direction output subunit is used to output the trend direction signal and mark it as an upward trend or a downward trend.

4. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 3, characterized in that: The joint information entropy calculation unit assigns preset weight coefficients w1, w2, and w3 to the sudden disturbance state, the model residual abnormal state, and the trend prediction abnormal state, respectively, where w3>w2>w1; In three consecutive first time periods, the candidate states appearing in each period are multiplied by the corresponding weight value, and the total score value S is accumulated. fault ; Wherein, the preset second time period includes three consecutive first time periods; The slow confirmation module triggers the fault confirmation signal based on the following fuzzy rules: When S fault ≥θ S ; where θ S is the threshold; Or the joint information entropy offset exceeds the threshold ΔH>∈ H ; where ∈ H is the threshold; Or the trend judgment subunit judges that the trend in the same direction is abnormal for two consecutive periods; The system is judged to have entered a fault state and a fault-tolerant trigger instruction is output.

5. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 4, characterized in that: The fault-tolerant control instruction output unit includes: A multi-model control subunit, used to switch between preset control models according to seabed topography characteristics, including flat slope model, upslope model and downslope model; The sliding mode control subunit is used to construct a sliding mode surface for the cable margin error and perform disturbance compensation on the winch control command in real time. The finite time domain prediction subunit is used to predict the future margin error trajectory based on the current state and provide feedforward adjustment to the sliding mode control subunit.

6. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 5, characterized in that: The multi-model control subunit is used to perform the following steps: Calculate the elevation angle of the seabed terrain corresponding to the current ship position based on the input seabed path data; Select the flat slope model, upslope model or downslope model based on the relative relationship between the seabed elevation angle and the current cable laying angle: load the predefined tension-vessel speed control parameters and margin release rate parameters according to the selected model; Based on the current ship speed v ship (t), water depth d(t) and the control parameters of the selected model, calculate the model control quantity u model (t); Among them, the flat slope model corresponds to the elevation angle γ(t) being less than the first set threshold; The uphill model corresponds to the elevation angle γ(t) being greater than the second set threshold, and the real-time laying angle α(t) being less than the elevation angle γ(t), and is selected; The downhill model corresponds to the elevation angle γ(t) being less than the negative threshold, and the real-time laying angle α(t) being greater than the elevation angle γ(t).

7. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 6, characterized in that: The model control quantity u model (t) is calculated as a linear combination or as a ratio based on the target residual release rate; Linear combination form: u model (t) = k v ·v ship (t)+k d d(t)+k c , where k v ,k d ,k c is the scale parameter corresponding to the model; The ratio form of the release rate based on the target margin is: Among them, L target (t) is the current target cable length, T release (t) is the release period.

8. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 7, characterized in that: The sliding mode control subunit is used to perform the following steps: Real-time calculation of the margin error e(t)=L between the actual cable length and the preset target length actual (t)-L target (t); Constructing sliding surface function based on margin error and its derivative Where λ is the sliding surface slope parameter; Output control compensation u based on sliding surface function sm (t) = -K s sign(s(t)), where K s is the sliding mode gain coefficient, sign(·) is the sign function; Perform boundary layer processing on the sign function sign(·) or use a super-helical sliding mode algorithm to alleviate control signal oscillations. The sliding mode control compensation u sm (t) and model control quantity u model (t) Output u after fusion total (t) = α·u model (t)+β·u sm (t)+γ·u pred (t), for the winch speed control system to call, where u pred (t) is the feedforward control quantity output by the finite time domain prediction subunit, α, β, γ are the control quantity fusion weight coefficients, satisfying α+β+γ=1, which are set according to empirical values or adjusted by the system operation status.

9. The fault identification and fault-tolerant control device for cable margin control of an optical cable laying vessel according to claim 8, characterized in that: The limited time domain prediction subunit is configured to perform the following steps: Establishing a prediction margin error trajectory equation based on a state space model or an identification model, wherein the state space model assumes that the system state change has short-term continuity and is suitable for control prediction in a slowly varying tension environment; Set the rolling minimization objective function within the prediction time domain Among them, e(t+k) is the prediction margin error, Δu(t+k) is the winch speed change, and ρ is the control balance coefficient; The optimal first few steps of the predicted future winch speed adjustment {u(t+1), u(t+2), ..., u(t+N)} are used as the feedforward control quantity u pred (t) Output to the sliding mode control subunit; The prediction time domain length N is dynamically adjusted according to the system load status and response characteristics.

Images