Whole-ship equipment health monitoring and early warning system for inland river ship ferry based on Internet of Things

Abstract

The invention discloses an inland river ship ferry whole ship equipment health monitoring and early warning system based on the Internet of Things, and relates to the technical field of ship equipment monitoring. A dynamic dual-threshold coupling mechanism is adopted, and the method comprises the steps that a noise estimation module combines a navigation state and a historical working condition library to generate a fundamental frequency floating range vector and a harmonic ratio threshold, and the problem of signal distortion caused by vibration fundamental frequency drift is solved; the collaborative analysis module constructs a three-dimensional incidence matrix of temperature gradient, envelope energy and current distortion rate, expands a harmonic ratio threshold into a dynamic window, and realizes multi-parameter synchronous out-of-limit judgment; and the health scoring module quantifies the equipment degradation degree through dynamic regular comparison of a waveform envelope line and a historical curve. And when the incidence matrix is in a continuous five-period super-dynamic window, outputting a micro-fault feature code, and mapping the micro-fault feature code to an equipment topology library in combination with the health score to generate an operation instruction. The fault recognition rate under the variable working conditions of the ship is improved, and predictive maintenance of early faults such as bearing abrasion and electric overload is supported.

Description

Technical Field

[0001] The present invention relates to the technical field of ship equipment monitoring, and in particular to an Internet of Things-based inland river ferry ship equipment health monitoring and early warning system. Background Art

[0002] As a key hub for water transportation, inland ferries require equipment health monitoring systems that are crucial for ensuring navigation safety and operational efficiency. Current ship health monitoring systems typically employ fixed-threshold alarm mechanisms and single-parameter independent analysis models, which offer a degree of reliability in steady-state scenarios, such as ocean-going vessels. These systems employ preset vibration amplitude thresholds or current distortion rate limits, combined with single-dimensional data excursions, to trigger alarms, providing basic protection against significant equipment failures.

[0003] However, the high-frequency speed and load changes, and start-stop operations unique to inland ferries (such as sudden draft changes after berthing and unloading, and sudden acceleration due to adverse currents) cause the fundamental frequency and harmonic components of equipment vibration to fluctuate dramatically depending on the operating conditions. Fixed thresholds cannot distinguish between normal operating fluctuations and early signs of faults. For example, when the main engine speed increases from 800 RPM to 1200 RPM, the fundamental frequency vibration amplitude of the drive shaft increases by 40%. Traditional systems mistakenly interpret this as bearing wear, leading to frequent false alarms. Furthermore, the forced increase in alarm thresholds to suppress false alarms masks weakly correlated micro-faults with multiple parameters (such as a 3% increase in vibration envelope energy, a slight increase of 0.5°C / min in temperature gradient, and a 0.8% fluctuation in current distortion rate during initial bearing spalling). This results in a high rate of missed early fault detection and frequent unexpected downtime for inland ferries.

[0004] There is an urgent need in this field to build a health monitoring mechanism that adapts to the dynamic operating characteristics of inland ferry ships. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides an inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things.

[0006] In order to achieve the above object, the technical solution of the present invention is as follows:

[0007] The present invention discloses an inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things, comprising:

[0008] The sensor group module is used to collect temperature values, vibration time domain signals and current harmonic distortion rate of the ship;

[0009] The noise estimation module is used to receive real-time navigation status data from the ship control terminal and generate the fundamental frequency floating range vector and the first harmonic ratio threshold based on the pre-stored historical operating condition database;

[0010] a parameter adjustment module, configured to determine whether the vibration time domain signal is within the fundamental frequency floating range vector, and if so, to time align the vibration time domain signal and output time alignment data;

[0011] An envelope extraction module, configured to perform bandpass filtering and Hilbert transform on the time-aligned data to generate a current waveform envelope;

[0012] A health scoring module is used to compare the current waveform envelope with a pre-stored historical health curve using a dynamic time warping algorithm to output a health score;

[0013] A collaborative analysis module constructs a correlation matrix M based on the verified temperature value, the time-aligned data, and the current harmonic distortion rate; expands the first harmonic ratio threshold into a second dynamic threshold window; the second dynamic threshold window = {current harmonic window, envelope energy window, temperature gradient window};

[0014] When the correlation matrix M synchronously exceeds the second dynamic threshold window within five consecutive sampling periods, outputting a micro-fault feature code;

[0015] The early warning execution module is used to map the micro-fault feature code and the health score to a pre-stored equipment topology relationship library, generate and output an operation instruction to the ship control terminal.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. This invention significantly improves the system's anti-interference capabilities under variable vessel operating conditions through a fundamental frequency fluctuation range mechanism driven by navigational conditions. When a vessel encounters complex scenarios such as fluctuating water currents and sudden changes in load, the equipment's vibration characteristic frequency bands adaptively adjust within tolerances, effectively preventing valid fault signals from being misfiltered or noise interference from being misidentified. This mechanism safeguards the integrity of vibration signals under challenging operating conditions such as sharp turns and shallow water navigation.

[0018] 2. This invention constructs a three-dimensional correlation model of temperature, current, and vibration energy, achieving decoupled judgment of multiple fault characteristics through physical constraints. This model accurately captures the multi-parameter co-evolution of micro-faults such as early bearing wear, while simultaneously eliminating false alarms caused by transient jumps in a single parameter. This significantly improves the reliability of early fault identification, enabling the system to maintain high-precision warnings even in complex anomaly scenarios such as strong electrical interference or localized overheating.

[0019] 3. This invention utilizes a health assessment system based on personalized equipment history curves, surpassing the insensitivity of traditional static thresholds to equipment performance degradation. By quantifying the timing deviation of equipment waveform characteristics, the system can proactively identify equipment in the slow degradation phase and dynamically adjust maintenance priorities based on the rate of degradation. This capability provides key support for predictive operations and maintenance for ferry fleets, effectively avoiding unexpected accidents such as electrical burnout and mechanical failure, and significantly extending the service life of critical equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them:

[0021] Figure 1 This is a system module connection diagram of the present invention;

[0022] Figure 2 A flow chart of the steps for the system of the present invention to work;

[0023] Figure 3 Module flow chart of the self-checking unit of the present invention;

[0024] Figure 4 A flowchart of the noise tolerance adaptive adjustment of the present invention;

[0025] Figure 5 is a flowchart of the steps of the collaborative analysis module of the present invention;

[0026] Figure 6 This is a flowchart of the steps of the health scoring module of the present invention. DETAILED DESCRIPTION

[0027] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0028] Existing technologies for inland ferry equipment health monitoring generally rely on fixed-threshold alarm mechanisms and single-parameter independent analysis models. While these systems can detect significant faults under steady-state conditions, they struggle to distinguish between normal operating fluctuations and early signs of faults when faced with the high-frequency speed and load variations, and start-stop operations unique to inland vessels. For example, a sudden change in main engine speed, which causes a normal increase in fundamental frequency vibration amplitude, can be misidentified by traditional systems as bearing wear. Furthermore, micro-faults with weak correlations among multiple parameters can be missed due to conservative threshold settings, leading to frequent unexpected downtime.

[0029] To address these issues, a monitoring mechanism adaptable to dynamic operating conditions is needed. This mechanism leverages early subtle parameter trend changes and data interaction to predictively identify and prevent failures in critical ferry equipment before they become apparent, thus avoiding sudden stoppages and equipment damage. The inventors discovered that traditional single-dimensional thresholds cannot reflect the coordinated changes in multiple parameters, and that early failures often manifest as subtle abnormal correlations between temperature, vibration, and current parameters. Based on this, they proposed capturing microfault characteristics through dynamic threshold windows and a multi-parameter correlation matrix, combining this with device topology to achieve precise positioning.

[0030] After introducing the basic concept of the present invention, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0031] Example:

[0032] like Figure 1 As shown in the figure, the IoT-based inland river ferry ship equipment health monitoring and early warning system includes:

[0033] The sensor group module is used to collect temperature values, vibration time domain signals and current harmonic distortion rates of the ship; the sensor group module deploys pluggable modular multi-parameter sensor units at key locations such as the engine, propeller, and cabin electrical equipment.

[0034] The noise estimation module is used to receive real-time navigation status data from the ship control terminal and, in combination with a pre-stored historical operating condition database, generate a fundamental frequency floating range vector and a first harmonic ratio threshold. The fundamental frequency floating range vector is a reasonable fluctuation range of the vibration fundamental frequency that is dynamically adjusted based on real-time operating conditions and historical data, and is used to distinguish normal operating condition fluctuations from abnormal frequency offsets.

[0035] The parameter adjustment module is used to determine whether the vibration time domain signal is within the fundamental frequency floating range vector. If so, the vibration time domain signal is time-aligned and the time-aligned data is output; the time-aligned data is the vibration time domain data after the signal acquisition time deviation is eliminated to ensure the consistency of the time base for subsequent envelope analysis.

[0036] Envelope extraction module, used to perform bandpass filtering and Hilbert transform on the time-aligned data to generate the current waveform envelope;

[0037] The health scoring module is used to compare the current waveform envelope with the pre-stored historical health curves using a dynamic time warping algorithm to output a health score. The module also outputs a health score value, Z, by comparing the root mean square error (ΔW) and error change rate (dV / dt) between the current waveform envelope and the historical health curves. The health score value, Z, is calculated as follows: Z = 100[α × |ΔW| + β × dV / dt], where α and β are device weight coefficients.

[0038] The collaborative analysis module constructs a correlation matrix M based on the verified temperature values, time-aligned data, and current harmonic distortion rate; expands the first harmonic ratio threshold into a second dynamic threshold window; the second dynamic threshold window = {current harmonic window, envelope energy window, temperature gradient window};

[0039] The correlation matrix M is a three-dimensional data matrix that integrates temperature gradient, envelope energy ratio, and current harmonic distortion rate, and is used to characterize the comprehensive characteristics of the device status. The second dynamic threshold window is a multidimensional threshold space composed of current harmonics, envelope energy, and temperature gradient. It is formed by expanding the first harmonic ratio threshold through a dynamic window conversion unit and is used to simultaneously monitor multi-parameter coordinated over-limit behavior.

[0040] When the correlation matrix M exceeds the second dynamic threshold window synchronously within 5 consecutive sampling periods, a micro-fault feature code is output;

[0041] The early warning execution module is used to map the micro-fault feature code and health score to the pre-stored equipment topology relationship library, generate and output operation instructions to the ship control terminal.

[0042] like Figure 2 As shown, the working principle of this application is: after the sensor group module collects multi-source data in real time, the noise estimation module matches the historical operating conditions according to the navigation status to generate a dynamic fundamental frequency range. The parameter adjustment module selects the vibration signals that meet the fundamental frequency range for time alignment to eliminate sampling deviations. The envelope extraction module extracts the waveform envelope through bandpass filtering and Hilbert transform, and the health scoring module performs dynamic time regularization comparison with the historical health curve to quantify the health of the equipment. The collaborative analysis module constructs the temperature, vibration, and current data into a correlation matrix. When the matrix synchronously exceeds the dynamic threshold window within a continuous sampling period, the micro-fault feature code is triggered. The early warning execution module maps the feature code to a specific device node in combination with the device topology relationship to generate a start-stop or power adjustment instruction.

[0043] In the specific implementation, different variant examples are performed:

[0044] Variant 1: Small passenger ferry application

[0045] When deploying the system on the 80-passenger ferry "Yuhang 08" (hull length 28 meters):

[0046] Environmental characteristics: Sharp bend section of the Yangtze River tributary (minimum curvature radius 150 meters);

[0047] Fundamental frequency floating range vector generation rules: When the hull length is less than 30m, the upper limit of the fundamental frequency floating range vector = standard value × 1.25, and the lower limit = standard value × 0.9;

[0048] By adjusting the range of the fundamental frequency floating range vector, the false alarm rate under sharp turning conditions can be reduced.

[0049] Variant 2: Heavy-duty cargo ship application

[0050] When deploying the system on the 5,000-ton cargo ship "Jiangyun Heavy Loading No. 1" (draft depth 5.8 meters):

[0051] Environmental characteristics: Deepwater section of the Three Gorges Reservoir (water depth > 50 meters);

[0052] When the draft of the cargo ship is greater than 4 meters, linear compensation is performed, and the load compensation coefficient = min(1+0.02×(draft-4),1.3);

[0053] The first harmonic ratio threshold K = basic threshold × load compensation coefficient;

[0054] By adjusting the value of the first harmonic ratio threshold K, the micro-fault recognition rate under heavy-load conditions can be improved.

[0055] This solution uses the fundamental frequency floating range vector and dynamic threshold window to achieve adaptive threshold adjustment, the correlation matrix and synchronous over-limit detection mechanism to identify complex fault modes, and combines health scores with fault location coordinates to achieve precise control. Through the above technical solutions, this application effectively distinguishes between normal fluctuations and early faults under the dynamic working conditions of inland ships, reduces the false alarm rate, and improves the sensitivity of micro-fault detection. Through multi-parameter collaborative analysis and equipment topology mapping, accurate fault location and targeted control instruction generation are achieved, avoiding the risk of sudden downtime caused by conservative threshold settings in traditional systems.

[0056] like Figure 3 As shown, the present application further proposes that each sensor unit of the sensor group module includes a self-checking unit, and the self-checking unit specifically performs the following steps:

[0057] During a preset acquisition cycle, the sensor group module performs drift detection on the reference voltage of each sensor unit and generates a voltage stability mark. The reference voltage drift detection measures the reference voltage fluctuation range of the sensor unit in real time through a voltage monitoring circuit and compares it with the reference voltage source using an analog-to-digital converter. This is used to identify sensor signal acquisition deviations caused by circuit aging or environmental interference.

[0058] The currently collected temperature value of the sensor is compared with the temperature value of the adjacent sensor to generate a temperature difference, and the temperature difference is compared with the preset tolerance δ. If the temperature difference exceeds the preset tolerance δ three times in a row, a sensor failure code is generated. The preset tolerance δ refers to the maximum allowable temperature difference between adjacent sensor nodes. For example, the preset tolerance δ can be dynamically set based on the historical extreme temperature of the equipment operation to determine whether the sensor has abnormal temperature measurement due to local damage or loose installation.

[0059] The voltage stability mark and sensor failure code are respectively embedded in the collected current harmonic distortion rate and the verified temperature value and transmitted to the collaborative analysis module; the collaborative analysis module receives the embedded mark and failure code. When a sensor failure code is detected, the temperature value of the corresponding node is ignored and the temperature value is collected again.

[0060] In specific implementation, the voltage drift detection unit samples the reference voltage at a frequency of 32 times / cycle and calculates the difference between the reference voltage and the nominal value. Standard deviation .when When the 0.05V threshold (set according to the ISO17025 calibration standard) is exceeded, a voltage instability flag is generated, indicating that power supply fluctuations or sensor aging may have caused signal distortion.

[0061] The adjacent temperature difference verification unit adopts a spatial thermal field modeling strategy to obtain the temperature values ​​of the target sensor and the three nearest adjacent nodes (Ttgt, Tn1, Tn2, Tn3);

[0062] Calculate the normalized temperature gradient: ΔTmax=max(|Ttgt-Tni|) / di;

[0063] Where di is the sensor spacing (meters). When di ≤ 1 meter, di = 1.0. When di > 1 meter, di = actual distance.

[0064] If ΔTmax is greater than δ (δ is preset according to the equipment type: δ = 0.4°C / m for power equipment and δ = 0.6°C / m for auxiliary equipment) for three consecutive times, a sensor failure code is generated, indicating physical damage or installation abnormality.

[0065] The data encapsulation unit encapsulates the voltage mark, failure code, original temperature value, and current distortion rate into a data frame with CRC check according to the protocol. After the collaborative analysis module parses the data frame, if a failure code is detected, it immediately starts the adjacent node temperature interpolation compensation: take the arithmetic mean of the temperatures of the three adjacent points, and superimpose the historical baseline temperature difference of the node. (in, is the difference between the historical mean temperature of the target node and the historical mean temperature of the neighboring nodes), and generates the compensation temperature value T compensated Replace the original data.

[0066] This solution uses a dual verification mechanism of reference voltage drift and neighborhood temperature difference, combined with a rule for determining the number of consecutive deviations, to accurately distinguish between normal operating fluctuations and sensor hardware failures. Through the above technical solution, this application can effectively identify sensor data anomalies caused by circuit aging, mechanical vibration, or installation position offset, and prevent erroneous data from entering the subsequent analysis process. By dynamically shielding the temperature data of failed sensors, misjudgments caused by single-node failures can be prevented, and the accuracy of multi-parameter collaborative analysis can be improved. At the same time, the embedded transmission mechanism of the voltage stability mark provides a quantitative basis for data credibility assessment, which is beneficial for the system to maintain monitoring continuity under complex working conditions.

[0067] like Figure 4 As shown, the present application further proposes that the noise estimation module performs adaptive adjustment of noise tolerance, and the specific steps include:

[0068] Based on the main engine speed and load rate in the real-time navigation status data, the noise fundamental frequency floating model in the historical operating condition database is called to obtain the background noise energy;

[0069] When the energy spectrum of the real-time vibration time domain signal exceeds 20% of the background noise energy, the dynamic scaling of the sampling frequency is triggered, and the dynamically scaled sampling frequency instruction is fed back to the sensor group module.

[0070] Specifically, the module receives the main engine speed (RPM) and load rate (%) data from the ship control terminal in real time, and uses the hash matching algorithm to search for similar working conditions in the historical working condition database (speed error ±2%, load error ±5%). After a successful match, the noise fundamental frequency floating model under this working condition is called, which contains the fundamental frequency floating range vector and background noise energy , background noise energy Obtained through 1 / 3 octave spectrum analysis collected during the previous trouble-free operation period.

[0071] When the real-time vibration signal is input, the system performs short-time Fourier transform to generate the energy spectrum , perform energy integration within the frequency band defined by the fundamental frequency floating model. If the following formula is satisfied:

[0072]

[0073] The dynamic scaling engine is triggered:

[0074] High-frequency scenarios (base frequency > 50 Hz): Increase the sampling rate to 10 kHz to avoid high-frequency impact signal loss;

[0075] Low-frequency scenarios (base frequency ≤ 50 Hz): Reduce the sampling rate to 2 kHz to suppress low-frequency environmental noise;

[0076] The scaling command is fed back to the sensor group module via an encrypted protocol packet with a timestamp, triggering dynamic reconfiguration of the sensor (such as AD sampling rate switching and anti-aliasing filter bandwidth adjustment).

[0077] This solution constructs a dynamic benchmark by establishing a noise fundamental frequency floating model and combines the quantitative comparison of real-time vibration energy and background noise to achieve intelligent adjustment of the acquisition frequency, which not only avoids the accumulation of invalid data under high-frequency conditions but also ensures the capture of fault characteristics under low-frequency conditions.

[0078] Through the above technical solution, this application effectively solves the contradiction between the accuracy and efficiency of vibration signal acquisition under variable operating conditions of inland ferries. By dynamically adjusting the sampling frequency, it suppresses the interference of background noise on the effective signal, provides a high-quality data basis for subsequent envelope extraction and health scoring, and significantly improves the recognition accuracy of micro-fault features.

[0079] The present application further proposes that the parameter adjustment module further includes:

[0080] The frequency extraction unit performs fast Fourier transform on the vibration time domain signal and outputs the real-time fundamental frequency value;

[0081] The floating window comparison unit compares the real-time fundamental frequency value with the fundamental frequency floating range vector;

[0082] The sampling decision unit determines whether the vibration time domain signal is within the fundamental frequency floating range vector, and outputs time-aligned data if it is.

[0083] Otherwise, in response to the real-time fundamental frequency value being out of the fundamental frequency floating range vector, execute:

[0084] When the real-time fundamental frequency value is greater than the upper limit of the fundamental frequency floating range vector, an instruction to increase the sampling rate to 10kHz is generated;

[0085] When the real-time fundamental frequency value is less than the lower limit of the fundamental frequency floating range vector, an instruction to reduce the sampling rate to 2kHz is generated;

[0086] The command feedback unit transmits the command to the sensor group module for dynamic acquisition. When the parameter adjustment module and the noise estimation module have conflicting conditions, the noise estimation module takes priority.

[0087] Specifically, when the real-time fundamental frequency of the vibration signal is within the pre-stored floating range, the system maintains the original sampling rate and performs time alignment processing to ensure data consistency; when the fundamental frequency exceeds the upper limit, it is determined that there is a high-frequency abnormal component, and the sampling rate needs to be increased to 10kHz to capture transient impact signals; when the fundamental frequency is lower than the lower limit, it is determined that there is a low-frequency drift phenomenon, and the sampling rate needs to be reduced to 2kHz to reduce redundant data. In the specific implementation process, the sampling rate can be adjusted according to the type of equipment and is not limited to 10kHz or 2kHz. By dynamically adjusting the sampling rate, undersampling distortion under high-frequency conditions is avoided, and storage and computing overhead under low-frequency conditions is reduced. The instruction feedback unit sends the adjustment instruction to the sensor group module in real time to form a closed-loop control mechanism.

[0088] This solution optimizes system resource utilization while ensuring signal integrity through dynamic fundamental frequency detection and adaptive sampling rate adjustment. Through this technical solution, this application effectively resolves the conflict between accuracy and efficiency in vibration signal acquisition under variable operating conditions on inland river ferries. It accurately captures fault characteristics by adjusting the sampling rate when the fundamental frequency is abnormal, while also reducing the amount of invalid data under normal operating conditions, thereby improving the sensitivity of micro-fault detection and the economic efficiency of system operation.

[0089] The present application further proposes that the specific steps of the envelope extraction module include:

[0090] The dynamic bandwidth calculation unit obtains a width value ΔF based on the difference between the upper and lower limits of the fundamental frequency floating range vector, and calculates the bandwidth BW; the difference operator is combined with the proportional coefficient calculation to achieve the bandwidth value by multiplying the difference between the upper and lower limits of the fundamental frequency floating range vector by a preset coefficient. For example, the coefficient can be set to 0.5-0.8, thereby matching the bandwidth range with the actual vibration frequency change trend of the equipment.

[0091] The center frequency setting unit uses the median of the fundamental frequency floating range vector as the center frequency of the bandpass filter; by taking the arithmetic mean of the fundamental frequency floating range vector as the center frequency, the filtering range covers the current main vibration frequency band of the equipment.

[0092] The Hilbert execution unit uses the center frequency and bandwidth as parameters to perform filtering and transformation on the time-aligned data to generate the current waveform envelope.

[0093] The specific implementation process is as follows: Receive the fundamental frequency floating range vector output by the noise estimation module , calculate the characteristic frequency bandwidth: ΔF= - , generate signal analysis bandwidth: BW=k1×ΔF+k2, among which, for power equipment, k1=1.2, k2=3.0; for auxiliary equipment, k1=0.8, k2=2.0; the vibration energy of diesel engines is concentrated in the frequency band of 1.2 times ΔF, and the motor equipment needs to be expanded to 0.8 times ΔF + 2Hz to cover electromagnetic harmonics.

[0094] Calculate the median fundamental frequency , which is calculated as follows:

[0095]

[0096] Construct bandpass filter parameters: passband range = [ -BW / 2, +BW / 2];

[0097] Extract the current waveform envelope Env(t) through Hilbert transform: H(s)=Hilbert(s(t)); Env(t)=√[s(t)²+H(s(t))²].

[0098] The parameters of the current waveform envelope generation algorithm are defined as follows:

[0099] symbol definition Physical meaning Ship scene role H(s) Hilbert(s(t)) Perform Hilbert transform on the original signal s(t) Convert the real signal into an analytical signal and obtain the orthogonal components s(t) Input time domain vibration signal Original vibration waveform collected by the sensor (time domain) Fault characteristics such as mechanical impact / bearing wear Env(t) √[s(t)²+H(s(t))²] Signal instantaneous amplitude (current waveform envelope) Demodulate the fault impulse characteristics modulated on the carrier

[0100] This solution dynamically calculates the bandwidth and center frequency, ensuring that the filtering parameters always track the actual vibration state of the equipment. For example, it automatically expands the bandwidth when the fundamental frequency fluctuation intensifies, ensuring the complete extraction of effective signals. Through the above technical solution, this application can adapt to the fundamental frequency fluctuations of inland ferry vessels under variable speed and load conditions, dynamically adjust the envelope extraction parameters, eliminate the characteristic signal distortion caused by traditional fixed bandwidth filtering, provide accurate waveform envelope data for subsequent health score calculations, and effectively improve the accuracy of early micro-fault feature identification.

[0101] This application further proposes that the specific steps for the health score module to output the health score include:

[0102] Perform empirical mode decomposition on historical health curves to generate multiple eigenmode components;

[0103] Calculate the projection energy ratio of the current waveform envelope on each eigenmode component;

[0104] Calculate the trend deviation entropy value based on the projection energy ratio;

[0105] The health score is output based on the trend deviation entropy value and the preset calculation rules.

[0106] In the trend deviation entropy calculation step, the Shannon information entropy formula is used to quantify the abnormality of the projection energy distribution:

[0107]

[0108] Where H is the trend deviation entropy (unit: bit);

[0109] is the projection energy ratio of the current waveform envelope on the kth eigenmode component;

[0110] n is the total number of components involved in the calculation (usually the first 5 components);

[0111] For example, when bearings experience early wear, the energy distribution exhibits a concentrated effect:

[0112] Normal state: ;

[0113] Wear status: ;

[0114] The entropy value dropped by 0.54 bits, triggering a decrease in the health score (the Z value dropped by 12 points), accurately capturing early degradation characteristics.

[0115] Specifically, the historical health curve is broken down into multiple intrinsic mode components through empirical mode decomposition, each of which represents the vibration characteristics of a different frequency band. The proportion of the projected energy of the current waveform envelope on each component reflects the difference in matching between the current vibration state and the historical health state. By calculating the entropy value of these energy distributions, the degree of deviation of the current vibration envelope from the historical benchmark can be quantified. For example, when the entropy value exceeds the preset threshold, it indicates that the device vibration pattern has abnormally shifted, and the health score will be reduced according to the preset rules. This process achieves a refined assessment of the device health status through dynamic decomposition and multi-dimensional energy analysis.

[0116] This solution adaptively extracts multi-scale features from historical curves through empirical mode decomposition. Combined with projected energy ratios and entropy calculations, it effectively distinguishes normal operating fluctuations from early signs of faults. For example, a sudden change in engine speed, resulting in a fundamental frequency shift, might be misidentified as an anomaly by traditional methods. However, this solution, through analysis of the energy distribution of the decomposed components, accurately identifies energy changes in the same frequency band as the historical healthy curve, avoiding false positives.

[0117] Through the above-mentioned technical solution, this application can dynamically adapt to the time-varying characteristics of equipment vibration, improving the accuracy of health scores. Through multi-scale decomposition and entropy quantification, it enhances sensitivity to subtle fault characteristics, resolving the problem of missed early fault detection caused by fixed thresholds in traditional methods. Furthermore, the calculation of energy distribution deviation effectively suppresses misjudgments caused by operating condition fluctuations, providing a reliable basis for inland ferry equipment health monitoring.

[0118] like Figure 5As shown, this application further proposes that the collaborative analysis module specifically includes:

[0119] The temperature gradient processing unit receives the verified temperature values ​​and calculates the real-time temperature difference gradient vector ▽T between adjacent sensor nodes. The real-time temperature difference gradient vector refers to the temperature change rate of adjacent sensor nodes. Specifically, it can be achieved by differentially calculating the temperature difference between adjacent nodes and dividing it by the sampling interval time. It is used to characterize abnormal fluctuations in the local temperature field of the device.

[0120] Envelope energy quantization unit, time-aligned data contains vibration spectrum feature vectors with timestamps, extracts the 1-3 times frequency band amplitude of the time-aligned data, and calculates the envelope energy ratio value E n ;E n =(A2+A3) / A1, where A1 is the fundamental frequency amplitude, A2 is the second harmonic amplitude, and A3 is the third harmonic amplitude, all of which are extracted from the vibration spectrum by fast Fourier transform;

[0121] The envelope energy ratio refers to the energy distribution ratio of the vibration signal in a specific frequency band. It can be achieved by integrating the signal after bandpass filtering and is used to quantify the abnormal energy accumulation of the equipment vibration characteristics.

[0122] The correlation matrix construction unit is based on the real-time temperature gradient vector ▽T, the envelope energy ratio En value and the current harmonic distortion rate D i , generate a three-dimensional correlation matrix M=[▽T,En,D i ];

[0123] a dynamic window conversion unit, which expands the first harmonic ratio threshold into a three-dimensional threshold space and uses the three-dimensional threshold space as a second dynamic threshold window;

[0124] The synchronous over-limit detection unit activates and outputs a microfault signature when the values ​​of all dimensions in the correlation matrix M simultaneously exceed the corresponding window range of the second dynamic threshold window within five consecutive sampling periods. The synchronous over-limit detection unit monitors whether the parameters of each dimension of the correlation matrix simultaneously exceed the corresponding threshold window within five consecutive sampling periods. It only determines a valid microfault signature when the synchronous over-limit is continuously exceeded.

[0125] This solution realizes multi-parameter collaborative analysis by constructing a three-dimensional correlation matrix, and combines it with a dynamic threshold window adjustment mechanism to enable the judgment criteria to adapt to the parameter fluctuation characteristics of inland ferry ships under variable speed and load conditions. Compared with the single-dimensional threshold triggering method, the three-dimensional threshold space and synchronous detection conditions can effectively distinguish between normal operating fluctuations and real fault characteristics. Through the above technical solution, the present application can accurately identify early equipment failures caused by abnormal temperature gradients, sudden changes in envelope energy, and current harmonic distortion, and avoid false alarms caused by single parameter fluctuations or instantaneous interference. Through the three-dimensional dynamic threshold window and multi-cycle synchronous detection mechanism, the judgment reliability is significantly improved while retaining fault sensitivity, solving the technical problem of missed detection of multi-parameter weakly correlated micro-faults in traditional systems.

[0126] The present application further proposes that the specific steps performed by the dynamic window conversion unit include:

[0127] The threshold space construction subunit expands the first harmonic ratio threshold K into a three-dimensional threshold space, which includes:

[0128] Current harmonic window: [0.85K×(1-0.05|▽T|), 1.15K×(1+0.03|▽T|)];

[0129] Envelope energy window: [1.2-0.1K+0.02·ΔF, 1.8+0.1K-0.01·ΔF];

[0130] Temperature gradient window: [▽T min ×(1-0.1·D i ), ▽T max ×(1+0.15·D i )]; where ▽Tmin and ▽Tmax are the equipment safety temperature difference thresholds, which are pre-stored in the historical operating condition database;

[0131] The working condition response subunit calculates the rate of change η of the fundamental frequency floating range vector and adjusts the window boundary of the three-dimensional threshold space according to the rate of change η:

[0132] If η>0.2, all window boundaries are expanded by 15%;

[0133] If η<0.05, the window boundary of the current harmonic window is compressed to ±8%;

[0134] The output subunit generates a second dynamic threshold window = {current harmonic window, envelope energy window, temperature gradient window} according to the adjusted window boundary.

[0135] Among them, in the threshold space construction sub-unit, the current harmonic window refers to the dynamic range set for the current harmonic distortion rate. Specifically, it can be achieved by using the temperature gradient as a correction factor to dynamically scale the initial threshold, and by introducing the temperature gradient change to compensate for the operating condition drift of the current parameter. The envelope energy window refers to the dynamic range set based on the vibration envelope energy ratio. Specifically, it can be achieved by using the bandwidth parameter of the fundamental frequency floating range vector for boundary calculation, and by combining the fundamental frequency change characteristics to enhance the robustness of the envelope energy criterion. The temperature gradient window refers to the dynamic range constructed according to the equipment safety temperature difference threshold. Specifically, it can be achieved by using the current harmonic distortion rate as a weight factor to adjust the safety threshold boundary, and by correlating current parameters to reflect the correlation characteristics of temperature anomalies.

[0136] The first harmonic ratio threshold K (scalar value) is received and combined with the device physical characteristic parameter library to perform three-dimensional expansion:

[0137] Current harmonic dimension: basic window range: [0.85K, 1.15K];

[0138] Temperature gradient compensation: window lower limit = 0.85K × (1-0.05|▽T|), window upper limit = 1.15K × (1+0.03|▽T|);

[0139] Among them, |▽T| is the modulus of the real-time temperature gradient vector;

[0140] Envelope energy dimension: Basic window range: [1.2-0.1K, 1.8+0.1K]

[0141] Frequency coupling correction: Window lower limit = 1.2-0.1K + 0.02·ΔF, window upper limit = 1.8+0.1K-0.01·ΔF; where ΔF is the fundamental frequency floating range vector width;

[0142] Temperature gradient dimension: basic window range: [▽T_min,▽T_max] (pre-stored equipment safety temperature difference range);

[0143] Current distortion rate compensation: Window lower limit = ▽T_min×(1-0.1·distortion rate), window upper limit = ▽T_max×(1+0.15·distortion rate);

[0144] The rate of change of the fundamental frequency floating range vector output by the response noise estimation module , which is calculated as follows:

[0145]

[0146] like >0.2 (violent fundamental frequency fluctuations), apply ±15% boundary expansion to all dimensional windows;

[0147] like <0.05 (fundamental frequency stability), compressing the current harmonic dimension window to ±8% tolerance;

[0148] Output the mathematical expression of the second dynamic threshold window:

[0149] The second dynamic threshold window = {current harmonic window: [window lower limit current, window upper limit current],

[0150] Envelope energy window: [lower window energy, upper window energy],

[0151] Temperature gradient window: [window lower limit temperature, window upper limit temperature]};

[0152] Among them, the compensation relationship between the dimension of the correlation matrix M and the corresponding threshold window is as follows:

[0153] Correlation matrix M dimension Corresponding threshold window Dynamic compensation factor Current harmonic distortion Current harmonic window Temperature gradient ▽T Envelope energy value En Envelope Energy Window Fundamental frequency width ΔF Temperature gradient ▽T Temperature gradient window Current distortion rate

[0154] This application constructs a three-dimensional dynamic threshold space, jointly analyzes the current harmonic distortion rate, envelope energy, and temperature gradient, and adjusts the threshold window boundaries in real time based on the fundamental frequency change rate, thus addressing the technical shortcomings of traditional single-dimensional thresholds, which have poor adaptability under variable operating conditions. At the same time, by introducing temperature gradients to correct the boundaries of the current harmonic window and using fundamental frequency bandwidth parameters to set the envelope energy window, collaborative optimization among multiple parameters is achieved, overcoming the misjudgment problem caused by isolated parameter analysis in existing technologies.

[0155] Through the above technical solution, the present application can accurately distinguish between normal operating fluctuations and early micro-fault characteristics under high-frequency speed and load conditions of inland ferry ships. The construction of a three-dimensional dynamic threshold space significantly improves the ability to capture multi-parameter correlation features, avoiding false alarms caused by fluctuations in a single parameter. The window boundary adjustment mechanism based on the fundamental frequency change rate enables the system to adaptively match the detection sensitivity under different navigation conditions, effectively suppressing false alarms while ensuring the micro-fault detection rate. The cross-correction mechanism of temperature gradient and current harmonic parameters enhances the adaptability of the threshold window to sudden changes in equipment load and changes in ambient temperature, solving the technical problem of high missed detection rate of traditional methods under dynamic conditions.

[0156] This application further proposes that the early warning execution module includes:

[0157] The topology mapping unit matches the micro-fault signature code with the device hierarchy code in the device topology relationship library to generate the fault location coordinates;

[0158] The command decision unit calls the corresponding operation instructions based on the preset alarm range of the health score and the fault location coordinates. The operation instructions include equipment start and stop commands and power adjustment parameters;

[0159] The multi-level push unit pushes operation instructions to the equipment controller of the ship control terminal and transmits the encrypted instruction backup package to the cloud server at the same time.

[0160] When the micro-fault feature code is activated, the topology mapping unit retrieves the corresponding hierarchical code from the device topology relationship library by parsing the device identification field in the feature code. The hierarchical code is mapped and converted with the coordinate system of the three-dimensional model of the ship to generate fault location coordinates that include physical location and logical affiliation. After receiving the health score and location coordinates, the instruction decision unit selects the corresponding operation strategy according to the preset alarm interval in which the score is located: when the score falls into the primary alarm interval, the power adjustment parameters are called for load reduction operation; when the score falls into the severe alarm interval, the equipment shutdown command is directly sent. The multi-level push unit transmits the generated operation instructions to the equipment controller of the ship control terminal in real time for execution, and at the same time encapsulates the instruction content, timestamp and equipment status snapshot into an encrypted data packet, which is transmitted to the cloud server through an independent communication link for dual storage.

[0161] In some specific implementations, the device topology relationship library may use a graph database to store device connection relationships, and quickly match the associated paths of faulty devices through a graph traversal algorithm; the encrypted instruction backup package may use sharding encryption technology to divide the instruction content into multiple data blocks and encrypt them separately before transmission.

[0162] Compared to existing technologies, traditional ship health monitoring systems trigger fixed control commands based solely on a single alarm signal, failing to dynamically adjust response strategies based on fault location and health status. Existing technologies also lack redundant backup mechanisms for command transmission, posing the risk of command loss if the ship's communication link is interrupted. This solution precisely locates faults through device topology mapping, combines health score grading with triggering differentiated control strategies, and employs encrypted multi-level push notifications to ensure reliable command transmission.

[0163] like Figure 6 As shown, this application further proposes that the health scoring module also includes:

[0164] The characteristic peak extraction unit extracts the peak sequence of the waveform envelope corresponding to the current micro-fault characteristic code after the early warning execution module outputs the operation instruction;

[0165] Curve deviation calculation unit, calculates the dynamic time warping distance W between the peak sequence and the historical health curve;

[0166] The curve library update unit generates an updated history curve by weighted fusion when the dynamic time warping distance W exceeds the preset deviation threshold three times in a row and the health score drops by more than 5% for two consecutive cycles;

[0167] The updated historical curve will replace the historical health curve of the corresponding device in the pre-stored historical health curve library.

[0168] After the early warning execution module outputs an operation instruction, the characteristic peak extraction unit immediately locks onto the waveform envelope corresponding to the microfault signature code and extracts a peak sequence containing amplitude and timing information (such as the zero-crossing points and extreme points of the bearing wear characteristic waveform). The curve deviation calculation unit uses a dynamic time warping algorithm to perform nonlinear path matching between the peak sequence and the pre-stored historical health curve and calculate the minimum cumulative distance W. This distance quantifies the morphological difference between the current fault waveform and the historical baseline, overcoming the limitation of traditional root mean square error (RMS) insensitivity to waveform phase shift.

[0169] When the system detects that the distance W exceeds the preset deviation threshold (this threshold is dynamically adjusted based on the device type and operating life) three times in a row, and the health score drops by more than 5% for two consecutive cycles, the curve library update unit starts the weight fusion engine. The engine weights the current waveform envelope with the original historical curve according to the preset ratio:

[0170] New curve = α × historical curve + (1-α) × current envelope (α = 0.7)

[0171] The generated new curve is then updated to the historical health curve library, replacing the original baseline data for the corresponding device. This process uses the alpha coefficient to control the update amplitude, gradually absorbing the performance degradation characteristics of the device while preserving the credibility of the historical data.

[0172] This solution dynamically updates historical health curves, retaining the initial health status characteristics of the equipment while progressively integrating the latest operating data, eliminating the risk of misjudgment caused by long-term performance degradation while maintaining sensitivity to sudden abnormal conditions.

[0173] Through the above technical solution, this application effectively solves the problem of misjudgment caused by the fixed historical curve in traditional health monitoring systems. By dynamically updating the baseline curve, it achieves continuous calibration of the device health status. When the device experiences gradual performance degradation, the system can automatically adjust the health score baseline to avoid false alarms caused by mismatches between the historical curve and the current status. When the device suddenly anomalies, it can still accurately identify significant deviations based on the latest health curve, improving the accuracy and reliability of early fault detection.

[0174] The present application further proposes that the collaborative analysis module also includes a behavior compensation unit. The specific steps of the behavior compensation unit include:

[0175] Receive crew operation instruction data and count the frequency of operation instructions;

[0176] When the operation instruction frequency exceeds the threshold number of times within the preset time window, a high-frequency operation flag is generated and current distortion rate compensation is started;

[0177] Current distortion rate compensation applies a time decay compensation coefficient to the current distortion rate data to generate a corrected current harmonic distortion rate;

[0178] Using the corrected current harmonic distortion rate Replace the original current harmonic distortion rate D i , input to the construction process of the correlation matrix M.

[0179] The behavior compensation unit receives the real-time operation instruction stream transmitted by the ship control bus, and uses the sliding time window algorithm to count the instruction triggering frequency per unit time. When it is detected that the instruction frequency exceeds the preset threshold (such as switching instructions > 5 times / minute), a high-frequency operation mark is generated and the exponential decay engine is activated. The engine builds an attenuation model based on the electromagnetic transient characteristics of the equipment, and applies a time-related compensation coefficient γ(t)=e^(-λ·t) to the original current distortion rate, where λ is the attenuation constant adaptive to the equipment type (motor type λ=0.25, pump and valve type λ=0.15), and t is the duration after the most recent high-frequency operation is triggered. The generated corrected distortion rate Will replace the original current harmonic distortion rate D i Input correlation matrix construction unit to eliminate non-fault distortion interference from the source.

[0180] This solution, through the innovative design of an adaptive attenuation model for equipment type and continuous time-domain dynamic compensation, effectively solves common problems in the field of ship monitoring while maintaining crew operational flexibility. It reduces the false alarm rate during high-frequency operation to a negligible level, avoids false operation and maintenance instructions triggered by operational interference, and fully retains the characteristics of actual equipment failures (such as continuous distortion caused by insulation degradation), maintaining the fault capture rate. Without the need for manual configuration of compensation parameters, the system automatically matches the optimal attenuation model based on the equipment type. This technology provides a pollution-free electrical data foundation for predictive maintenance, significantly improving the reliability of the ship's intelligent operation and maintenance system.

[0181] The following is a specific application example based on the inland ferry equipment health monitoring system:

[0182] The Yangtze Pearl, a Yangtze River freight ferry operating from Chongqing to Yichang, collects real-time equipment status through a sensor network deployed on the main engine, propeller, and power distribution compartment. When the ship passed through the rapids of the Three Gorges Reservoir, the main engine entered heavy-load acceleration mode. The noise estimation module detected a sudden 23% increase in speed and, using a historical operating condition database, matched the fundamental frequency fluctuation range vector to [28.5, 31.3] Hz. The parameter adjustment module captured the real-time vibration fundamental frequency of 29.8 Hz (within the fluctuation range), time-aligned the vibration signal, and transmitted it to the envelope extraction unit.

[0183] At this time, the collaborative analysis module detected three sets of abnormal collaborative phenomena: the temperature gradient of the main engine cylinder rose to 4.2℃ / m (the temperature difference of the neighboring sensors continued to exceed the standard), the current harmonic distortion rate reached 0.41 (exceeding the daily average by 35%), and the thruster envelope energy value En=0.52 (the energy in the characteristic frequency band of bearing wear surged).

[0184] The dynamic window conversion unit generates a three-dimensional threshold space based on the first harmonic ratio threshold K=0.35:

[0185] Current harmonic window: [0.29, 0.41]

[0186] Envelope energy window: [1.15,1.68]

[0187] Temperature gradient window: [3.8,5.2]℃ / m

[0188] When the correlation matrix M exceeds the limit synchronously within five consecutive sampling periods, the system activates the micro-fault feature code "P01" (early wear of the propeller bearing).

[0189] The health score module compares the current waveform envelope with the historical curve and finds that the phase offset has increased significantly and the rate of change has accelerated, and the health score has dropped to 72 points (the health threshold is 80 points).

[0190] The early warning execution module maps the signature code to the device topology library, highlights the thruster position (30.76°N, 111.28°E) on the electronic nautical chart, and simultaneously pushes two operational instructions:

[0191] Reduced load operation command: thruster power is limited to 75% of the rated value;

[0192] Lubrication path switching command: start the backup oil circuit forced flushing system;

[0193] After the crew receives the instructions on the control terminal, the system automatically records the execution process and uploads the encrypted log to the cloud operation and maintenance platform.

[0194] This solution maintains the continuity of vibration signal analysis during rapids, avoiding monitoring interruptions caused by sudden changes in operating conditions in traditional systems. It separates the wear characteristics of thruster bearings from complex interference, improves the sensitivity of micro-fault identification compared to single-dimensional analysis, intuitively locates faulty equipment on electronic charts, and simultaneously pushes executable maintenance plans, shortening the emergency response path.

[0195] This system enabled the Yangtze River Pearl to avoid suspension accidents caused by sudden propeller failure during its voyages during the flood season, and extended the bearing replacement cycle through early intervention, verifying the practical value of predictive maintenance in the field of inland shipping.

[0196] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. The IoT-based inland river ferry ship equipment health monitoring and early warning system is characterized by: include: The sensor group module is used to collect temperature values, vibration time domain signals and current harmonic distortion rate of the ship; The noise estimation module is used to receive real-time navigation status data from the ship control terminal and generate the fundamental frequency floating range vector and the first harmonic ratio threshold based on the pre-stored historical operating condition database; a parameter adjustment module, configured to determine whether the vibration time domain signal is within the fundamental frequency floating range vector, and if so, to time align the vibration time domain signal and output time alignment data; An envelope extraction module, configured to perform bandpass filtering and Hilbert transform on the time-aligned data to generate a current waveform envelope; A health scoring module is used to compare the current waveform envelope with a pre-stored historical health curve using a dynamic time warping algorithm to output a health score; A collaborative analysis module constructs a correlation matrix M based on the verified temperature value, the time-aligned data, and the current harmonic distortion rate; Expanding the first harmonic ratio threshold into a second dynamic threshold window; the second dynamic threshold window = {current harmonic window, envelope energy window, temperature gradient window}; When the correlation matrix M synchronously exceeds the second dynamic threshold window within five consecutive sampling periods, outputting a micro-fault feature code; The early warning execution module is used to map the micro-fault feature code and the health score to a pre-stored equipment topology relationship library, generate and output an operation instruction to the ship control terminal.

2. The IoT-based inland river ferry ship equipment health monitoring and early warning system according to claim 1 is characterized by: Each sensor unit of the sensor group module includes a self-checking unit, and the self-checking unit specifically performs the following steps: The sensor group module performs drift detection on the reference voltage of each sensor unit within a preset acquisition period and generates a voltage stability mark; Comparing the currently collected temperature value of the sensor with the temperature value of the adjacent sensor to generate a temperature difference, comparing the temperature difference with a preset tolerance δ, and generating a sensor failure code if the temperature difference exceeds the preset tolerance δ three times in a row; The voltage stability mark and the sensor failure code are respectively embedded in the collected current harmonic distortion rate and the verified temperature value and transmitted to the collaborative analysis module; the collaborative analysis module receives the embedded mark and failure code, and when the sensor failure code is detected, ignores the temperature value of the corresponding node.

3. The inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things according to claim 1 is characterized by: The noise estimation module performs adaptive adjustment of noise tolerance, and the specific steps include: According to the main engine speed and load rate in the real-time navigation status data, calling the noise fundamental frequency floating model in the historical operating condition database and obtaining background noise energy; When the energy spectrum of the real-time vibration time domain signal exceeds 20% of the background noise energy, the dynamic scaling of the sampling frequency is triggered, and the dynamically scaled sampling frequency instruction is fed back to the sensor group module.

4. The inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things according to claim 1 is characterized by: The parameter adjustment module also includes: a frequency extraction unit, performing a fast Fourier transform on the vibration time domain signal and outputting a real-time fundamental frequency value; A floating window comparison unit performs interval comparison on the real-time fundamental frequency value and the fundamental frequency floating range vector; a sampling decision unit, for determining whether the vibration time domain signal is within the fundamental frequency floating range vector, and outputting the time alignment data if so; Otherwise, in response to the real-time fundamental frequency value being out of the fundamental frequency floating range vector, executing: When the real-time fundamental frequency value is greater than the upper limit of the fundamental frequency floating range vector, generating an instruction to increase the sampling rate to 10kHz; When the real-time fundamental frequency value is less than the lower limit of the fundamental frequency floating range vector, generating an instruction to reduce the sampling rate to 2kHz; The instruction feedback unit transmits the instruction to the sensor group module for dynamic collection.

5. The inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things according to claim 3 is characterized by: The specific steps of the envelope extraction module include: a dynamic bandwidth calculation unit, which obtains a width value ΔF according to a difference between an upper limit and a lower limit of a fundamental frequency vector of the fundamental frequency floating range vector, and calculates a bandwidth BW; a center frequency setting unit, which uses the median of the fundamental frequency floating range vector as the center frequency of the bandpass filter; The Hilbert execution unit performs filtering and transformation on the time-aligned data with the center frequency and the bandwidth as parameters to generate the current waveform envelope.

6. The inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things according to claim 1 is characterized by: The specific steps of the health score module to output the health score include: Performing empirical mode decomposition on the historical health curve to generate multiple eigenmode components; Calculating the projection energy ratio of the current waveform envelope on each eigenmode component; Calculating a trend deviation entropy value according to the projection energy ratio; The health score is output based on the trend deviation entropy value and preset calculation rules.

7. The inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things according to claim 1 is characterized by: The collaborative analysis module specifically includes: A temperature gradient processing unit receives the verified temperature value and calculates a real-time temperature difference gradient vector ▽T of adjacent sensor nodes; Envelope energy quantization unit, the time-aligned data contains the vibration spectrum feature vector with a timestamp, extracts the 1-3 times frequency band amplitude of the time-aligned data, and calculates the envelope energy ratio value E n ; The correlation matrix construction unit is based on the real-time temperature gradient vector ▽T, the envelope energy ratio En value and the current harmonic distortion rate D i , generate the three-dimensional correlation matrix M=[▽T,En,D i ]; a dynamic window conversion unit, which expands the first harmonic ratio threshold into a three-dimensional threshold space and uses the three-dimensional threshold space as the second dynamic threshold window; The limit-crossing synchronous detection unit is configured to activate and output the micro-fault feature code when the values ​​of all dimensions in the correlation matrix M simultaneously deviate from the corresponding window range of the second dynamic threshold window within five consecutive sampling periods.

8. The IoT-based inland river ferry ship equipment health monitoring and early warning system according to claim 7 is characterized by: The specific steps performed by the dynamic window conversion unit include: The threshold space construction subunit expands the first harmonic ratio threshold K into the three-dimensional threshold space, where the three-dimensional threshold space includes: Current harmonic window: [0.85K×(1-0.05|▽T|), 1.15K×(1+0.03|▽T|)]; Envelope energy window: [1.2-0.1K+0.02·ΔF, 1.8+0.1K-0.01·ΔF]; Temperature gradient window: [▽T min ×(1-0.1·D i ), ▽T max ×(1+0.15·D i )]; where ▽Tmin and ▽Tmax are the equipment safety temperature difference thresholds, which are pre-stored in the historical operating condition database; The operating condition response subunit calculates the change rate η of the fundamental frequency floating range vector and adjusts the window boundary of the three-dimensional threshold space according to the change rate η: If η>0.2, all window boundaries are expanded by 15%; If η<0.05, the window boundary of the current harmonic window is compressed to ±8%; The output subunit generates the second dynamic threshold window = {current harmonic window, envelope energy window, temperature gradient window} according to the adjusted window boundary.

9. The IoT-based inland river ferry ship equipment health monitoring and early warning system according to claim 7 is characterized by: The early warning execution module includes: A topology mapping unit matches the micro-fault feature code with the device hierarchy code in the device topology relationship library to generate fault location coordinates; An instruction decision unit, which calls corresponding operation instructions according to the preset alarm interval of the health score and the fault location coordinates, wherein the operation instructions include equipment start and stop commands and power adjustment parameters; The multi-level push unit pushes the operation instruction to the device controller of the ship control terminal and transmits the encrypted instruction backup package to the cloud server at the same time.

10. The inland river ferry ship equipment health monitoring and early warning system based on the Internet of Things according to claim 6 is characterized by: The health scoring module also includes: A characteristic peak extraction unit, which extracts a peak sequence of a waveform envelope corresponding to a current micro-fault characteristic code after the warning execution module outputs the operation instruction; a curve deviation calculation unit, for calculating a dynamic time warping distance W between the peak sequence and the historical health curve; A curve library updating unit generates an updated history curve by weighted fusion when the dynamic time warping distance W exceeds a preset deviation threshold for three consecutive times and the health score decreases by more than 5% for two consecutive cycles; The updated historical curve replaces the historical health curve of the corresponding device in the pre-stored historical health curve library.

Images