Ship power system fault diagnosis method and system based on multi-source data fusion

Abstract

The present application relates to the technical field of ship power control and intelligent operation and maintenance, in particular to a ship power system fault diagnosis method and system based on multi-source data fusion, which collects multi-source control monitoring data to form a multi-source control state feature set around the ship power control system, generates a scene risk regulation amount, and establishes a cascading causal event sequence to generate a power chain propagation time index and a power chain propagation regulation amount, and then forms a dynamic regulation parameter to complete multi-source fusion diagnosis, fault prediction and health state evaluation, and output a predicted health management result. The present application unifies and correlates each control link through the multi-source control state feature set, combines the scene risk regulation amount and the cascading causal event sequence to generate the dynamic regulation parameter, realizes fusion diagnosis, fault prediction, health state evaluation and fault output mode dynamic regulation, and improves monitoring integrity, diagnosis reliability and working condition adaptability.

Description

Technical Field

[0001] This invention relates to the field of ship power control and intelligent operation and maintenance technology, and more specifically, to a method and system for diagnosing ship power system faults based on multi-source data fusion. Background Technology

[0002] Ship propulsion control systems typically consist of power plant supply and distribution control links, propulsion control links, steering control links, dynamic positioning control links, and communication and data acquisition links, which work together to complete power supply, propulsion execution, course maintenance, position maintenance, and operational data acquisition. With the development of ship automation, integrated electric propulsion, and dynamic positioning technologies, the coupling relationships between these control links have become increasingly tight. The power plant supply and distribution status not only affects propulsion power output but may also affect steering response, dynamic positioning thrust distribution, and the transmission quality of critical monitoring data.

[0003] Existing methods for diagnosing ship power system failures can typically identify component anomalies, parameter exceedances, and operational degradation based on equipment alarms, single-link operating parameters, or historical operating data, and have certain application value in routine operation and maintenance. However, in high-risk navigation scenarios such as navigation in bridge areas, harbors, narrow waterways, berthing operations, unberthing operations, dynamic positioning operations, and recovery from a complete power failure, the ship's maneuvering margin changes rapidly, the power control link response time is short, and the status of a single link or single device often cannot fully reflect the development process of the failure.

[0004] In actual operation, when voltage fluctuations, frequency deviations, abnormal power distribution, or circuit breaker malfunctions occur in the power supply and distribution control link of a power plant, their impact may be transmitted along the power supply relationship to the propulsion control link, steering control link, and dynamic positioning control link, manifesting as limited propulsion power, delayed rudder angle response, thrust distribution reconfiguration, or decreased position holding capability. Simultaneously, multi-source control and monitoring data may also suffer from inconsistent sampling periods, timestamp offsets, communication delays, data gaps, data degradation, and abnormal messages, leading to inconsistencies in timing or differences in credibility between evidence from different sources.

[0005] Based on this, the present invention proposes a method and system for fault diagnosis of ship propulsion system based on multi-source data fusion. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for diagnosing ship power system faults based on multi-source data fusion.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A fault diagnosis method for ship propulsion systems based on multi-source data fusion is used for control status monitoring, fault prediction, and health status management of ship propulsion control systems. The ship propulsion control system includes power supply and distribution control links, propulsion control links, steering control links, dynamic positioning control links, and communication and data acquisition links. The method includes the following steps:

[0009] S1, collects multi-source control monitoring data from multiple data sources to form a multi-source control status feature set with timestamps, link sources and data validity status;

[0010] S2, generates scenario risk control quantities based on multi-source control state feature sets;

[0011] S3, establish a cascaded causal event sequence based on the multi-source control state feature set; the cascaded causal event sequence is used to characterize the process of abnormal state of power supply and distribution control link propagating to propulsion control link, steering control link and dynamic positioning control link; based on the cascaded causal event sequence, generate power chain propagation time index and power chain propagation control quantity;

[0012] S4 generates dynamic control parameters based on scenario risk control quantity, power chain propagation control quantity, multi-source control state feature set, and data validity state, so as to dynamically control the multi-source fusion diagnosis process, fault prediction process, health status evaluation process, and fault output mode; the fault output mode includes at least one of component-level fault output mode and link cascade risk output mode.

[0013] S5, based on the dynamic control parameters generated by S4, performs multi-source fusion diagnosis, fault prediction, and health status evaluation on the multi-source control and monitoring data collected by S1, and generates predictive health management results.

[0014] In one embodiment, the generation of scenario risk control quantity includes: identifying dynamic high-risk navigation scenarios based on multi-source control state feature set, establishing scenario risk coupling relationship between navigation scenario constraints, ship motion state and power maneuvering capability, calculating the remaining maneuvering time margin index of the ship under dynamic high-risk navigation scenarios, and generating scenario risk control quantity based on the remaining maneuvering time margin index and scenario risk coupling relationship.

[0015] In one embodiment, the dynamic high-risk navigation scenario includes at least one of bridge area navigation, harbor navigation, narrow waterway navigation, berthing operation, unberthing operation, dynamic positioning operation, and ship-wide power loss recovery process.

[0016] In one embodiment, the generation of the remaining maneuvering time margin index includes: under dynamic high-risk navigation scenarios, determining the spatial risk margin, motion trend risk, and power maneuvering capability margin based on navigation scenario constraints, ship motion state, and power maneuvering capability; and generating a remaining maneuvering time margin index characterizing the ship's ability to maintain a safe maneuvering state based on the spatial risk margin, motion trend risk, and power maneuvering capability margin.

[0017] In one embodiment, the generation of scenario risk control quantity includes: forming a scenario risk state vector based on the scenario type of the dynamic high-risk navigation scenario, the remaining maneuvering time margin index, the degree of compression of the spatial risk avoidance margin, the rate of change of the motion trend risk quantity, and the degree of decay of the power maneuvering capability margin; and generating scenario risk control quantity based on the scenario risk state vector.

[0018] The scenario risk control measures include at least one of the following: scenario risk level, control margin compression coefficient, scenario risk change rate, scenario control priority, and fault prediction sensitivity correction.

[0019] In one embodiment, the generation of the power chain propagation time index and the power chain propagation control quantity includes: determining the time difference of the propagation of the abnormal state of the power supply and distribution control link to the response state of the propulsion control link, steering control link, and dynamic positioning control link based on the cascaded causal event sequence; calculating the power chain propagation speed index based on the time difference, the propagation hierarchy between the corresponding control links, and the degree of abnormality of the response state, and generating the power chain propagation time index based on the power chain propagation speed index; and generating the power chain propagation control quantity based on the cascaded causal event sequence, the power chain propagation speed index, and the power chain propagation time index.

[0020] In one embodiment, the power chain propagation control quantity includes at least one of the following: propagation path weight, cascade propagation rate correction, cascade critical time compression coefficient, fault source location weight, and link cascade risk priority.

[0021] In one embodiment, in S4, the generation of dynamic control parameters includes: generating diagnostic credibility indicators and diagnostic credibility control quantities based on the multi-source control state feature set and data validity state; constructing a risk propagation credibility control state including the scenario risk dimension, the power chain propagation dimension, and the diagnostic credibility dimension based on the scenario risk control quantity, the power chain propagation control quantity, and the diagnostic credibility control quantity; and generating dynamic control parameters based on the risk propagation credibility control state.

[0022] In one embodiment, the diagnostic confidence adjustment factor includes at least one of the following: data source evidence weight, outlier data suppression coefficient, degraded data compensation coefficient, credible evidence gating coefficient, and diagnostic output confidence correction factor.

[0023] In one embodiment, the ship propulsion system fault diagnosis system based on multi-source data fusion includes:

[0024] The data acquisition module is used to collect multi-source control and monitoring data from multiple data sources, forming a multi-source control status feature set with timestamps, link sources, and data validity status.

[0025] The risk generation module is used to generate scenario risk control parameters based on a multi-source control state feature set.

[0026] The propagation generation module is used to establish a cascaded causal event sequence based on a multi-source control state feature set, and to generate a power chain propagation time index and a power chain propagation control quantity based on the cascaded causal event sequence.

[0027] The parameter generation module is used to generate dynamic control parameters based on scenario risk control quantities, power chain propagation control quantities, multi-source control state feature sets, and data validity states.

[0028] The result generation module is used to perform multi-source fusion diagnosis, fault prediction, and health status evaluation on multi-source control and monitoring data based on the dynamic control parameters, and generate predictive health management results.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] This invention collects multi-source control monitoring data from multiple data sources and forms a multi-source control status feature set with timestamps, link sources, and data validity status. This enables the operation information of power plant power supply and distribution control links, propulsion control links, steering control links, dynamic positioning control links, and communication acquisition links to be correlated and processed under the same data basis. This improves the completeness, temporal consistency, and data reliability of ship power control status monitoring and reduces diagnostic bias caused by judgments from a single link or a single data source.

[0031] Based on the multi-source control state feature set, scenario risk regulation quantities are generated, and a cascaded causal event sequence is established. This enables correlation analysis of the propagation process of abnormal states in the power supply and distribution control link to the propulsion control link, steering control link, and dynamic positioning control link. The propagation time index and propagation regulation quantity of the power chain are obtained, so that fault diagnosis is no longer limited to component abnormality identification, but can further reflect the fault propagation path, propagation time, and link cascade risk.

[0032] Based on scenario risk control parameters, power chain propagation control parameters, multi-source control state feature sets, and data validity status, dynamic control parameters are generated to dynamically control the multi-source fusion diagnosis process, fault prediction process, health status evaluation process, and fault output mode. This enables the predicted health management results to take into account both component-level fault output modes and link cascade risk output modes, thereby improving the fault prediction adaptability, health status evaluation accuracy, and operation and maintenance decision reference value of the ship's power control system in complex operating scenarios. Attached Figure Description

[0033] Figure 1 A flowchart illustrating the fault diagnosis method for ship propulsion systems based on multi-source data fusion provided in an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of the process for generating scenario risk control parameters according to an embodiment of the present invention;

[0035] Figure 3 A schematic diagram illustrating the generation process of cascaded causal event sequences, dynamic chain propagation control quantities, and dynamic control parameters provided in an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the structure of a ship power system fault diagnosis system based on multi-source data fusion, provided in an embodiment of the present invention. Detailed Implementation

[0037] Example 1: Refer to Figures 1 to 3 A fault diagnosis method for ship propulsion systems based on multi-source data fusion is used for control status monitoring, fault prediction, and health status management of ship propulsion control systems. The ship propulsion control system includes power plant power supply and distribution control links, propulsion control links, steering control links, dynamic positioning control links, and communication acquisition links. The method includes the following steps:

[0038] S1 collects multi-source control and monitoring data from multiple data sources to form a multi-source control status feature set with timestamps, link sources, and data validity status; sets data access rules around the power plant power supply and distribution control link, propulsion control link, steering control link, dynamic positioning control link, and communication acquisition link, and synchronously collects multi-source control and monitoring data such as generator voltage, current, frequency, circuit breaker status, propulsion speed, propulsion power, rudder angle command, rudder angle feedback, dynamic positioning thrust allocation command, communication message delay, message packet loss status, and acquisition terminal online status;

[0039] For example, during berthing operations in the harbor, the power supply and distribution control link experiences bus frequency fluctuations, the propulsion control link simultaneously adjusts propulsion power, the steering control link corrects rudder angles, the dynamic positioning control link changes thrust distribution, and the communication acquisition link records changes in message delay. All of the above data are written into the acquisition queue according to a unified time base, and the corresponding link source is retained.

[0040] The collected multi-source control and monitoring data are calibrated with timestamps, marked with link sources, and judged for data validity. The timestamp calibration adopts a combination of shipboard unified clock, sampling period compensation, and message arrival time correction. The link source mark is assigned values ​​according to the power plant power supply and distribution control link, propulsion control link, steering control link, dynamic positioning control link, and communication acquisition link. The data validity status is judged based on data exceeding limits, data mutation, sampling interruption, duplicate messages, time drift, and link communication anomalies.

[0041] Any feature record in the multi-source control state feature set is denoted as

[0042]

[0043] in, Number the data source. For sampling sequence number, For the message source time, The message arrival time, For the calibrated timestamp, As the source of the link, For the name of the monitoring parameter, To monitor parameter values, The data is in a valid state. To determine the corresponding credibility, the calibrated timestamp is determined by the following formula:

[0044]

[0045]

[0046] in, For the first Clock offset of each data source This is the time calibration sample set. The criteria for determining data exceeding limits, data mutation, sampling interruption, duplicate messages, time drift, and link communication anomalies are denoted as follows: If the condition is true, assign a value of 1; if the condition is false, assign a value of 0. The corresponding confidence level is determined by the following formula:

[0047]

[0048] in, The coefficients are non-negative, and the sum of the coefficients is no greater than one. If... ,but It is in a valid state; if ,but It is in a degenerate state; if ,but This is an abnormal state; if there is no corresponding data within the sampling time window, then... The missing state is where The data exceeding the limit threshold is taken from the rated range of the equipment or the alarm limit; the data mutation threshold is taken from the quantile value of the change rate of historical normal voyage data or the allowable change rate of the equipment; the sampling interruption threshold is taken from the sampling period multiple; the time drift threshold is taken from the synchronization accuracy of the shipborne unified clock; and the link communication abnormality threshold is taken from the message delay limit and the packet loss rate limit.

[0049] For example, if a certain propulsion power data does not exceed the physical threshold, but its timestamp lags behind the power plant frequency fluctuation data in the same period, and there is a message delay in the communication acquisition link, then the propulsion power data is marked as having reduced validity. It is then written into the multi-source control state feature set along with the corresponding timestamp and link source, so that the multi-source control state feature set simultaneously includes numerical status, time sequence status, link affiliation, and trust status.

[0050] S2 generates scenario risk control parameters based on a multi-source control state feature set. Based on the timestamp, link source, and data validity status in the multi-source control state feature set, scenario discrimination is performed on the ship's location, navigation mission, and maneuvering conditions. Navigation in bridge areas, navigation within harbors, navigation in narrow waterways, berthing operations, unberthing operations, dynamic positioning operations, and the ship's power loss recovery process are considered as candidate types of dynamic high-risk navigation scenarios. The scenario type determination result is then formed by combining propulsion power changes, rudder angle response, dynamic positioning thrust allocation, power station power supply fluctuations, and communication acquisition delays.

[0051] The set of candidate types for dynamic high-risk navigation scenarios is denoted as , This includes navigation in bridge areas, navigation within harbors, navigation in narrow waterways, berthing operations, unberthing operations, dynamic positioning operations, and the ship's power recovery process after a complete power outage. Any candidate type. The scene matching score is recorded as follows:

[0052]

[0053]

[0054]

[0055] in, This is the matching value between the ship's current position and the bridge area, port area, narrow waterway, wharf boundary, dynamic positioning operation area, or the area where the ship has lost power and is being restored. The matching value for the mission corresponding to the navigation plan, berthing / unberthing operation instructions, or dynamic positioning mode. These are the motion matching values ​​corresponding to changes in speed, lateral speed, rate of change of heading angle, and relative distance to obstacles. To advance the dynamic matching values ​​corresponding to power changes, rudder angle response, dynamic positioning thrust distribution, and power supply and distribution control link fluctuations in the power plant; The weights are non-negative, and the sum of the weights is one. This is currently the most suitable dynamic high-risk navigation scenario. For scene type coefficients, The calibration value is set for the scenario type, and its range is from zero to one. It is determined by the ship handling manual, berthing and unberthing operation procedures, dynamic positioning operation restrictions or historical voyage calibration data.

[0056] For example, if a ship experiences a continuous decrease in speed, a rapid reduction in lateral distance, frequent adjustments in propulsion power, and a lag in rudder angle feedback during berthing operations in port, it is considered a dynamic high-risk navigation scenario under berthing operations.

[0057] Under the identified dynamic high-risk navigation scenarios, establish the scenario risk coupling relationship between navigation scenario constraints, ship motion state and dynamic maneuvering capability. The navigation scenario constraints include channel width, bridge pier safety distance, wharf boundary, no-navigation zone boundary and dynamic positioning allowable offset range. The ship motion state includes speed, bow angle change rate, lateral speed, yaw trend and relative obstacle distance change. The dynamic maneuvering capability includes available propulsion power, rudder efficiency margin, dynamic positioning thrust distribution margin and power supply stability of the power station power distribution control link.

[0058] The set of constraints on the navigation scenario is denoted as The space safety margin is determined by the minimum safe distance difference between the ship's current position and the constraint boundary, denoted as:

[0059]

[0060]

[0061] in, Number the constraint boundaries, For the ship to the The distance of each constraint boundary, For the first The safety distance corresponding to each constraint boundary The boundary safety distance given by navigation rules, port area rules or operating procedures. This refers to the safety compensation amount for the ship's hull dimensions. This is the positioning error compensation amount. The motion trend risk amount is denoted as:

[0062]

[0063]

[0064]

[0065] in, For speed, For lateral velocity, For the heading angle, The rate of change of the heading angle. This refers to the relative distance between the vessel and obstacles, bridge piers, dock boundaries, no-navigation zone boundaries, or dynamically positioned allowable offset boundaries. The sampling time interval, These are the limits for cruising speed, lateral speed, rate of change of heading angle, and speed relative to an obstacle. The weights are non-negative, and the sum of the weights is one. The margin of power handling capability is denoted as:

[0066]

[0067] in, For available propulsion power, For the propulsion power required in the current scenario, As a rudder efficiency margin, For the rudder effect required in the current scenario, To allocate margin for dynamic positioning thrust. The thrust allocation required for dynamic positioning. To ensure the stability of the power supply and distribution control link of the power plant. It is a positive number;

[0068] Available propulsion power, propulsion power required for the current scenario, rudder efficiency margin, rudder efficiency required for the current scenario, dynamic positioning thrust allocation margin, thrust allocation required for dynamic positioning, and power supply stability are determined by the following formula:

[0069]

[0070]

[0071]

[0072]

[0073]

[0074]

[0075]

[0076] in, For the number of generator sets, Number the generator set. For the first The online status of the generator sets For the first The rated power of the generator set For the first Availability factor of generator sets, For the basic load of the entire ship, For safety backup power, To advance power commands, This represents the minimum control power required for the current dynamic high-risk navigation scenario. For rudder angle command, For rudder angle feedback, For the allowable error of the rudder angle, This represents the maximum rudder angle. The rate of change of the target heading angle. This represents the maximum rate of change of the heading angle. For the number of thrusters or propellers, Number the thruster or propulsion unit. For the first The online status of each thruster or propeller. For the first The maximum usable thrust of a single propeller or thruster For the first A thruster or propulsion unit commands thrust. For environmental disturbance forces, To maintain the required control in position, For vector norm, For bus frequency, For the rated frequency, Bus voltage Rated voltage, For frequency tolerance, For voltage tolerance, This represents the current load power. The weights are non-negative, and the sum of the weights is one. It is a positive number;

[0077] In the berthing operation example, when the space risk margin continues to shrink, the motion trend risk increases rapidly, and the power control capability margin decreases due to limited propulsion power, the remaining control time margin index is calculated according to the degree of compression of the space risk margin, the rate of change of the motion trend risk, and the degree of decay of the power control capability margin.

[0078] The remaining maneuvering time margin index is determined by the minimum of the time corresponding to spatial risk avoidance margin, the time corresponding to motion trend risk, and the time corresponding to dynamic maneuvering capability margin, denoted as:

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086] in, This is an indicator of remaining manipulation time margin. This refers to the time corresponding to the spatial safety margin. The time corresponding to the risk level of the movement trend. This refers to the time corresponding to the margin of power handling capability. The speed at which a ship approaches a constraint boundary or obstacle in a navigation scenario. The positive rate of change of the risk quantity of the movement trend. The rate of decrease in power handling capability margin. The sampling time interval, It is a positive number. The scenario risk state vector is denoted as:

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] in, This represents the scenario risk state vector. To normalize the remaining manipulation time margin, The degree of compression of space safety margin, The degree of reduction in the margin of power handling capability. To pre-set a safety margin, As a safety distance benchmark, Score the risk of the scenario. The weights are non-negative, and the sum of the weights is one. Scenario risk level. Control margin compression factor Scenario risk change rate Scene control priority and fault prediction sensitivity correction amount Determine them respectively using the following formulas:

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] in, This is the threshold for the scenario risk level, and ; The gain coefficient is used to manipulate the margin compression factor. The gain coefficient for the fault prediction sensitivity correction; The weights are non-negative, and the sum of the weights is one.

[0099] The scenario risk control quantity includes at least one of the following: scenario risk level, control margin compression coefficient, scenario risk change rate, scenario control priority, and fault prediction sensitivity correction quantity. For example, in narrow waterway navigation, if the remaining control time margin index is lower than the preset safety margin, the degree of compression of the space avoidance margin is high, and the degree of decay of the power control capability margin is obvious, then the scenario risk level and scenario control priority are increased, and the fault prediction sensitivity correction quantity is increased.

[0100] S3. A cascaded causal event sequence is established based on the multi-source control state feature set. The cascaded causal event sequence is used to characterize the process of abnormal states in the power supply and distribution control link of the power plant propagating to the propulsion control link, steering control link, and dynamic positioning control link. Based on the cascaded causal event sequence, power chain propagation time indicators and power chain propagation control quantities are generated. According to the timestamps, link sources, and data validity status in the multi-source control state feature set, the state changes of the power supply and distribution control link, propulsion control link, steering control link, and dynamic positioning control link are time-aligned. Events such as voltage drop, frequency offset, circuit breaker abnormal jump, propulsion power fluctuation, propulsion speed following deviation, rudder angle command and rudder angle feedback deviation, and dynamic positioning thrust distribution mutation are written into the same time axis according to their occurrence time and combined with the link source to form event nodes.

[0101] Any event node in a cascaded causal event sequence is denoted as:

[0102]

[0103] in, Number the event nodes. As the source of the link, For event type, This represents the start time of the abnormal or response state. For duration, To indicate the degree of abnormality, The credibility of the evidence for the event is determined by the degree of anomaly, which is based on the magnitude of the deviation, its duration, and the slope of the change.

[0104]

[0105] in, For the monitoring parameters corresponding to the event, This is the normal baseline value. To allow for deviation, To allow for the duration, The slope of the change in the monitoring parameter corresponding to the event. To allow for varying slopes, The weights are non-negative, and the sum of the weights is one. If the power supply and distribution control link event... Events related to propulsion control link, steering control link, or dynamic positioning control link If the following conditions are met simultaneously, then and Write the cascaded causal event sequence according to the time of occurrence:

[0106]

[0107]

[0108]

[0109]

[0110]

[0111] Where a is the source event number and b is the target event number. The start time of the source event. The start time of the target event. For the time window of dissemination, This is the value for determining the energy supply relationship or control dependency. It is set to one if the energy supply relationship or control dependency exists, and zero if it does not exist. For the degree of abnormality of the target event, Relevance to changes in event state To assess the credibility of evidence related to the source event, To assess the credibility of evidence related to the target event, The threshold for the degree of abnormality. The relevance threshold, This is the threshold for the credibility of evidence;

[0112] For example, during dynamic positioning operations, the power supply and distribution control link experiences a decrease in bus frequency, followed by a limitation in propulsion power in the propulsion control link, a lag in rudder angle feedback in the steering control link, and a reconfiguration of thrust distribution in the dynamic positioning control link. A cascaded causal event sequence is established according to the chronological order of the above events and the link affiliation.

[0113] Based on the cascaded causal event sequence, the time difference of the propagation of the abnormal state of the power supply and distribution control link to the response state of the propulsion control link, steering control link, and dynamic positioning control link is determined. Combined with the propagation hierarchy relationship between the corresponding control links and the degree of abnormality of the response state, the power chain propagation speed index is calculated. The time difference is determined by the difference between the start time of the abnormal state of the power supply and distribution control link and the start time of each response state. The propagation hierarchy relationship is determined according to the association order of power supply impact, execution response, and control hold. The degree of abnormality of the response state is determined based on the deviation amplitude, duration, and change slope.

[0114] Power plant power distribution control link events To target control link event The time difference of occurrence is denoted as:

[0115]

[0116] The propagation level distance is denoted as It is a propagation relationship diagram. The number of edges in the shortest path from link a to link b; For the control link set, This refers to a set of energy supply relationships, control dependencies, or manipulation-maintenance relationships. The propagation speed of the kinetic chain is indicated by:

[0117]

[0118] The propagation time of the kinetic chain is marked as:

[0119]

[0120] The propagation path weight is denoted as:

[0121]

[0122] in, This is the set of valid propagation relationships in a cascading causal event sequence. This constitutes an effective propagation relationship between the source event and the target event. for Any valid transmission relationship in China For the event Credibility of evidence For the event Credibility of evidence For the event The degree of abnormality, For the event To the event The time difference of occurrence. The cascading propagation rate correction, cascading critical time compression factor, fault source location weight, and link cascading risk priority are respectively denoted as:

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] in, This is the correction factor for the cascade propagation rate. The propagation speed benchmark is determined from historical normal voyages, sea trial calibration data, or simulation calibration data. The critical time compression factor for cascaded cascades. To determine the weight of the source of the fault, Prioritize the risk of cascading links for a single effective propagation relationship. Prioritize the risk of link cascading at the current moment. The weights are non-negative, and the sum of the weights is one. It is a positive number.

[0129] The propagation relationship diagram consists of the power supply and distribution control link, propulsion control link, steering control link, and dynamic positioning control link of the power plant. ,in For the control link set, This is a set of power supply relationships, control dependencies, or manipulation-maintenance relationships. If the link... To link If there is a power supply relationship, control dependency relationship, or manipulation maintenance relationship, then ,otherwise Propagation level distance For propagation relationship diagram Midlink To link The number of edges in the shortest path. The propagation time window is determined based on historical calibration samples:

[0130]

[0131] in, Links in historical calibration samples Anomaly to Link The time difference in response to anomalies For the first Quantile values. The correlation of event state changes is determined by the maximum normalized cross-correlation within the time window:

[0132]

[0133] in, and Links and links The normalized sequence of anomalous states, For event analysis time window, and These are the mean values ​​within the corresponding time windows;

[0134] The propagation time index of the power chain is generated based on the propagation speed index of the power chain, and the cascaded causal event sequence, the propagation speed index of the power chain, and the propagation time index of the power chain are jointly used to generate the power chain propagation control quantity. The power chain propagation control quantity includes at least one of the following: propagation path weight, cascaded propagation rate correction amount, cascaded critical time compression coefficient, fault source location weight, and link cascaded risk priority.

[0135] Propagation speed benchmark Determined from historical normal voyages, sea trial calibration data, or simulation calibration data, the median value of the corresponding propagation relationship is taken:

[0136]

[0137] in, This is the set of propagation velocity indicators in the calibration samples. If historical calibration samples are missing, initial values ​​are set according to the propagation level distance and the sampling period of the ship control system.

[0138]

[0139] in, For link To link The initial number of propagation sampling periods, To standardize the sampling period, the weights in the link cascading risk priority are determined according to normalization constraints:

[0140]

[0141]

[0142] When calibration samples are missing, Take equal weight initial values; after obtaining calibration samples, update according to the minimum link cascade risk identification error criterion;

[0143] For example, when an abnormal state of the power supply and distribution control link causes power limitation in the propulsion control link in a short period of time, and further causes abnormal thrust distribution in the dynamic positioning control link, the propagation path weight from the power supply and distribution control link to the propulsion control link is increased, the fault source location weight is increased, and the link cascading risk priority is improved.

[0144] S4 generates dynamic control parameters based on scenario risk control quantity, power chain propagation control quantity, multi-source control state feature set, and data validity state, so as to dynamically control the multi-source fusion diagnosis process, fault prediction process, health status evaluation process, and fault output mode; the fault output mode includes at least one of component-level fault output mode and link cascade risk output mode.

[0145] Based on the multi-source control state feature set and data validity state, the integrity, timing consistency, link consistency and anomaly credibility of the data collected from each link are jointly judged to generate diagnostic credibility index and diagnostic credibility control quantity.

[0146] The diagnostic reliability metrics are determined by data integrity, time synchronization reliability, link consistency, and the proportion of abnormal conflicts. Data integrity, time synchronization reliability, link consistency, and the proportion of abnormal conflicts are denoted as follows:

[0147]

[0148]

[0149]

[0150]

[0151] Diagnostic reliability is indicated as:

[0152]

[0153] in, To determine the effective number of samples, To determine the required number of samples, For the number of data sources, For the first The calibrated timestamps of each data source To unify the time base, For the first Time synchronization threshold for each data source It is a set of adjacent control link relationships. This refers to the relationship between adjacent control links. The correlation between changes in the state of adjacent control links. The number of conflicting and unusual pieces of evidence. This represents the total number of abnormal pieces of evidence. As a diagnostic reliability indicator, The weights are non-negative, and the sum of the weights is one. The values ​​are positive. The data source evidence weight, outlier suppression coefficient, degraded data compensation coefficient, credible evidence gating coefficient, and diagnostic output confidence correction are denoted as:

[0154]

[0155]

[0156]

[0157]

[0158]

[0159] in, For the first Weight of evidence from each data source For the first Diagnostic reliability metrics corresponding to each data source For the first Each data source corresponds to a level of credibility. This is the outlier suppression coefficient. For degraded data compensation coefficients, For credible evidence gating coefficient, The threshold for credible evidence. The diagnostic output confidence correction value;

[0160] Among them, the data source evidence weight is determined according to the data stability, sampling continuity and matching degree with the status of adjacent links of the power plant power supply and distribution control link, propulsion control link, steering control link, dynamic positioning control link and communication acquisition link; the abnormal data suppression coefficient is determined according to the data exceeding the limit, mutation, duplicate message, time drift and sampling interruption; the degraded data compensation coefficient is determined according to the situation where the data validity status is reduced but still has trend reference value; the credible evidence gating coefficient is determined according to the degree of mutual corroboration between multi-source data; and the diagnostic output confidence correction amount is determined according to the number of valid evidence, the proportion of conflicting evidence and the credibility of key link data.

[0161] For example, during berthing operations, if there is a slight time drift in the propulsion power data, while the power station frequency, rudder angle feedback, and dynamic positioning thrust distribution all show a consistent trend, then the evidence weight of the propulsion power data source should be reduced, a degradation data compensation coefficient should be applied, and its trend participation value should be retained.

[0162] The scenario risk control quantity, power chain propagation control quantity, and diagnostic credibility control quantity are jointly organized according to the scenario risk dimension, power chain propagation dimension, and diagnostic credibility dimension to construct a risk propagation credibility control state, and dynamic control parameters are generated based on the risk propagation credibility control state.

[0163] The credible control status of risk propagation is recorded as follows:

[0164]

[0165] The dynamic control parameters are denoted as:

[0166]

[0167] in, To ensure a credible and controlled state of risk propagation, For a set of dynamically adjustable parameters, To integrate weights, The fault prediction trigger threshold, Output weights for link cascading risks. This is a correction factor for health status assessment. This is the fault output mode. The fusion weight, fault prediction trigger threshold, link cascading risk output weight, and health status evaluation correction amount are respectively denoted as:

[0168]

[0169]

[0170]

[0171]

[0172] The fault output mode is determined by the following formula:

[0173]

[0174] in, The weights are non-negative, and the sum of the weights is one. The minimum fault prediction trigger threshold, The baseline fault prediction trigger threshold, To trigger the threshold adjustment coefficient, Output a baseline weight for link cascading risk. Output weight adjustment coefficients for link cascading risk. This is a correction factor for health status assessment. The threshold for the fault output mode. It is a positive number;

[0175] Among them, the scenario risk dimension reflects the scenario risk level, manipulation margin compression coefficient, scenario risk change rate, scenario control priority, and fault prediction sensitivity correction amount; the power chain propagation dimension reflects the propagation path weight, cascade propagation rate correction amount, cascade critical time compression coefficient, fault source location weight, and link cascade risk priority; and the diagnostic credibility dimension reflects the data source evidence weight, abnormal data suppression coefficient, degraded data compensation coefficient, credible evidence gating coefficient, and diagnostic output confidence correction amount.

[0176] For example, when the frequency of the power supply and distribution control link decreases, the power of the propulsion control link is limited, and the thrust distribution of the dynamic positioning control link is abnormal during dynamic positioning operations, and there is a message delay in the communication acquisition link, the dynamic adjustment parameters increase the weight of the link cascading risk output mode, compress the cascading critical time, reduce the impact of delayed data on the diagnostic conclusions, and increase the weight of fault source location and the fault prediction sensitivity correction amount.

[0177] S5, based on the dynamic control parameters generated by S4, performs multi-source fusion diagnosis, fault prediction, and health status evaluation on the multi-source control and monitoring data collected by S1, and generates predictive health management results. The system reads the dynamic control parameters generated by S4, and uniformly organizes the multi-source control monitoring data collected by S1 according to timestamp, link source, and data validity status. Based on the data source evidence weight, abnormal data suppression coefficient, degradation data compensation coefficient, credible evidence gating coefficient, and diagnostic output confidence correction in the dynamic control parameters, the system performs layered fusion of data evidence from the power plant power supply and distribution control link, propulsion control link, steering control link, dynamic positioning control link, and communication acquisition link. During the fusion diagnosis process, when the scenario risk level is high, the participation weight of propulsion power, rudder angle feedback, dynamic positioning thrust allocation, and power plant frequency data that are directly related to maneuvering safety is increased. When the link cascading risk priority is high, the matching strength of the propagation path evidence between the power plant power supply and distribution control link and the propulsion control link, steering control link, and dynamic positioning control link is increased. Suppression or compensation processing is applied to the data corresponding to time drift, sampling interruption, duplicate messages, and communication delay to form diagnostic results for component abnormal status, link abnormal status, and cascading propagation status.

[0178] For example, in dynamic positioning operations, when the bus frequency drops, propulsion power is limited, thrust allocation is frequently reconfigured, and there is message delay in the communication acquisition link, the fusion diagnostic results will take the abnormality of the power supply and distribution control link of the power plant as high-weight evidence and make a correlation judgment on the response abnormalities of the propulsion control link and the dynamic positioning control link.

[0179] Based on the above-mentioned integrated diagnostic results and dynamic control parameters, fault prediction and health status evaluation are carried out. The scenario risk change rate, control margin compression coefficient, cascade propagation rate correction, cascade critical time compression coefficient, fault source location weight and diagnostic output confidence correction are jointly incorporated into the prediction calculation to generate a predicted health management result that includes fault type, fault location, fault development trend, remaining safety control margin, link cascade risk priority, health status level and diagnostic confidence level.

[0180] The normalized anomaly scores for each data source are denoted as follows:

[0181]

[0182] in, For the first Normalized anomaly scores for each data source For the first The monitoring parameter values ​​of each data source. For the first Normal baseline values ​​for each monitoring parameter For the first Allowable deviation for each monitoring parameter The result is a positive number. The fault prediction score and health status index are denoted as follows:

[0183]

[0184]

[0185] in, To score the fault prediction, As an indicator of health status, For the number of data sources, For the first Weighting of data sources For the first Each data source has a credible evidence gating coefficient. Output weights for link cascading risks. Prioritize the risk of link cascading. This is the correction amount for fault prediction sensitivity. Score the risk of the scenario. The diagnostic output confidence correction value. This is a correction factor for health status assessment. When... When, output the fault prediction result; when At that time, the system outputs either a normal monitoring result or a deterioration in health status. The health status level is recorded as:

[0186]

[0187] in, Health status level The threshold for health status level, and Predicted health management outcomes include at least one of the following: fault type, fault location, fault progression trend, remaining maneuver time margin, link cascading risk priority, health status level, and diagnostic confidence level. Prediction bias is denoted as:

[0188]

[0189] The diagnostic reliability index has been updated to:

[0190]

[0191] The data source evidence weights have been updated to:

[0192]

[0193] in, To predict bias, To predict the forecast values ​​in health management outcomes, This corresponds to the measured value or manually confirmed value for the next sampling period. To calibrate the upper limit for the corresponding parameter, To calibrate the lower limit for the corresponding parameter, C_d(t) represents the updated diagnostic confidence index, and C_d(t) represents the current diagnostic confidence index. To update the step size for diagnostic reliability metrics, and , For the first The updated evidence weights for each data source For the first Current evidence weights for each data source For the first Each data source corresponds to a prediction bias. For the first The step size for updating the weights of evidence from each data source, and ;

[0194] For example, during berthing operations, when the space safety margin shrinks rapidly, propulsion power fluctuations are accompanied by rudder angle feedback lag, and the dynamic control parameters show an increase in the fault prediction sensitivity correction, the prediction health management results give a comprehensive conclusion of propulsion control link execution capability degradation, steering control link response lag, and berthing maneuver margin reduction, and form corresponding results according to component-level fault output mode or link cascade risk output mode.

[0195] In this embodiment, the threshold, weight, and coefficient are determined based on at least one of the following: equipment manufacturer's rated parameters, equipment alarm limits, ship handling manual, berthing and unberthing operation procedures, dynamic positioning operation restrictions, electronic nautical chart data, port boundary data, sea trial calibration data, historical normal voyage data, simulation calibration data, and manual confirmation results. For parameters without manufacturer alarm limits or explicit operation limits, the threshold is set according to the quantile value of historical normal voyage data.

[0196]

[0197] in, For the monitoring parameters whose thresholds are to be set, This is the average of historical normal voyage data. For the first quantile value For parameters Corresponding threshold. Any weighted reorganization is uniformly denoted as... }, which satisfies:

[0198]

[0199]

[0200] in, The first in the equity restructuring Each weight, This represents the number of weights. The parameter table is denoted as... Any parameter record in the parameter table is denoted as:

[0201]

[0202] in, For the nth parameter record, For parameter names, For the control link to which it belongs, Dimensions are given by the unit of measurement. For the range of values, As the data source, The sampling period is For threshold or alarm limit, To adapt to dynamic, high-risk navigation scenarios, the parameter generation module reads the parameter table. The system generates data validity status, scenario risk control amount, power chain propagation control amount, diagnostic credibility control amount, and dynamic control parameters by considering the range of values, thresholds or alarm limits, sampling period, and adapted dynamic high-risk navigation scenarios.

[0203] Example 2: Refer to Figure 4 A fault diagnosis system for ship propulsion systems based on multi-source data fusion includes:

[0204] The data acquisition module collects multi-source control and monitoring data from multiple data sources, forming a multi-source control state feature set with timestamps, link sources, and data validity status. This module serves as the foundation for aggregating the operational data of the ship's power control system. It collects multi-source control and monitoring data from multiple data sources for power plant power supply and distribution control links, propulsion control links, steering control links, dynamic positioning control links, and communication acquisition links, retaining timestamps, link sources, and data validity status during the acquisition process. The multi-source control state feature set generated by this module reflects the state differences, temporal relationships, and data reliability of different control links within the same operating period, providing a unified data foundation for generating scenario risk control quantities, cascading causal event sequences, and dynamic control parameters.

[0205] The risk generation module is used to generate scenario risk control quantities based on a multi-source control state feature set. This module is responsible for quantifying risks in dynamic high-risk navigation scenarios. It identifies dynamic high-risk navigation scenarios based on the multi-source control state feature set, establishes scenario risk coupling relationships by combining navigation scenario constraints, ship motion state, and power maneuvering capabilities, and calculates the remaining maneuvering time margin index. The scenario risk control quantities generated by this module can reflect changes in maneuvering margin, risk trends, and fault prediction sensitivity during ship navigation in bridge areas, harbors, narrow waterways, berthing operations, unberthing operations, dynamic positioning operations, and ship-wide power loss recovery processes.

[0206] The propagation generation module is used to establish a cascaded causal event sequence based on a multi-source control state feature set, and to generate power chain propagation time indices and power chain propagation control quantities based on the cascaded causal event sequence. This module is responsible for constructing the abnormal propagation relationship between control links. Based on the multi-source control state feature set, this module establishes a cascaded causal event sequence and generates power chain propagation time indices and power chain propagation control quantities by focusing on the propulsion control link, steering control link, and dynamic positioning control link of the abnormal state in the power plant's power supply and distribution control link. This module can reflect the time difference of occurrence of abnormal states between different control links, the propagation hierarchy, the degree of abnormality of the response state, and the priority of cascaded link risks, facilitating the differentiation between single-link abnormalities and multi-link cascaded abnormalities.

[0207] The parameter generation module generates dynamic control parameters based on scenario risk control quantities, power chain propagation control quantities, multi-source control state feature sets, and data validity status. This module is responsible for forming control parameters during the diagnosis, prediction, and health status evaluation processes. It integrates scenario risk control quantities, power chain propagation control quantities, multi-source control state feature sets, and data validity status to generate dynamic control parameters. Furthermore, by combining diagnostic credibility indicators and diagnostic credibility control quantities, this module constructs a risk propagation credibility control state encompassing scenario risk, power chain propagation, and diagnostic credibility dimensions, ensuring that the dynamic control parameters simultaneously include constraint information related to scenario risk, power chain propagation, and data credibility.

[0208] The result generation module is used to perform multi-source fusion diagnosis, fault prediction, and health status evaluation on multi-source control monitoring data based on the dynamic control parameters, and generate predictive health management results. This module is responsible for outputting the fault diagnosis, fault prediction, and health status evaluation results. Based on the dynamic control parameters, this module performs multi-source fusion diagnosis, fault prediction, and health status evaluation on multi-source control monitoring data, and generates predictive health management results. The predictive health management results output by this module correspond to fault output modes, which include at least one of component-level fault output modes and link-cascaded risk output modes, reflecting the fault risk, health status, and diagnostic reliability of the ship's power control system under its current operating state.

[0209] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A fault diagnosis method for ship propulsion systems based on multi-source data fusion, used for control status monitoring, fault prediction, and health status management of ship propulsion control systems. The ship propulsion control system includes a power plant power supply and distribution control link, a propulsion control link, a steering control link, a dynamic positioning control link, and a communication and data acquisition link. Its characteristic is that... Includes the following steps: S1, collects multi-source control monitoring data from multiple data sources to form a multi-source control status feature set with timestamps, link sources and data validity status; S2, generates scenario risk control quantities based on multi-source control state feature sets; S3, establish a cascaded causal event sequence based on the multi-source control state feature set; Cascaded causal event sequences are used to characterize the process by which abnormal states in the power supply and distribution control link of a power plant propagate to the propulsion control link, steering control link, and dynamic positioning control link. Based on the cascaded causal event sequence, a power chain propagation time index and a power chain propagation control quantity are generated. S4 generates dynamic control parameters based on scenario risk control quantity, power chain propagation control quantity, multi-source control state feature set, and data validity state, so as to dynamically control the multi-source fusion diagnosis process, fault prediction process, health status evaluation process, and fault output mode; the fault output mode includes at least one of component-level fault output mode and link cascade risk output mode. S5, based on the dynamic control parameters generated by S4, performs multi-source fusion diagnosis, fault prediction, and health status evaluation on the multi-source control and monitoring data collected by S1, and generates predictive health management results.

2. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 1, characterized in that, The generation of scenario risk control quantities includes: identifying dynamic high-risk navigation scenarios based on multi-source control state feature sets, establishing scenario risk coupling relationships between navigation scenario constraints, ship motion state, and power maneuvering capabilities, calculating the remaining maneuvering time margin index of the ship under dynamic high-risk navigation scenarios, and generating scenario risk control quantities based on the remaining maneuvering time margin index and the scenario risk coupling relationship.

3. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 2, characterized in that, Dynamic high-risk navigation scenarios include at least one of the following: navigation in bridge areas, navigation in harbors, navigation in narrow waterways, berthing operations, unberthing operations, dynamic positioning operations, and the process of restoring power to the entire ship after a power outage.

4. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 2, characterized in that, The generation of the remaining maneuvering time margin index includes: under dynamic high-risk navigation scenarios, determining the spatial risk margin, motion trend risk, and dynamic maneuvering capability margin based on navigation scenario constraints, ship motion state, and dynamic maneuvering capability; and generating the remaining maneuvering time margin index, which characterizes the ship's ability to maintain a safe maneuvering state, based on the spatial risk margin, motion trend risk, and dynamic maneuvering capability margin.

5. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 4, characterized in that, The generation of scenario risk control quantities includes: forming a scenario risk state vector based on the scenario type of dynamic high-risk navigation scenarios, the remaining maneuvering time margin index, the degree of compression of spatial risk avoidance margin, the rate of change of motion trend risk quantity, and the degree of decay of power maneuvering capability margin; and generating scenario risk control quantities based on the scenario risk state vector. The scenario risk control measures include at least one of the following: scenario risk level, control margin compression coefficient, scenario risk change rate, scenario control priority, and fault prediction sensitivity correction.

6. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 1, characterized in that, The generation of the power chain propagation time index and the power chain propagation control quantity includes: determining the time difference of the propagation of the abnormal state of the power supply and distribution control link to the response state of the propulsion control link, steering control link, and dynamic positioning control link based on the cascaded causal event sequence; calculating the power chain propagation speed index based on the time difference, the propagation hierarchy between the corresponding control links, and the degree of abnormality of the response state, and generating the power chain propagation time index based on the power chain propagation speed index; and generating the power chain propagation control quantity based on the cascaded causal event sequence, the power chain propagation speed index, and the power chain propagation time index.

7. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 6, characterized in that, The propagation control parameters of the power chain include at least one of the following: propagation path weight, cascade propagation rate correction, cascade critical time compression coefficient, fault source location weight, and link cascade risk priority.

8. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 7, characterized in that, In S4, the generation of dynamic control parameters includes: generating diagnostic credibility indicators and diagnostic credibility control quantities based on the multi-source control state feature set and data validity state; constructing a risk propagation credibility control state that includes the scenario risk dimension, the power chain propagation dimension, and the diagnostic credibility dimension based on the scenario risk control quantity, the power chain propagation control quantity, and the diagnostic credibility control quantity; and generating dynamic control parameters based on the risk propagation credibility control state.

9. The method for fault diagnosis of ship propulsion systems based on multi-source data fusion according to claim 8, characterized in that, The diagnostic confidence adjustment measures include at least one of the following: data source evidence weight, outlier suppression coefficient, degraded data compensation coefficient, credible evidence gating coefficient, and diagnostic output confidence correction.

10. A ship propulsion system fault diagnosis system based on multi-source data fusion, applied to the ship propulsion system fault diagnosis method based on multi-source data fusion as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect multi-source control and monitoring data from multiple data sources, forming a multi-source control status feature set with timestamps, link sources, and data validity status. The risk generation module is used to generate scenario risk control parameters based on a multi-source control state feature set. The propagation generation module is used to establish a cascaded causal event sequence based on a multi-source control state feature set, and to generate a power chain propagation time index and a power chain propagation control quantity based on the cascaded causal event sequence. The parameter generation module is used to generate dynamic control parameters based on scenario risk control quantities, power chain propagation control quantities, multi-source control state feature sets, and data validity states. The result generation module is used to perform multi-source fusion diagnosis, fault prediction, and health status evaluation on multi-source control and monitoring data based on the dynamic control parameters, and generate predictive health management results.

Images