Intelligent operation and maintenance system of smart ship

Abstract

This invention relates to the field of vessel maintenance technology, and discloses an intelligent maintenance system for smart vessels, including an onboard data acquisition unit, a communication transmission unit, a cloud-based maintenance platform, and a mobile terminal application. The onboard unit collects operational status data in real time and sends it to the cloud platform. The cloud platform performs timestamp alignment and noise reduction on the received data, performs multi-classification fusion diagnosis by combining aquatic environmental attributes and physical operating parameters; it calls an evaluation model to generate a comprehensive priority score, allocates maintenance work orders based on operator load and spatial distance; it calculates the search radius based on remaining battery power to match the optimal battery swapping station node; and it generates pre-maintenance work orders based on the equipment's historical operating load to predict remaining lifespan. The mobile terminal application controls task flow through a finite state machine and uses cryptographic hashing and digital signatures to ensure the security of feedback data. This invention improves the accuracy of system fault diagnosis and the execution efficiency of maintenance task scheduling.

Description

Technical Field

[0001] This invention relates to the field of ship maintenance technology, specifically to an intelligent maintenance system for smart ships. Background Technology

[0002] As the intelligence level of ships and vessels increases, the need for real-time monitoring and unified scheduling of their operational status is growing. However, current ship maintenance systems still have some technical shortcomings in practical applications.

[0003] Existing systems are susceptible to interference from sensor noise and network transmission delays when collecting low-level operational status data, leading to timing discrepancies and abnormal deviations in the data. In anomaly identification and fault diagnosis, current technologies often rely solely on simple threshold comparisons of single sensor physical parameters, failing to consider the vessel's surrounding water environment and multidimensional physical characteristics for comprehensive judgment. This makes diagnostic results vulnerable to external environmental interference, resulting in false alarms and insufficient objectivity and accuracy in fault diagnosis.

[0004] Existing work order processing and battery swapping scheduling mechanisms typically employ a simplistic assignment method, resulting in inefficient resource allocation. Regarding personnel scheduling, the system lacks comprehensive consideration of the operator's actual jurisdictional waters, current unfinished tasks, and the Euclidean distance between the operator and the target vessel, leading to inefficient task allocation. In the matching process for vessel refueling nodes, current technology often directly searches for the nearest device, failing to use the theoretical voyage distance calculated from the remaining battery power as a spatial retrieval constraint for comprehensive evaluation. This can result in battery swapping equipment matching exceeding the vessel's actual reachability, reducing the rationality of scheduling.

[0005] Existing equipment maintenance models are mostly reactive repairs after a failure occurs. The system cannot extrapolate and predict equipment degradation based on the cumulative load stress of historical operating data, making it difficult to generate pre-maintenance work orders for preventative maintenance. Furthermore, on-site personnel lack strict control over task flow nodes when performing maintenance tasks; and after task completion, the uploaded on-site handling feedback data lacks effective cryptographic verification and digital signature mechanisms during interactive transmission, posing a risk of data tampering and making it difficult to guarantee the integrity and security of completed business data. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an intelligent operation and maintenance system for smart vessels, which solves at least one of the following problems: existing vessel operation and maintenance systems rely solely on a single physical parameter for judgment, resulting in low diagnostic accuracy; existing vessel operation and maintenance systems have limited dimensions for anomaly identification; unreasonable task priority assessment; and low efficiency in scheduling operation and maintenance personnel.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: an intelligent operation and maintenance system for smart boats, including a shipborne data acquisition unit deployed on the boat, used to collect operating status data including positioning coordinates, battery power and communication status;

[0008] A communication transmission unit is used to send the operating status data to a cloud-based operation and maintenance platform;

[0009] The mobile terminal application, configured on the operator's end, is used to obtain real-time location and number of pending tasks, receive and display distributed system tasks, receive operation input instructions, and collect credential files containing on-site image data and send them to the cloud operation and maintenance platform.

[0010] The cloud-based operation and maintenance platform is used to receive the operating status data, perform anomaly identification based on the battery level and communication status; generate an anomaly event when an anomaly is triggered and extract the anomaly duration; generate a comprehensive priority score by weighting the anomaly duration, battery level, and location coordinates; encapsulate the anomaly event and the comprehensive priority score into a system task; calculate the spatial distance between the location coordinates and the real-time location, select the optimal operator based on the number of pending tasks, and distribute the system task to the corresponding mobile terminal application.

[0011] Preferably, the cloud-based operation and maintenance platform performs timestamp alignment processing on the received operational status data, uses a resampling algorithm to map various types of underlying data to fixed time intervals, and applies a sliding median filter or Kalman filter to denoise and clean the timestamp-aligned operational status data, identifying and removing abrupt outliers in the data sequence, and using a linear interpolation algorithm to fill in missing values. This solution eliminates temporal misalignment and disorder caused by network latency and sensor noise through standardized mapping and filtering of underlying data, providing a high-quality input data source for subsequent diagnostic models.

[0012] Preferably, the operational status data further includes motor operational status parameters; the positioning coordinates include a set of latitude and longitude coordinates and a ground speed parameter; the cloud-based operation and maintenance platform extracts the operational status data to generate a device operational feature vector and inputs it into the rule engine. The rule engine performs anomaly identification based on set multi-dimensional logical conditions and generates anomaly events when an anomaly is triggered: a low battery anomaly event is triggered when the battery level is lower than a preset threshold; a disconnection anomaly event is triggered when the number of consecutive packet losses recorded in the communication status exceeds the tolerance limit or the disconnection duration is greater than the judgment period; a boundary crossing anomaly event is triggered when the extracted latitude and longitude coordinate set exceeds the preset water area electronic fence range and the ground speed parameter is lower than the drift threshold; the corresponding anomaly event is generated through the above determinations. This solution combines multi-dimensional operational status data with the rule engine to execute logical judgments, improving the sensitivity and accuracy of triggering abnormal events on the vessel.

[0013] Preferably, when generating the comprehensive priority score, the cloud-based operations and maintenance platform extracts the penalty score converted from the abnormal duration corresponding to the triggered abnormal event, the risk score converted from the battery power, the risk weight value mapped based on the latitude and longitude coordinate set, and the fixed importance coefficient of the vessel equipment. These are then multiplied by the corresponding sub-dimensional weight factor parameters, and the results are summed to obtain the comprehensive priority score. The cloud-based operations and maintenance platform compares and matches the comprehensive priority score with the corresponding mapping table of warning levels, assigning a clear ranking identifier to the system tasks. Through this multi-parameter summation calculation mechanism, the system can convert a single hardware alarm into a quantitative security rating with a global perspective, achieving a reasonable ranking of operations and maintenance tasks.

[0014] Preferably, the cloud-based operations and maintenance platform extracts the event type of the abnormal event, generates a corresponding abnormal state classification code based on the event type, and encapsulates the abnormal state classification code and the explicit sorting identifier into the system task. The cloud-based operations and maintenance platform reads the abnormal state classification code and the latitude and longitude coordinate set in the operating status data, and filters out operators with matching jurisdictional water areas and corresponding business skill classification tags as candidate objects from its internal personnel attribute database. When multiple qualified candidate objects exist, the cloud-based operations and maintenance platform prioritizes assigning the system task to the candidate object with the fewest pending tasks. If the number of pending tasks is the same, the system task is assigned to the mobile terminal application bound to the candidate object with the closest straight-line Euclidean distance to the target vessel. This matching mechanism ensures the matching of the task execution personnel's professional skills, while optimizing spatial movement costs and personnel workload.

[0015] Preferably, when the low battery anomaly event is triggered, the cloud-based operation and maintenance platform calculates the maximum theoretical sailing distance based on the battery level and uses it as the search radius. It then performs a range query in its internal spatial geographic database to extract all battery swapping cabinets falling within the search radius, generating a candidate battery swapping cabinet set. The platform calculates a comprehensive matching score between the target vessel and each battery swapping cabinet in the candidate set. This comprehensive matching score is calculated by multiplying the scores of sub-functions in four dimensions—spatial distance, available resources, equipment health, and time cost—by their respective decision weight parameters and summing the results. The cloud-based operation and maintenance platform extracts the battery swapping cabinet with the highest comprehensive matching score as the optimal battery swapping cabinet node, performs a resource locking operation on the optimal node, and issues a scheduling instruction data packet. This solution uses battery level as a dynamic spatial constraint, and through automatic platform-triggered scheduling and multi-dimensional decision calculations, ensures the accuracy and high availability of vessel energy replenishment node scheduling.

[0016] Preferably, the cloud-based operation and maintenance platform extracts the stator current parameter from the ground speed parameter and the motor operating status parameter to construct a time-series data stream. It then uses a cumulative sum algorithm to perform abrupt change point detection and extract abnormal physical features. Next, it extracts the latitude and longitude coordinate sets corresponding to the time intervals triggering the abrupt changes to construct an abnormal loitering point dataset. A density-based spatial clustering algorithm is used to calculate the spatial geometric center coordinates of the loitering point clusters. The spatial geometric center coordinates are then intersected with the preset water area electronic fence range to extract water environment attribute labels. The change in heading angle between adjacent latitude and longitude coordinate sets is calculated. The abnormal physical features, the change in heading angle, and the water environment attribute labels are one-hot encoded and concatenated to generate a multi-dimensional feature vector. This vector is input into a pre-trained multi-class fusion diagnostic model to calculate the probability distribution of fault classification. The maximum probability value is compared with the diagnostic confidence threshold, and the confirmed fault type is output and updated in the system task. The system utilizes spatial environmental information and the rate of change of operating parameters for fusion inference, eliminating the possibility of false alarms from a single physical sensor and improving the accuracy of fault type diagnosis in complex water environments.

[0017] Preferably, the cloud-based operation and maintenance platform extracts historical operating data of the vessel's core power equipment to construct a time-series data stream of equipment operating load. Based on a single sampling period, it accumulates electrical load stress and thermal load stress over time to generate a quantified equipment health degradation index sequence. A time-series prediction model is used to extrapolate and predict the health degradation index values ​​at various future time points, calculating the remaining service life variable required to reach the critical failure threshold. It then determines whether the remaining service life variable is less than or equal to a pre-configured early warning threshold. If it is greater, continuous monitoring continues; if it is less than or equal to, the corresponding maintenance material list is queried from the central database, and the maintenance material list is integrated to generate a predictive pre-maintenance work order data package, which is then sent as a system task. By quantifying the cumulative damage of equipment loads and predicting lifespan, the system achieves a technological leap from passive maintenance to proactive preventative maintenance.

[0018] Preferably, the mobile terminal application has a built-in finite state machine engine. This engine reads the initial attributes of the system task and, based on the received operation input instructions, controls the unidirectional ordered transitions of the system task between the pending response state, the progressing state, the processing state, and the completed state. The finite state machine engine reads a constraint matrix containing the results of multimedia attachment non-empty verification and spatial coordinate distance comparison, and, in conjunction with the received operation input instructions, triggers the unidirectional ordered transitions: when the spatial coordinate distance comparison result meets a set threshold, a transition to the processing state is triggered; when the multimedia attachment non-empty verification passes, a transition to the completed state is triggered. Based on the transition results, the finite state machine engine updates the state identifier bit and asynchronously uploads the data frame containing the state change code to the cloud-based operation and maintenance platform. This mechanism, through strict logical condition control, constrains the business workflow of on-site operators, ensuring precise closed-loop control of on-site progress by the platform.

[0019] Preferably, the operation input instruction includes a completion submission instruction. After receiving the completion submission instruction, the mobile terminal application obtains the credential file, extracts the core fields of the credential file, calculates a fixed-length digital digest using a cryptographic hash function, and encrypts the digital digest using the private key parameter allocated in the local secure storage area of ​​the mobile terminal application to generate a digital signature variable. The mobile terminal application uploads the credential file with the digital signature variable attached to it to the cloud operation and maintenance platform. The cloud operation and maintenance platform decrypts the digital signature variable using the corresponding public key parameter, performs a cryptographic hash calculation on the received data to generate a local verification hash value, compares the decrypted digital digest with the local verification hash value for equality, and writes the credential file into the audit database if they match. This solution constructs a secure verification channel for on-site data upload, ensuring the tamper-proof characteristics of the business completion evidence chain.

[0020] The intelligent operation and maintenance system for smart boats provided by this invention has the following beneficial effects:

[0021] This invention constructs a closed-loop operation and maintenance (O&M) link, from status monitoring and anomaly identification to work order processing, through collaborative interaction between a cloud computing center, ship-side sensing terminals, and on-site processing terminals. The invention collects operational status data such as the vessel's positioning coordinates, battery level, and communication status via an onboard data acquisition unit and uploads it to a cloud-based O&M platform. The cloud-based O&M platform performs anomaly identification based on battery level and communication status, generating anomaly events and extracting the anomaly duration when an anomaly is triggered. It then calculates a comprehensive priority score by combining the anomaly duration, battery level, and positioning coordinates, and further selects the optimal operator and distributes tasks based on the operator's real-time location and the number of pending tasks. This achieves more accurate identification of vessel anomalies and more rational scheduling of O&M tasks, improving fault diagnosis accuracy, task response timeliness, and overall O&M efficiency.

[0022] This invention removes sensor noise and outlier interference by aligning and filtering the underlying operational status data with timestamps. At the same time, it uses a density-based spatial clustering algorithm to extract aquatic environmental tags and performs multi-dimensional feature splicing and fusion diagnosis of abnormal physical features, changes in heading angle, and environmental attributes. This eliminates false alarms caused by abnormal indicators of a single physical sensor and improves the objectivity and accuracy of diagnosing vessel fault types.

[0023] This invention introduces a multi-dimensional comprehensive scoring mechanism into the system task flow and battery swapping scheduling. Work orders are allocated based on the operator's jurisdictional waters, the current unfinished task load, and the spatial Euclidean distance. At the same time, the theoretical navigation distance converted from the remaining battery power is used as a spatial retrieval constraint to calculate the comprehensive matching score of the battery swapping cabinet equipment. This optimizes the allocation logic of personnel scheduling and improves the rationality and execution efficiency of matching vessel refueling nodes.

[0024] This invention uses a load stress accumulation and autoregressive integral moving average model based on historical operating data to extrapolate the equipment attenuation index and generate pre-maintenance work orders in advance, thus realizing preventive maintenance of the equipment. In addition, the task flow nodes are controlled by the finite state machine built into the mobile terminal, and hash calculation and digital signature technology are used to perform bidirectional verification of the on-site handling feedback data packets, which standardizes the flow sequence of on-site operations and ensures the integrity and security of the completion certificate data during interactive transmission. Attached Figure Description

[0025] Figure 1 This is an overall architecture diagram of an intelligent operation and maintenance system for a smart vessel according to an embodiment of the present invention;

[0026] Figure 2 This is a flowchart illustrating the data processing and business execution process of the intelligent operation and maintenance system according to an embodiment of the present invention.

[0027] Figure 3 This is a schematic diagram of the login interface of a mobile terminal application according to an embodiment of the present invention;

[0028] Figure 4 This is a schematic diagram of the main interface of the mobile terminal application's workbench according to an embodiment of the present invention;

[0029] Figure 5 This is a schematic diagram of the task center list interface of a mobile terminal application according to an embodiment of the present invention;

[0030] Figure 6 This is a schematic diagram of the interface for a mobile terminal application to perform battery swapping scheduling and matching according to an embodiment of the present invention.

[0031] Figure 7 This is a schematic diagram of the task on-site handling and feedback interface of a mobile terminal application according to an embodiment of the present invention;

[0032] Figure 8 This is a schematic diagram of the multi-dimensional spatial scheduling and control of intelligent vessels according to an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions in the embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings in this specification.

[0034] See Figure 1This invention provides an intelligent operation and maintenance system for smart vessels, comprising: an onboard data acquisition unit, a communication transmission unit, a cloud-based operation and maintenance platform, and a mobile terminal application.

[0035] The shipborne data acquisition unit is deployed on each operating vessel within the waterway. This unit establishes a data connection with the vessel's internal battery management system and motor controller. Following a set sampling period, the shipborne data acquisition unit collects real-time operational status data of the vessel. This data includes positioning coordinates, remaining battery power, voltage data, current data, motor operating parameters, and communication heartbeat status. A communication transmission unit connects the shipborne data acquisition unit to a cloud-based operations and maintenance platform. The cloud-based platform receives and processes the data, generates system tasks, and distributes them to mobile terminal applications for on-site handling.

[0036] like Figure 2 As shown, this embodiment of the invention provides a smart vessel intelligent operation and maintenance method based on the above system, including the following steps:

[0037] S10 collects the vessel's operating status data in real time according to the set sampling period; the operating status data includes positioning coordinates, remaining battery power, voltage data, current data, motor operating status parameters, and communication heartbeat status.

[0038] The S20 uses wireless network communication technology to package the collected operating status data, attach a timestamp, and send it to the cloud operation and maintenance platform via wireless network.

[0039] S30: The cloud-based operation and maintenance platform receives and stores reported operational status data, performs format parsing, noise reduction and cleaning, and timestamp alignment on multi-source data to generate structured equipment operation feature vectors; based on the rule engine and data analysis model, it performs at least one of anomaly identification, resource scheduling calculation, and equipment health assessment on the equipment operation feature vectors.

[0040] S40: The cloud-based operations and maintenance platform identifies anomalies in the device operation feature vector generated by S30 based on the rule engine and data analysis model. When the feature parameters corresponding to the identification result meet the preset alarm threshold or combined triggering conditions, the identification result is determined as an abnormal event, and the duration of the anomaly is extracted. A comprehensive priority score is generated by weighting the anomaly duration, battery level, and location coordinates, and the abnormal event and the comprehensive priority score are encapsulated into a structured system task. The system task may also include the location coordinates of the triggering anomaly, the occurrence time, and anomaly-related data indicators.

[0041] S50: For the system tasks generated by S40, the cloud-based operation and maintenance platform obtains the real-time location feedback from the mobile terminal application of each operator and the number of pending tasks for each operator; calculates the spatial distance between the location coordinates of the triggered anomaly and the real-time location of each operator, and selects the optimal operator based on the number of pending tasks for each operator; and distributes the structured system tasks to the mobile terminal application corresponding to the optimal operator.

[0042] In an alternative embodiment, in step S50, the cloud-based operations and maintenance platform obtains the real-time location feedback from the mobile terminal application corresponding to each operator and the number of pending tasks for each operator for the system task generated in S40; determines candidate responsibility jurisdiction areas based on the location coordinates of the triggered anomaly and the pre-configured jurisdictional water boundary information in the background; calculates the spatial distance between the location coordinates of the triggered anomaly and the real-time location of each operator among the operators corresponding to the candidate responsibility jurisdiction areas; and selects the optimal operator based on the number of pending tasks for each operator; and distributes the structured system task to the mobile terminal application corresponding to the optimal operator.

[0043] S60, the mobile terminal application obtains the real-time location of the operator and counts or obtains the number of tasks to be done. The real-time location and the number of tasks to be done are sent to the cloud operation and maintenance platform for the cloud operation and maintenance platform to perform task distribution. After receiving the system tasks pushed by the cloud operation and maintenance platform, the task list is classified and displayed in the interface according to the comprehensive priority score and task type, and the current location of the vessel, details of abnormal events and handling guidance steps are displayed.

[0044] S70 receives operation input commands from on-site maintenance personnel; after completing on-site troubleshooting, repair or battery replacement operations, it calls the hardware interface to collect on-site image data as evidence, and sends or transmits the evidence file and processing record text to the cloud maintenance platform.

[0045] The S80 cloud-based operations and maintenance platform receives the returned processing credentials and result feedback, updates the task's workflow status, changes the workflow status from pending to completed, and removes the abnormal flag from the corresponding vessel's operating status; it also associates and stores the processing data, operator accounts, processing timestamps, and credential files to complete the closed-loop archiving of operations and maintenance business data.

[0046] The shipborne data acquisition unit establishes a hardware communication connection with the battery management system and the motor controller. This hardware communication connection uses a controller area network (CLAN) bus or a serial communication interface. For the underlying hardware wiring and communication protocol parsing of the CLAN bus and serial communication interface, those skilled in the art can refer to existing industry standards for setup; the specific implementation methods are well-known technologies in this field and will not be elaborated upon here.

[0047] The shipborne data acquisition unit acquires underlying operational data according to a set sampling period. The acquired data parameters are categorized into power battery parameters, motor drive parameters, spatial position parameters, and communication link parameters. Power battery parameters include individual cell voltage, bus current, cell temperature, and remaining battery capacity. Motor drive parameters include the three-phase current output by the motor controller, bus voltage, motor speed, and fault diagnostic codes. Spatial position parameters are generated by the shipborne satellite positioning module and include longitude coordinates, latitude coordinates, and ground speed. Communication link parameters include the round-trip latency of data packets and the heartbeat connectivity status between the shipborne terminal and the cloud-based operations and maintenance platform.

[0048] The cloud-based operations and maintenance platform receives multi-source operational status data continuously reported by the shipborne data acquisition unit. Due to differences in the hardware sampling frequencies of different sensors, the cloud-based platform performs timestamp alignment on the received heterogeneous data. Timestamp alignment employs a resampling algorithm, using the cloud-based operations and maintenance platform's system clock as a global reference, mapping various underlying data to fixed time intervals. The step size of these time intervals is set according to the system's early warning response requirements.

[0049] The cloud-based operations and maintenance platform performs noise reduction and cleaning, as well as missing value imputation, on the time-aligned data to eliminate the impact of transient physical interference from sensors on subsequent early warning calculations. Noise reduction and cleaning employs a sliding median filter or a Kalman filter to identify and remove abrupt outliers exceeding physical limits in the data sequence. When outliers are removed or data loss occurs due to network jitter, the system triggers a missing value imputation mechanism. When the number of consecutively lost data points is less than a preset threshold, the system uses a linear interpolation algorithm for numerical fitting and imputation; when the number of consecutively lost data points is greater than or equal to the preset threshold, the system marks all data within that time period as invalid to prevent erroneous data from participating in early warning determination.

[0050] After time alignment and quality cleaning, the cloud-based operations and maintenance platform encapsulates fragmented sensor data into structured device operation feature vectors. (Definition of the vessel / boat) exist The running state vector at time t is Its construction formula is:

[0051] ;

[0052] In the formula, Representative boats exist The remaining battery percentage at any given time; Representative boats exist Speed ​​relative to ground at any given moment; Representative boats exist The set of latitude and longitude coordinates at any given moment; Representative boats exist The communication heartbeat status parameter at any given time can be represented by a Boolean value to indicate whether the connection is established or disconnected, or by an integer value to record the number of consecutive packet losses. Representative boats exist The motor operating status parameter matrix at any given time, covering stator current and rotor speed; A unique equipment identification number representing the target operating vessel accessing the system within the waters; This represents the time-series nodes corresponding to the underlying operational data acquired by the shipborne data acquisition unit. The cloud-based operations and maintenance platform will then use the operational status vector... It is stored in a time-series database and serves as the standard input source for subsequent anomaly identification rule engine and hierarchical early warning model.

[0053] The cloud-based operations and maintenance platform uses the generated structured operational status vector as input data and transmits it to the internally configured rule engine. Based on pre-defined multi-dimensional logical conditions, the rule engine performs real-time judgment and comparison on various parameters in the operational status vector to identify abnormal events during the vessel's operation.

[0054] The rule engine sets independent triggering logic for different types of abnormal events. When the remaining battery power parameter is lower than a preset safe percentage threshold, the rule engine determines that a low battery abnormal event is triggered. When the number of consecutive packet losses recorded by the communication heartbeat status parameter exceeds the system's set tolerance limit, or when the duration of the heartbeat disconnection is longer than a preset judgment period, the rule engine determines that a disconnection abnormal event is triggered.

[0055] The cloud-based operations and maintenance platform constructs the boundary coordinate model of the waterway electronic fence in memory. The rule engine extracts the latitude and longitude coordinate set from the running status vector and calculates whether the coordinates exceed the physical range defined by the boundary coordinate model. For the underlying geometric algorithm determining the spatial location and polygon boundary inclusion relationship, those skilled in the art can refer to existing geographic information system standards and specifications for development and configuration; this is well-known technology in the field and will not be elaborated upon here. When the coordinates exceed the boundary limit and the ground speed parameter is lower than the set drift threshold, the rule engine determines that a boundary crossing anomaly event has been triggered.

[0056] After an abnormal event is triggered, the cloud-based operations and maintenance platform invokes a tiered early warning and assessment model to quantitatively analyze the event. This model comprehensively considers the severity of the abnormal state and environmental risks, calculating a comprehensive priority score to determine the order in which the system handles the event. The formula for calculating the comprehensive priority score is as follows:

[0057] ;

[0058] In the formula, Representative target vessel A comprehensive priority score for abnormal events triggered at the current calculation moment; This variable represents the duration of the abnormal event from the time it was triggered until the current time. The time penalty function is represented by a non-linear monotonically increasing mathematical model. The larger the value of the duration variable, the higher the output penalty score. This represents the remaining battery power value recorded by the system at the time the anomaly was triggered; This represents a low battery risk assessment function, whose output is negatively correlated with the battery level; that is, the lower the battery level, the higher the risk score. A set of spatial latitude and longitude coordinates representing the moment the anomaly was triggered; The risk level mapping function represents spatial location. The system assigns weight values ​​according to the electronic fence attributes of the water area corresponding to the coordinates. The weight value of the location in the deep water area or the restricted area is greater than the weight value of the location in the near-shore berthing area. The fixed importance coefficient representing the equipment on the vessel is pre-configured in the database based on the asset type and passenger capacity of the vessel. , , , The system represents the weighting factors of various sub-dimensions that are independently set by the system based on the actual needs of water area operation and management.

[0059] The cloud-based operations and maintenance platform establishes a mapping table between comprehensive priority scores and warning levels. The platform extracts the calculated comprehensive priority score and compares it with the specified ranges in the mapping table. When the score falls into the highest-level range, the platform marks the corresponding event as a high-risk outage or serious fault warning; when the score falls into the intermediate range, it marks it as a low-battery swap warning requiring follow-up scheduling; and when the score falls into the basic range, it marks it as a regular boundary violation warning. This quantitatively calculated mapping mechanism ensures that even when multiple vessels experience concurrent anomalies, the system outputs a list of warning events with clear ordering identifiers.

[0060] After obtaining the list of warning events with comprehensive priority scores, the cloud-based operations and maintenance platform initiates an automated mapping process for structured work orders. The platform extracts key data fields from the warning events, including the unique equipment identifier of the vessel triggering the anomaly, the current latitude and longitude coordinates, the anomaly status classification code, the duration variable, and the system timestamp of the occurrence. The platform then encapsulates and combines these extracted key data fields into a standard-format structured work order data package according to a preset database mapping template. This data package includes a unique task tracking sequence number automatically generated by the system and the comprehensive priority score corresponding to the warning event.

[0061] The cloud-based operations and maintenance platform maintains an attribute database for on-site operations and maintenance personnel. This database records the account identification code, business skill category tags, and the area of ​​water under their jurisdiction for each on-site operator. Business skill category tags correspond to different types of abnormal event handling permissions, such as electrical repair, mechanical maintenance, or battery replacement. The area of ​​water under jurisdiction is physically bounded in the system using a set of polygonal spatial electronic fence coordinates. For the underlying code implementation of the account permission table structure and basic information management, those skilled in the art can refer to existing relational database design specifications; its construction pattern is well-known in the field and will not be elaborated upon here.

[0062] The cloud-based operations and maintenance platform performs task dispatch and matching calculations based on the generated structured work order data packets. The platform reads the latitude and longitude coordinates from the work order data packet and traverses the jurisdictional water areas of each operator in the attribute database. The platform uses a ray-crossing algorithm or a loop-counting algorithm to compare whether the coordinate data falls within a certain spatial electronic fence polygon, thereby determining the spatial responsibility of the target vessel. After spatial matching is completed, the platform reads the abnormal status classification codes from the work order data packet and selects operators with corresponding business skill classification tags as candidate objects from the set of personnel whose spatial responsibility has been determined.

[0063] When multiple qualified operators exist within the pool of candidates matching space and skills, the cloud-based operations and maintenance platform triggers its internal load balancing mechanism. The platform retrieves the total number of pending work orders under each candidate's current account in real time and performs a numerical comparison. The platform prioritizes assigning the currently generated structured work order to the candidate with the fewest pending work orders. If multiple candidates have the same total number of pending work orders, the platform calculates the linear Euclidean distance between the work order's latitude and longitude coordinates and the candidate's mobile terminal's real-time coordinates, assigning the work order to the operator closest to the target vessel.

[0064] After the cloud-based operations and maintenance platform determines the account of the person ultimately receiving the task, it calls the message push interface to send the task data stream. The platform sends the structured work order data packet to the mobile terminal application bound to the corresponding operator via wireless network. The mobile terminal application persistently records the received work order data packet in its local storage device and updates the list of tasks to be processed in the application interface. The cloud-based operations and maintenance platform synchronously modifies the status flag of the task record in the central database, changing the status flag from the initial pending status to the dispatched status, thereby completing the closed loop of automated business data flow from the triggering of underlying hardware anomalies to the front-end personnel's response and instruction reception.

[0065] The mobile terminal application provides a one-click battery swap trigger control in its user interface. When on-site maintenance personnel input a trigger command through the interactive interface, or when the cloud-based maintenance platform detects that the target vessel's remaining battery power is below a preset safety threshold, the system generates a battery swap request data packet. This data packet contains the target vessel's unique device identifier, real-time latitude and longitude coordinates, and the current remaining battery power percentage. The mobile terminal application or the underlying early warning module sends this battery swap request data packet to the cloud-based maintenance platform's scheduling interface. For the underlying communication protocol parsing of application-layer data requests and platform interface responses, those skilled in the art can refer to existing network transport layer protocol specifications for deployment and configuration; this is well-known technology in the field and will not be elaborated upon here.

[0066] The cloud-based operations and maintenance platform receives and parses the battery swapping request data packet, extracting the target vessel's real-time latitude and longitude coordinates and remaining battery power. Based on the current remaining battery power and the vessel's historical energy consumption baseline, the platform calculates the maximum theoretical voyage distance the vessel can currently support. Using the real-time latitude and longitude coordinates as the central origin and the maximum theoretical voyage distance as the search radius, the platform performs a range query within its internal spatial geographic database. The platform extracts all battery swapping cabinet devices falling within this search radius, constructing an initial set of battery swapping cabinet nodes. This dynamic spatial constraint mechanism based on actual remaining battery power avoids including remote devices exceeding the vessel's range in the calculation.

[0067] The cloud-based operations and maintenance platform performs hard-constraint screening on the operational status of each device in the initial set of battery swapping cabinet nodes. The platform queries the IoT database to obtain the current communication heartbeat status, underlying hardware fault codes, and real-time sensor data from each battery compartment for each swapping cabinet. The platform introduces device status indicator variables. Used to characterize the battery swapping cabinet Basic availability. When the communication link of the battery swapping cabinet j is in a normal online state and no hardware fault codes hindering the operation of the mechanical mechanism are reported to the cloud, the system assigns a value. When the device is offline or has a mechanical failure, the system assigns a value. .

[0068] The cloud-based operation and maintenance platform synchronously reads the battery swapping cabinet. Charging and discharging data uploaded by the internal battery management system. Platform statistics for the battery swapping cabinet. The number of currently fully charged batteries that are not locked by other concurrent battery swapping tasks. The logic for determining a fully charged state is that the state of charge (SOC) value of the batteries in a single battery compartment reaches the usable threshold set by the system.

[0069] The cloud-based operations and maintenance platform applies the aforementioned hard constraints to the initial set of battery swapping cabinet nodes for data filtering. When a battery swapping cabinet simultaneously satisfies the equipment status indicator variables... And the number of available batteries At that time, the platform retains it in the candidate sequence. The platform removes device nodes that do not meet any of the above hard constraints, and finally generates a candidate battery swapping cabinet set C. Each battery swapping cabinet in set C... All possess the physical hardware capabilities to immediately replace the batteries for the target vessels. The cloud-based operations and maintenance platform writes the candidate battery swapping cabinet set C into an in-memory database, serving as the input benchmark for subsequent multi-objective weighted intelligent matching algorithms.

[0070] After generating a set of candidate battery swapping cabinets, the cloud-based operations and maintenance platform calls a multi-objective weighted intelligent matching algorithm to quantitatively evaluate each battery swapping cabinet node in the set. The algorithm traverses each battery swapping cabinet in the candidate set and calculates the comprehensive matching score between the target vessel and the battery swapping cabinet. The comprehensive matching score consists of normalized sub-functions in four dimensions, covering spatial distance, available resources, equipment health, and time cost.

[0071] The cloud-based operations and maintenance platform calculates the score of a sub-function representing the spatial distance dimension. The platform extracts the real-time latitude and longitude coordinates of the target vessel and the fixed coordinates of the target battery swapping station. Based on a waterway electronic fence map model, the platform uses a path planning algorithm to calculate the planned route distance between two points. For the grid partitioning and shortest path search in the path planning algorithm, those skilled in the art can refer to the A* or Dijkstra algorithm for deployment; their underlying pathfinding logic is well-known technology in the field and will not be elaborated here. The spatial distance sub-function adopts an exponential decay model; the shorter the planned route distance, the higher the score.

[0072] The cloud-based operations and maintenance platform calculates the sub-function score for the available resource dimension. The platform reads the number of fully charged batteries currently available in the target battery swapping cabinet. The system uses a linear normalization function to map the number of fully charged batteries for calculation; the more fully charged batteries available, the higher the corresponding resource dimension score.

[0073] The cloud-based operations and maintenance platform calculates a sub-function score for the device health dimension. The platform retrieves historical charge-discharge cycle records and internal resistance test data of available batteries in the target battery swapping cabinet via an IoT interface to calculate the average health status level of the available batteries. The platform assigns higher priority scores to battery packs with higher average health status levels to ensure the stability of the boat's range after battery swapping.

[0074] The cloud-based operations and maintenance platform calculates the score for a sub-function related to time cost. The platform counts the number of battery swapping reservations received but not yet completed at the target battery swapping station, and, combined with the standard time for a single battery swapping operation, estimates the current queuing time for that station. The time cost sub-function uses a linear penalty model; the longer the queuing time, the greater the penalty.

[0075] The formula for calculating the overall matching score is:

[0076] ;

[0077] In the formula, Representative target vessel Battery swapping cabinets in the candidate battery swapping cabinet set The overall matching score between them; Representative target vessel With battery swapping cabinet The planned distance between the routes; The representative spatial distance sub-function has the following mathematical expression: , Represents the distance attenuation coefficient; Represents battery swapping cabinet The number of fully charged batteries currently available; The sub-function representing available resources is a linearly normalized function; Represents battery swapping cabinet Average health status level of the internally available batteries; Represents the device health sub-function; Represents battery swapping cabinet Current queue waiting time; The time cost sub-function is represented by the mathematical expression: Represents the time penalty coefficient; , , , The decision weight parameters representing the four dimensions of system configuration satisfy... Physical constraints.

[0078] After the cloud-based operations and maintenance platform calculates the comprehensive matching score for all devices within the candidate battery swapping cabinet set, it performs optimal solution selection. The platform extracts the battery swapping cabinet with the highest comprehensive matching score as the optimal battery swapping cabinet node. The objective function to be solved is:

[0079] ;

[0080] In the formula, This represents the value of the variable that maximizes the objective function; This indicates the constraints that the formula needs to satisfy; Representative of the candidate battery swapping cabinets; This variable represents the equipment status indicator. It takes the value of 1 when the battery swapping cabinet is online and without faults, and 0 otherwise. This represents the number of fully charged batteries currently available at the optimal battery swapping station. The cloud-based operations and maintenance platform uses the calculated optimal battery swapping station node as the target execution location for the battery swapping task and triggers subsequent resource locking and scheduling command issuance procedures. A multi-dimensional weighted matching calculation mechanism balances the battery swapping efficiency of an individual vessel with the energy network load of the entire waterway.

[0081] After acquiring the optimal battery swapping cabinet node, the cloud-based operations and maintenance platform triggers the underlying concurrent control program. The platform generates a corresponding battery swapping reservation record in the central database. The battery swapping reservation record includes the unique equipment identifier of the target vessel, the equipment number of the optimal battery swapping cabinet, and the assigned target battery compartment identifier.

[0082] The cloud-based operations and maintenance platform performs a distributed lock resource locking operation on the target battery compartment to prevent multiple concurrent battery swapping requests from repeatedly allocating the same fully charged battery. The platform calls atomic operation instructions from the in-memory database to bind the target battery compartment identifier code with the unique tracking number of the battery swapping reservation record. For the underlying code logic of performing distributed locks and atomic operations based on the in-memory database, those skilled in the art can refer to existing high-concurrency system architecture specifications for writing such code; it is well-known technology in the field and will not be elaborated upon here.

[0083] After resource locking is completed, the cloud-based operations and maintenance platform synchronously updates the real-time status data of the optimal battery swapping cabinet in the global cache. The platform decrements the number of fully charged batteries currently available for the cabinet by one. This decrement operation changes the available resource input parameter when performing multi-target matching calculations for subsequent battery swapping requests, maintaining the dynamic consistency of global battery swapping resource data.

[0084] The cloud-based operations and maintenance platform sets a valid time threshold for locked battery swapping reservation records. This threshold is calculated based on the planned route distance from the target vessel to the optimal battery swapping station and the vessel's current ground speed, with a fixed time tolerance parameter added. If the system clock exceeds this valid time threshold and the platform does not receive a physical battery swapping action trigger signal from the optimal battery swapping station, the cloud-based operations and maintenance platform determines that the battery swapping task has timed out. The platform automatically releases the distributed lock on the target battery compartment, resets the available status of the fully charged battery, and synchronously restores the cached available battery quantity parameter.

[0085] The cloud-based operations and maintenance platform sends a scheduling instruction data packet containing the coordinates of the battery swapping cabinet, the target battery compartment identifier, and the effective time threshold to the mobile terminal application of the corresponding operations and maintenance personnel on the target vessel. The mobile terminal application receives and parses the scheduling instruction data packet, rendering the physical location coordinates of the target battery swapping cabinet in the operation interface. The mobile terminal application calls its built-in map service component to generate a waterway navigation path, starting from the vessel's real-time location and ending at the optimal battery swapping cabinet location. The waterway navigation path data is graphically presented as a vector trajectory on the interface, guiding on-site operations and maintenance personnel or vessel drivers to the target battery swapping cabinet to perform the physical battery replacement operation. Upon arrival at the site, the operations and maintenance personnel scan the hardware identification code of the target battery compartment using the mobile terminal application, sending a physical replacement confirmation instruction to the cloud-based operations and maintenance platform, completing the closed loop from cloud-based scheduling calculation to on-site resource acquisition.

[0086] The auxiliary diagnostic module of the cloud-based operations and maintenance platform acquires continuous equipment operation feature vectors from the time-series database. The system extracts the ground speed parameter and the stator current parameter from the motor drive parameters from the feature vectors, and constructs one-dimensional time-series data streams respectively.

[0087] The system employs a cumulative sum algorithm to perform abrupt change detection on the time-series data stream, identifying moments when the vessel's operating parameters deviate from the normal baseline. The underlying iterative loop logic and array traversal operations of the cumulative sum algorithm can be coded by those skilled in the art with reference to existing statistical signal processing standards; these are well-known techniques in the field and will not be elaborated upon here.

[0088] The system defines the physical quantities at time [time]. The sampled value The system calculates the local mean of the physical quantity within the current sliding time window based on a set sliding time window. with standard deviation System calculation time Standardized deviation parameter The calculation formula is: The system calculates the positive offset accumulation and parameters respectively. With negative offset cumulative sum parameter This is used to capture sudden increases and decreases in time-series data. The calculation formula is:

[0089] ;

[0090] ;

[0091] In the formula, Represents the previous moment Positive offset cumulative sum; Represents the previous moment Negative offset accumulation and parameters; system settings in the initial state. and ; This parameter represents the drift tolerance. Configuring this parameter is used to filter out minor data oscillations caused by normal water flow fluctuations or wind and wave interference, thus preventing the algorithm from being falsely triggered.

[0092] The system sets a threshold for detecting mutations. The system compares the cumulative sum parameter calculated in real time with the judgment threshold. When the cumulative sum parameter shifts positively... At that time, the system records the time. This is a positive mutation point, and the cumulative sum parameter is reset to zero; when the cumulative sum parameter is shifted negatively... At that time, the system records the time. This is a negative mutation point; the cumulative sum parameter is synchronously reset to zero.

[0093] The system combines the records of abrupt changes in ground speed parameters and stator current parameters on the same time axis to extract underlying abnormal physical characteristics. When the ground speed parameter triggers a negative abrupt change point, and the stator current parameter triggers a positive abrupt change point within the same time window, the system extracts a high-load stagnation characteristic of the power system. This characteristic characterizes the physical phenomenon of an abnormal increase in the output current of the ship's motors while the speed decreases, mapping to mechanical states such as propeller entanglement or drive shaft jamming.

[0094] When the ground speed parameter triggers a negative abrupt change, while the stator current parameter within the same time window does not trigger an abrupt change, and the three-phase current and bus voltage reported by the motor controller remain within the set range, the system extracts the abnormal speed reduction characteristics of the non-power source. The system stores the extracted high-load stagnation characteristics of the power system and the abnormal speed reduction characteristics of the non-power source, along with their corresponding timestamps, into a cache as temporal feature inputs for subsequent spatiotemporal trajectory fusion and multi-classification diagnostic models.

[0095] After acquiring the abrupt change characteristics of the time-series data, the cloud-based operations and maintenance platform synchronously extracts the spatial trajectory data of the vessel within the corresponding time window. The platform retrieves the continuous set of latitude and longitude coordinates of the target vessel from the time-series database and constructs a discrete sequence of spatial trajectory points.

[0096] The platform calculates the change in heading angle between adjacent trajectory points to quantify the vessel's heading stability. (Time point defined) The latitude and longitude coordinates are ,in and Representing time respectively The longitude and latitude values; and Indicates the time of the previous adjacent trajectory point ( The latitude and longitude coordinates of ( ). Time The heading vector is .

[0097] The platform calculates the angle between two consecutive heading vectors using the dot product formula, which is used as the change in heading angle. The calculation formula is as follows:

[0098] ;

[0099] In the formula, Represents the current moment The heading vector; Represents the previous moment The heading vector; and These represent the magnitudes of the corresponding vectors. The change in heading angle within the time window... When the variance exceeds the set heading deflection threshold, the platform extracts heading loss-of-control features.

[0100] The platform extracts the set of latitude and longitude coordinates corresponding to the time intervals where negative mutations in ground speed parameters are triggered, and constructs this set as an anomalous loitering point dataset. The system calls a density-based spatial clustering algorithm to calculate the spatial distribution of the anomalous loitering point dataset. For the core point determination and neighborhood expansion distance threshold setting logic in the density-based spatial clustering algorithm, those skilled in the art can refer to standard data mining clustering algorithm specifications for deployment; these are well-known techniques in the field and will not be elaborated upon here. The system eliminates scattered coordinate drift noise through clustering calculations, generates effective loitering point clusters, and uses the arithmetic mean method to calculate the spatial geometric center coordinates of each loitering point cluster.

[0101] The platform retrieves a pre-set water area electronic fence model from the central database. The water area electronic fence model consists of closed geographic polygons formed by multiple latitude and longitude coordinate points. The system pre-binds water environment attribute tags to each closed geographic polygon. These environmental attribute tags include shallow water areas, areas with dense aquatic vegetation, reef areas, and regular main channels.

[0102] The system performs a spatial intersection operation between the generated cluster of lingering points' geometric center coordinates and the water area electronic fence model. The system uses a ray intersection algorithm to determine whether the geometric center coordinates fall inside a closed polygon with environmental attribute labels. The system emits a ray in a single direction from the geometric center coordinates and counts the number of intersections between the ray and the polygon boundary. When the number of intersections is odd, the system determines that the geometric center coordinates fall inside the current polygon and extracts the environmental attribute label bound to that polygon.

[0103] The system performs multi-dimensional logical cross-fusion of extracted spatial environmental attribute labels and physical features output by the time-series data mutation detection module. When the time-series data outputs characteristics of high-load stagnation of the power system, and the environmental attribute label extracted by the spatial intersection operation is a densely planted area, the system combines the two to generate an auxiliary diagnostic conclusion of propeller entanglement in aquatic plants. When the time-series data outputs characteristics of abnormal deceleration of non-power sources, and the extracted environmental attribute label is a shallow area, the system generates an auxiliary diagnostic conclusion of physical grounding of the hull. The underlying fusion mechanism of multi-dimensional data provides a clear data support foundation for maintenance personnel to dispatch skill tags in the front-end work order system.

[0104] The cloud-based operations and maintenance platform constructs multi-dimensional feature vectors, which serve as the standard computational input for the multi-classification fusion diagnostic model. The platform extracts anomalous physical features from the time-series data mutation point detection module, heading angle changes from the spatial trajectory module, and aquatic environmental attribute labels extracted from geographic intersection operations.

[0105] The platform performs numerical transformation on the aforementioned heterogeneous feature data. It performs one-hot encoding on discrete state variables such as the direction of sudden changes in stator current and speed, as well as environmental attribute labels such as dense vegetation areas and shallow water areas, converting them into binary numerical sequences. The platform then concatenates the continuous course deflection variance with the processed binary numerical sequence to generate a normalized multidimensional feature vector.

[0106] The platform inputs the normalized multidimensional feature vector into the internally configured multi-class fusion diagnostic model. The multi-class fusion diagnostic model is constructed using a multi-layer neural network with a logistic regression classifier to calculate the probability distribution of the current multidimensional feature vector belonging to each known fault type.

[0107] The model calculates the feature vector belonging to the first... Predicted probability of fault states The calculation formula is as follows:

[0108] ;

[0109] In the formula, A multidimensional feature vector representing the fusion of temporal and spatial attributes; This represents the total number of predefined fault categories in the system database. The specific fault categories include mechanical repairs, ship grounding, motor overload faults, and battery management system communication anomalies. This represents the index of the target fault category currently being calculated, and its value ranges from 1 to... Positive integers; The corresponding number in the representative model structure The classifier weight vector for each type of fault; Represents the corresponding number Bias parameters for fault types; This represents the transpose operation of the weight vector; Represents an exponential function with the natural constant as its base; The feature vectors calculated by the representative model Belongs to the Linear raw scores for fault-like states; This represents the index used to iterate through all fault categories when calculating the denominator; and Representing the corresponding number Classifier weight vector and bias term parameters for fault types; Representative eigenvector Belongs to the Linear raw scores for fault-like states; Represents all The indexed scores of each fault category are summed.

[0110] The system obtains the set of predicted probability distributions output by the model. The platform extracts the maximum probability value from the distribution set and compares it with the system's set diagnostic confidence threshold. When the maximum probability value is greater than or equal to the diagnostic confidence threshold, the platform outputs the fault category index corresponding to that probability as the confirmed fault type of the target vessel. When the maximum probability value is less than the diagnostic confidence threshold, the platform determines that the current data features are insufficient to support automated diagnosis, marks the result as an unknown anomaly, and generates a manual review warning on the platform interface.

[0111] The platform establishes a mapping table between confirmed fault types and business skill classification tags in the central database. When the confirmed fault type is mechanical entanglement, the platform retrieves the mechanical cleaning skill tag from the mapping table; when the confirmed fault type is motor overload, the platform retrieves the electrical maintenance skill tag. The system encapsulates the confirmed fault type name, the characteristic data fragments that triggered the anomaly, and the retrieved business skill classification tags into the extended fields of the structured work order data package. The cloud-based operations and maintenance platform uses this skill classification tag as a hard constraint to trigger the underlying automated work order dispatch program, accurately assigning the task to operations and maintenance personnel with the corresponding maintenance qualifications. Based on a multi-dimensional data fusion model diagnostic mechanism, a single alarm prompt is transformed into a closed-loop task with clear handling instructions.

[0112] For backpropagation training, cross-entropy loss function calculation, and gradient descent update of network weight parameters of multi-class fusion diagnostic models on historical sample datasets, those skilled in the art can refer to existing deep learning computing framework standards and specifications for model tuning and deployment. The underlying network training logic is a well-known technology in this field and will not be elaborated here.

[0113] The cloud-based operations and maintenance platform acquires historical operating data of the ship's core power equipment from the time-series database. This core power equipment includes the propulsion motor and the battery pack. The platform extracts parameters such as motor stator current, motor operating temperature, battery discharge rate, and cell temperature within a set time window to construct a time-series data stream of the equipment's operating load.

[0114] The cloud-based operations and maintenance platform performs load stress quantification calculations on the time-series data stream of the operating load. Normal rated operating conditions result in minimal wear and tear on equipment lifespan, but prolonged overload operation or exceeding temperature limits accelerates the aging of physical components. The platform incorporates theories of physical fatigue damage and thermodynamic aging, mapping the collected underlying operating physical parameters into incremental health degradation over time.

[0115] For the rainflow counting algorithm for the fatigue cycle of equipment materials and the underlying logic derivation of the basic thermodynamic life model, those skilled in the art can refer to existing mechanical reliability analysis and electrical engineering specifications for code deployment, which are well-known technologies in this field and will not be elaborated here.

[0116] The system establishes a discretized equipment health degradation index model. The system uses a single sampling period as the calculation step size to numerically accumulate electrical and thermal load stresses over time. The formula for calculating the equipment health degradation index is:

[0117] ;

[0118] In the formula, Represents the target equipment's cumulative operating time The health degradation index at any given time shows a monotonically increasing trend as the equipment continues to operate. This represents the initial health degradation index of the target device, which is zero when the device is brand new and manufactured. Represents the first to the second The data from each discrete sampling node are accumulated and summed. Represents the index of a discrete sampling node in a time series; The variable represents the time interval between two adjacent sampling nodes; the part within square brackets represents the aging and decay rate of the device within a single step. Represents the electrical load loss weighting coefficient; Represents the device at any time The actual operating current sampling value; This represents the rated operating current parameter set on the equipment nameplate; The current load stress index is obtained through experimental calibration based on the material of the equipment hardware. Represents the weighting coefficient for thermodynamic load loss; Represents an exponential function with the natural constant as its base; The aging activation energy parameter representing the insulation material of the equipment; Represents the ideal gas constant. Represents the device at any time The actual operating absolute temperature sampling value; The reference absolute temperature parameter represents the design of the equipment.

[0119] When the cloud-based operations and maintenance platform receives new runtime load time-series data, it triggers incremental iterative calculations of the health decay index model. The platform will then accumulate the calculated health decay index. Numerical values ​​are truncated and scaled to map them to a percentage range of 0 to 100, generating a quantified assessment index for equipment aging. The platform allocates independent data table segments for each core power device in the central database and writes the timestamped health degradation index sequence to disk, achieving persistent storage of the equipment's aging trajectory throughout its entire lifecycle.

[0120] The platform constructs a set of device health threshold rules in memory. The system extracts a real-time updated health degradation index and compares it with the boundary threshold parameters set in the early warning rule set. The system divides the degradation threshold range into three segments: mild wear, moderate aging, and critical failure. When the calculated value of the health degradation index exceeds the lower limit of the critical failure segment, the platform outputs a judgment that the device performance is in a high-risk state. This quantified health degradation index provides the underlying numerical decision-making basis for the system's transition from passive fault response to proactive preventative intervention.

[0121] The cloud-based operations and maintenance platform extrapolates and predicts future degradation trends based on the historical time series of device health degradation indices stored in the database. The platform extracts health degradation index data points from the target device over a past period to construct a time series analysis sample. The system uses an autoregressive integral moving average (ARM) model to fit and predict the evolution of this time series. For the time series difference stabilization processing and the underlying algorithm logic for determining the order of autoregressive moving average parameters in the ARM model, those skilled in the art can refer to existing time series forecasting data analysis specifications for code writing. The model parameter estimation and white noise verification processes are well-known techniques in the field and will not be elaborated upon here.

[0122] The system generates predicted health degradation index values ​​for the device at various future time points using a time series prediction model. The platform retrieves the device's critical failure threshold parameter from memory. The system calculates the time span required for the target device to reach this critical failure threshold, generating the remaining useful life variable. The formula for calculating the remaining useful life variable is:

[0123] ;

[0124] In the formula, The variable representing the remaining useful life of the target device, calculated from the current moment. The variable representing the time span of future projections is measured in days or hours. Represents the current physical moment in which the system is performing lifetime prediction calculations; The device representing the output of the time series prediction model at future moments The predicted value of the health decline index; This represents the critical failure threshold parameter set by the system based on the equipment's factory hardware specifications and historical fault data.

[0125] The cloud-based operations and maintenance platform compares the calculated remaining service life variable with the system's configured early warning threshold. The early warning threshold is pre-set based on the underlying spare parts storage and scheduling cycle and the personnel shift planning cycle. When the value of the remaining service life variable is less than or equal to the early warning threshold, the platform determines that the core power equipment is about to enter a period of high incidence of physical failure and triggers the automated generation program of predictive pre-maintenance work orders.

[0126] The platform extracts the unique equipment identifier, equipment type classification code, latitude and longitude coordinates of the vessel currently belonging to the target equipment that triggered the warning, and the predicted critical failure timestamp. Based on the equipment type classification code, the platform queries the corresponding maintenance materials list in the central database. The maintenance materials list records the hardware consumables required to perform standard preventive maintenance on the equipment, covering special models of bearing assemblies, sealing ring accessories, or insulating coolant.

[0127] The system structurally integrates extracted basic equipment attributes, spatial geographic location data, and associated maintenance material list data, packaging them into a pre-maintenance work order data package. The cloud-based operations and maintenance platform attaches business skill classification tags to this pre-maintenance work order data package. The system then calls the underlying personnel attribute database and spatial matching algorithm to push the pre-maintenance work order data package to the mobile terminal application bound to the operations and maintenance personnel who have the corresponding equipment maintenance qualifications and who manage the water area.

[0128] The system marks pre-maintenance work orders triggered by remaining service life predictions as planned maintenance in the central database's work order record table, and sets a latest execution deadline for the work order based on the predicted critical failure timestamp. Upon receiving a pre-maintenance work order, the mobile application guides maintenance personnel to complete spare parts requisition and on-site equipment maintenance before the latest deadline. The pre-maintenance work order generation mechanism, based on health decay index extrapolation, transforms sudden hardware downtime into routine maintenance tasks with time redundancy, avoiding interruptions to vessel operations due to unplanned equipment damage.

[0129] After the cloud-based operations and maintenance platform sends out the dispatch instructions and structured work order data packets, the mobile terminal application creates a data image of the work order in its local storage device. The finite state machine engine built into the mobile terminal application reads the initial attributes of the work order data packet and initializes the lifecycle flag of the work order record to a pending response state. For the underlying system calls for record creation and memory state initialization in the mobile terminal's local database, those skilled in the art can refer to existing mobile operating system development specifications for implementation; the data persistence mechanism is a well-known technology in this field and will not be elaborated upon here.

[0130] Mobile terminal applications rely on a finite state machine engine to control the unidirectional, ordered transitions of work orders between different processing stages. The state transition calculation formula is:

[0131] ;

[0132] In the formula, This represents the work order status parameter for the next stage after triggering the transition calculation. The work order status parameter represents the current time point; This represents the vector of external trigger events currently received by the system, covering physical screen touch signals, GPS sensor data stream updates, and device QR code scanning and recognition results; The constraint matrix represents the current system's state transition, encompassing the results of non-empty verification of multimedia attachments and spatial coordinate distance comparison. This represents the state transition mapping function. The system reads the above input parameters based on the pre-configured logical routing table and performs the update output of the state flag bits.

[0133] Maintenance personnel trigger a work order reception command through the mobile application's interface. The finite state machine engine captures the touch confirmation signal from the external trigger event vector and verifies the data's validity constraints. The finite state machine engine transitions the work order's state parameter from the pending response state to the forward state. The application extracts the current local system clock to generate a response timestamp, calls the underlying network communication interface, and asynchronously uploads a data frame containing the device identifier, state change code, and response timestamp to the cloud-based maintenance platform, maintaining the synchronization consistency of state data between the edge and cloud.

[0134] The mobile application suspends a location monitoring service in the background, continuously acquiring the device's real-time latitude and longitude coordinates. The system calculates the Euclidean distance between the current real-time location and the coordinates of the target vessel or battery swapping cabinet set in the work order. The system compares the calculated distance value with the system-configured on-site check-in distance threshold. When the distance value is less than the threshold, a constraint condition matrix generates a satisfaction signal. The finite state machine engine reads this signal, determines that the personnel have physically arrived at the task site, automatically transitions the work order status parameter from "in progress" to "processing," and unlocks the specific business operation form on the interactive interface.

[0135] While the work order is in the processing state, maintenance personnel perform equipment repair or battery replacement actions. The mobile terminal application, by calling the underlying camera hardware driver, collects image data streams of the physical state of the equipment at the site and its post-repair state, and records the hardware serial number of the replacement consumables. The application constructs an execution data packet containing multimedia attachments and a text stream. When the maintenance personnel trigger the completion submission command, the application performs non-empty and format checks on the execution data packet. After the constraint matrix verification passes, the finite state machine engine changes the work order status parameters to the completed state. The underlying finite state machine logic, based on a strict precondition judgment mechanism, blocks invalid operations across stages, ensuring a strong binding between the on-site physical operation steps and the system data records.

[0136] When a work order is in the processing state, the mobile terminal application opens the on-site handling information collection interface. The application acquires the maintenance process record text entered by the maintenance personnel, the hardware serial numbers of consumed spare parts, and multimedia images of the repaired equipment. Simultaneously, the application calls the underlying operating system's location service and system clock to extract the latitude and longitude coordinates and precise timestamp of the current operation. For the underlying system implementation of the mobile terminal's hardware driver calls and multimedia file compression encoding, those skilled in the art can refer to existing mobile operating system development specifications for implementation. The data acquisition and encoding mechanisms are well-known technologies in this field and will not be elaborated upon here.

[0137] The mobile application serializes and concatenates collected heterogeneous data of various categories to construct a field handling feedback data packet. To ensure the immutability of the operation and maintenance execution data and the ability to conduct post-event audits and traceability, the application triggers a chain of evidence encryption and signature algorithm locally. The system extracts the core fields of the field handling feedback data packet and uses a cryptographic hash function to calculate a fixed-length digital digest. The formulas for calculating the digital digest and generating the digital signature are as follows:

[0138] ;

[0139] ;

[0140] In the formula, A unique digital digest parameter representing the data packet corresponding to the on-site handling data; This represents a secure hash algorithm mapping function, used to convert input data of arbitrary length into a fixed-length hash value; This represents the multimedia video data stream after the equipment has been repaired. This represents the character set of the maintenance process record text and spare parts serial number; Represents the precise timestamp at which the recording operation occurred; This represents the set of real-time latitude and longitude coordinates at the time the operation occurred. Represents the data string concatenation operator; Represents the generated digital signature variable; Represents an asymmetric encryption algorithm function; This parameter represents the private key allocated to the current operations and maintenance personnel account in the local secure storage area of ​​the mobile terminal.

[0141] The mobile terminal application appends the generated digital signature variables to the end of the on-site handling feedback data packet, forming a complete evidence chain encapsulation. The application establishes a secure transmission tunnel via a wireless network and uploads the evidence chain encapsulation to the cloud-based operation and maintenance platform. For key negotiation and secure tunnel establishment at the network transport layer, those skilled in the art can refer to existing transport layer security protocol standards for network deployment. The underlying communication encryption process is a well-known technology in the field and will not be elaborated here.

[0142] After receiving the evidence chain encapsulation, the cloud-based operations and maintenance platform retrieves the public key parameter bound to the corresponding operations and maintenance personnel's account. The platform uses this public key parameter to decrypt the digital signature variables, restoring the original digital digest parameter. The platform performs the same cryptographic hash calculation on the received multimedia images, recorded text, timestamps, and coordinate data sets to generate a local verification hash value for comparison. The platform performs a low-level byte-level equality comparison between the restored digital digest parameter and the local verification hash value. When the two values ​​are completely identical, the platform determines that the received on-site handling data has not been illegally tampered with or corrupted throughout its entire generation, storage, and transmission cycle.

[0143] After successful encryption verification, the cloud-based operations and maintenance platform writes the complete evidence chain into the read-only audit database. The platform assigns a unique archive index code and establishes a foreign key relationship between this code and the corresponding structured work order primary key record. The system updates the lifecycle status flag of the work order from "processing" to "closed-loop archived." Through a data interaction mechanism combining front-end hardware acquisition and back-end cryptographic verification, the platform transforms the dispersed physical intervention actions of operations and maintenance personnel into quantifiable and traceable electronic records.

[0144] In the above embodiments, the system underlying layer uses... Figure 1 and Figure 2The logic shown completes the calculation and status analysis of the vessel data. To intuitively convey these analysis results to frontline maintenance personnel and build a complete maintenance loop, this embodiment also provides a mobile terminal software interface, which is described below in conjunction with... Figures 3 to 7 Provide a detailed explanation of the operation process on the mobile terminal side.

[0145] like Figure 3 The image shows the login interface for this mobile terminal. Maintenance personnel must authenticate their identity by entering their assigned employee ID (e.g., OPS-8821) and password. This step ensures secure control over terminal operation permissions. After successful login, the system will retrieve data for the area currently managed by the personnel based on their employee ID.

[0146] After logging in, the system will redirect you to... Figure 4 The workbench homepage is shown below. The top of the page displays the current location of maintenance personnel (e.g., XX Lake, Central Lake Area) and their on-duty status; the middle section presents statistical results for "Today's Tasks, To-Dos, and Alerts" in a data dashboard format, all of which are distributed in real-time from the cloud; the bottom section includes quick access points and an emergency alert broadcast area. Furthermore, in the event of an emergency, users can click the "One-Click Battery Swap" shortcut button on the workbench, and the system will immediately display a pop-up message... Figure 6 The matching prompt layer shown is "Matching the nearest battery swapping station". At this point, the terminal sends a location request to the server, and the algorithm automatically sends the optimal battery swapping station location information based on the current coordinates, improving response efficiency.

[0147] In the routine operation and maintenance process, users can access the system via the bottom navigation bar. Figure 5 The task center interface shown is presented. This interface displays all events requiring intervention in a list format (All, To-Do, Warnings). For example, the "Boat #1024 Battery <15% System Warning" displayed in the list has an underlying triggering mechanism derived from the aforementioned (…). Figure 1 , Figure 2 The algorithm detects the results. Tasks in different states are distinguished by different labels (low battery, normal, fault) and operation buttons (process, to be done, under maintenance).

[0148] When the user Figure 5 After clicking "Go to Process" in the middle, you will enter the following... Figure 7 The task processing interface is shown below. The top of the interface displays the task number and a detailed description (e.g., guiding the user to a non-operational area to replace the battery). After processing, the maintenance personnel need to select the processing method below (e.g., on-site battery replacement), and then use the device's camera to take and upload photos or videos of the scene. Finally, they need to fill in a note to submit the work order. At this point, the terminal synchronizes the data transmitted from the scene to the cloud, thus completing the closed-loop processing of a single vessel anomaly.

[0149] The following describes in detail a complete embodiment of the collaborative operation of the various modules of the system of the present invention, using a typical water park operation scenario as an example, and provides specific comparative experimental data to verify its technical effect.

[0150] A water park is presumably equipped with 50 all-electric cruise boats, and a digital waterway electronic fence has been established, including a "main channel" and a "dense aquatic plant area." A cloud-based operations and maintenance platform has recorded the skill and location information of multiple operators. Operator A possesses skills in "electrical maintenance" and "mechanical cleaning" and is currently responsible for the central lake area; Operator B possesses skills in "battery replacement" and is currently responsible for the nearshore area.

[0151] The shipborne data acquisition unit collects operational status data of vessel #1024 in real time, with a sampling period of 1 second. At that moment, the communication transmission unit uploaded the data to the cloud-based operation and maintenance platform. After cloud-based resampling and filtering, the rule engine detected that the vessel's remaining battery power had dropped to 15% (below the preset threshold of 20%). Simultaneously, the cumulative algorithm detected a negative abrupt change in the ground speed parameter (sudden speed drop), and within the same time window, a positive abrupt change in the motor stator current parameter (sudden current increase). Based on this, the system extracted the abnormal physical feature of "high load shutdown of the power system".

[0152] The cloud-based operations and maintenance platform extracts the latitude and longitude coordinates within the aforementioned time window and uses a density-based spatial clustering algorithm to calculate the geometric center of the entanglement point. After performing ray cross-operation with the water area electronic fence model, it is determined that the spatial coordinates fall within the polygonal area of ​​"dense aquatic plant zone". The system concatenates the above physical features with environmental attribute labels into a multi-dimensional feature vector and inputs it into a multi-class fusion diagnostic model. The model outputs a probability of 92.5% for "propeller entanglement with aquatic plants," exceeding the 85% diagnostic confidence threshold. The system diagnoses the fault and automatically retrieves the "mechanical cleaning" and "battery replacement" business skill classification labels.

[0153] like Figure 8 As shown, the solid and dashed lines define the electronic fence boundaries of the regular main channel and the area with dense aquatic plants, respectively. The square element representing the abnormal vessel #1024 falls precisely within the area with dense aquatic plants, and a dashed warning circle is generated around it to indicate that it is in a high-risk stuck state; the remaining hollow dots represent cruise ships sailing normally.

[0154] The system searches the attribute database for matching personnel. Operator A, marked with a triangle, is 500 meters away from the abnormal vessel and currently has a workload of 1; operator B, marked with an inverted triangle, is 800 meters away and currently has a workload of 3. The system automatically generates [the appropriate personnel] through load balancing and spatial matching mechanisms. Figure 8The system dispatches a task along the arrow trajectory pointing to the abnormal vessel #1024, and accurately sends a structured chemical task data package containing fault coordinates and diagnostic conclusions to operator A's mobile terminal application.

[0155] After operator A's mobile terminal application receives the task, the built-in finite state machine engine changes the task status to "On the way". When operator A arrives at the site, the system changes the status to "Processing" based on GPS coordinate verification. After operator A performs the weed removal and battery replacement operations, he takes a picture of the restored propeller through the terminal. The mobile terminal extracts the image data, maintenance record text, timestamp, and coordinates, calculates a digital digest using a cryptographic hash function, and generates a digital signature variable using the private key, which is then uploaded. After the cloud public key decrypts and compares the data, ensuring the chain of evidence is tamper-proof, the task is finally marked as "closed-loop archived".

[0156] To verify the actual technical effectiveness of the present invention, a comparative experiment was conducted at the aforementioned water park for six months. The experimental subjects were divided into two groups: a control group using a traditional single-threshold alarm operation and maintenance system, and an experimental group using the intelligent operation and maintenance system of the present invention.

[0157] The system statistically analyzed the total number of various early warnings triggered by the underlying sensors during the experiment and their actual manual verification results. The specific data is shown in Table 1.

[0158] Table 1. Comparison of Fault Diagnosis Accuracy between Traditional Operation and Maintenance Systems and the Intelligent System of This Invention

[0159] Evaluation indicators Traditional operation and maintenance system (control group) The intelligent system of this invention (experimental group) Total number of alarms triggered 1420 times 856 times Actual number of failures 812 times 812 times Fault diagnosis accuracy 57.18% 94.86% Single sensor false alarm rate 42.82% 5.14%

[0160] Traditional systems rely solely on a single physical parameter threshold, making them susceptible to environmental factors such as sudden changes in water wave resistance, resulting in numerous false alarms. As shown in Table 1, the embodiments of this invention, by introducing a multi-dimensional feature vector and environmental attribute fusion mechanism, effectively filter out interference signals, significantly reducing the false alarm rate of a single sensor from 42.82% to 5.14%, and improving the fault diagnosis accuracy to 94.86%, thereby significantly improving the alarm reliability of the system.

[0161] The system extracts system flow log data from the generation of underlying alarms to the confirmation of completion by on-site maintenance personnel. Specific performance indicators are shown in Table 2.

[0162] Table 2. Comparison of Resource Scheduling and Execution Efficiency between Traditional Operation and Maintenance Systems and the Intelligent System of This Invention

[0163] Evaluation indicators Traditional operation and maintenance system (control group) The intelligent system of this invention (experimental group) Average task response time (from order acceptance to arrival) 28.5 minutes 14.2 minutes Battery swapping node matching failure exceeds range rate 16.4% 1.2% Equipment preventive maintenance interception rate 0% (Passive Repair) 82.5% Voucher data transmission tampering and loss rate 3.5% 0.00%

[0164] As shown in Table 2, the embodiments of the present invention optimize the work order flow logic by integrating priority scoring and load balancing allocation mechanisms, thereby reducing the average task response time by more than 50%. At the same time, by extrapolating and predicting the equipment health degradation index, a preventive maintenance interception rate of up to 82.5% is achieved, transforming traditional passive maintenance into proactive pre-maintenance. In addition, the cryptographic hash verification mechanism completely eliminates the risk of tampering and loss of credential data during transmission.

[0165] The above description is merely some specific implementations of this application and is not intended to limit the scope of protection of this application. Any variations or substitutions 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. An intelligent operation and maintenance system for smart vessels, characterized in that, include: The shipborne data acquisition unit is deployed on the vessel to collect operational status data, including positioning coordinates, battery level, and communication status. A communication transmission unit is used to send the operating status data to a cloud-based operation and maintenance platform; The mobile terminal application, configured on the operator's end, is used to obtain real-time location and number of pending tasks, receive operation input instructions, and collect on-site image data as evidence files to be sent to the cloud operation and maintenance platform. The cloud-based operation and maintenance platform is used to receive the operating status data, perform anomaly identification based on the battery level and the communication status, generate an anomaly event and extract the anomaly duration when an anomaly is triggered, and generate a comprehensive priority score by weighting the anomaly duration, battery level and positioning coordinates. The abnormal event and the comprehensive priority score are encapsulated into a system task; Calculate the spatial distance between the positioning coordinates and the real-time location, combine it with the number of pending tasks to select the optimal operator, and distribute the system tasks to the corresponding mobile terminal applications.

2. The intelligent operation and maintenance system for smart vessels according to claim 1, characterized in that, The cloud-based operation and maintenance platform performs timestamp alignment processing on the received operation status data, uses a resampling algorithm to map various underlying data to fixed time intervals, and applies a sliding median filter or Kalman filter to the timestamp-aligned operation status data for noise removal and cleaning, identifies and removes abrupt outliers in the data sequence, and uses a linear interpolation algorithm to fill in missing values.

3. The intelligent operation and maintenance system for smart vessels according to claim 1, characterized in that, The operating status data also includes motor operating status parameters; The positioning coordinates include a set of latitude and longitude coordinates and ground speed parameters; The cloud-based operation and maintenance platform extracts the operating status data to generate a device operating feature vector and inputs it into the rule engine. The rule engine performs anomaly identification based on the set multi-dimensional logical conditions and generates anomaly events when an anomaly is triggered.

4. The intelligent operation and maintenance system for smart vessels according to claim 3, characterized in that, The cloud-based operations and maintenance platform is used to generate the comprehensive priority score for: The penalty score for the abnormal duration corresponding to the triggered abnormal event, the risk score for the battery power conversion, the risk weight value based on the latitude and longitude coordinate set mapping, and the fixed importance coefficient of the boat equipment are extracted, multiplied by the corresponding sub-dimension weight factor parameters, and then the values ​​are summed to obtain the comprehensive priority score. The cloud-based operation and maintenance platform compares and matches the corresponding mapping table of the comprehensive priority score and the warning level, and assigns a clear sorting identifier to the system task.

5. The intelligent operation and maintenance system for smart vessels according to claim 4, characterized in that, The cloud-based operation and maintenance platform extracts the event type of the abnormal event, generates a corresponding abnormal state classification code based on the event type, and encapsulates the abnormal state classification code and the explicit sorting identifier into the system task. The cloud-based operation and maintenance platform reads the abnormal status classification code and the latitude and longitude coordinate set in the operation status data, and filters out the operators with matching jurisdictional water area and corresponding business skill classification tags as candidate objects from the internal personnel attribute database. When there are multiple qualified candidates, the cloud-based operation and maintenance platform will prioritize assigning the system task to the candidate with the fewest pending tasks. When the number of pending tasks is the same, the system task will be assigned to the mobile terminal application bound to the candidate that is closest to the target vessel in a straight line according to Euclidean distance.

6. The intelligent operation and maintenance system for a smart vessel according to claim 3, characterized in that, When the low battery abnormal event is triggered, the cloud-based operation and maintenance platform calculates the maximum theoretical sailing distance based on the battery power and uses it as the search radius. It then performs a range query in the internal spatial geographic database to extract all battery swapping cabinet devices that fall within the search radius, generating a candidate battery swapping cabinet set. Finally, it calculates the comprehensive matching score between the target vessel and each battery swapping cabinet device in the candidate battery swapping cabinet set. The cloud-based operation and maintenance platform extracts the battery swapping cabinet with the highest comprehensive matching score as the optimal battery swapping cabinet node, performs a resource locking operation on the optimal battery swapping cabinet node, and issues a scheduling instruction data packet.

7. The intelligent operation and maintenance system for smart vessels according to claim 3, characterized in that, The cloud-based operation and maintenance platform extracts the stator current parameter from the ground speed parameter and the motor operating status parameter to construct a time-series data stream, and uses a cumulative sum algorithm to perform abrupt change point detection and extract abnormal physical features; Extract the set of latitude and longitude coordinates corresponding to the time interval that triggered the mutation point to construct an abnormal retention point dataset. Calculate the spatial geometric center coordinates of the retention point clusters using a density-based spatial clustering algorithm. Perform a spatial intersection operation between the spatial geometric center coordinates and the preset water area electronic fence range to extract water area environmental attribute labels. Calculate the change in heading angle between adjacent sets of latitude and longitude coordinates; The abnormal physical features, the change in heading angle, and the water environment attribute labels are one-hot encoded and concatenated to generate a multi-dimensional feature vector. This vector is then input into a pre-trained multi-class fusion diagnostic model to calculate the probability distribution of the fault classification. The maximum probability value is obtained and compared with the diagnostic confidence threshold. The confirmed fault type is then output and updated in the system task.

8. The intelligent operation and maintenance system for smart vessels according to claim 1, characterized in that, The cloud-based operation and maintenance platform extracts historical operating data of the core power equipment of the vessel to construct a time-series data stream of equipment operating load. Based on a single sampling period, it accumulates the electrical load stress and thermal load stress in the time dimension to generate a quantitative equipment health degradation index sequence. A time series forecasting model is used to extrapolate and predict the health degradation index at various future time points. The remaining service life variable required to reach the critical failure threshold is calculated. It is determined whether the remaining service life variable is less than or equal to a pre-configured early warning threshold. If it is greater than the early warning threshold, continuous monitoring is performed. If it is less than or equal to the early warning threshold, the corresponding maintenance material list is queried in the central database. The maintenance material list is integrated to generate a predictive pre-maintenance work order data package, which is then sent as a system task.

9. The intelligent operation and maintenance system for smart vessels according to claim 1, characterized in that, The mobile terminal application has a built-in finite state machine engine. The finite state machine engine reads the initial attributes of the system task and controls the unidirectional ordered transitions of the system task between the pending response state, the proceeding state, the processing state, and the completed state according to the received operation input instructions.

10. The intelligent operation and maintenance system for a smart vessel according to claim 9, characterized in that, The finite state machine engine reads the constraint matrix containing the non-empty verification of multimedia attachments and the spatial coordinate distance comparison results, and triggers the unidirectional ordered transition in combination with the received operation input command; The finite state machine engine updates the state flag bit based on the transition result and asynchronously uploads the data frame containing the state change code to the cloud operation and maintenance platform.

Images