Intelligent shipping management system

Abstract

The invention relates to the technical field of shipping management information, and discloses an intelligent shipping management system which comprises an event perception and conversion module, an event bus and persistence module, a dynamic graph construction and analysis module, an intelligent process self-adaption engine module, a state projection and query service module and a unified application and interaction module. The dynamic graph construction and analysis module constructs and maintains an event causal association graph in real time. And the intelligent process self-adaptive engine module calculates a quantified cumulative influence index, and when the cumulative influence index exceeds a preset action threshold value, the engine generates a command to actively intervene the business process and sends the command to the event sensing and conversion module to form a control closed loop. The problems that a traditional system is lagged in response and lacks risk prediction capacity are solved, and prospective analysis and dynamic self-adaptive control over shipping business risks are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of shipping management information technology, in particular to a smart shipping management system. BACKGROUND

[0002] Modern shipping management relies on information systems to handle complex logistics, scheduling and monitoring tasks. However, existing shipping management systems still have some shortcomings when dealing with high dynamic and complex business scenarios.

[0003] Many existing systems focus on the recording and querying of business status in terms of function, and their running logic is often passive. When port congestion, equipment failure or adverse weather and other emergencies occur during shipping, these systems can usually only record the occurrence of the event, but lack the ability to make forward-looking predictions of the subsequent chain reactions triggered by the event. This results in management decisions being mostly reactive, making it difficult to intervene early before the risk spreads, thereby increasing operating costs or reducing service quality.

[0004] In addition, although existing systems have accumulated a large amount of operational data, these data are often treated as isolated information points. The deep and implicit causal relationships between events in the business process are difficult to effectively mine and utilize. When making assisted decisions, the system relies more on human experience or pre-set static business rules, and when faced with complex and rapidly changing shipping situations, this approach is difficult to provide scientific and dynamic decision support.

[0005] At the system architecture level, some traditional systems use a tightly coupled design, which results in a rigid system that is difficult to extend and has high maintenance costs. At the same time, their data storage method is usually a continuous overwrite of the current state, which makes it difficult to accurately trace the history of a business process (such as the complete life cycle of a container) and is not conducive to in-depth business audit and process optimization analysis. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a smart shipping management system, which aims to solve the problems of passive system response, lack of forward-looking risk prediction ability, and reliance on static rules for decision making, and difficulty in mining deep causal relationships between data in the prior art.

[0007] To achieve the above purpose, the present application realizes the following technical scheme: a smart shipping management system, which aims to realize forward-looking analysis and dynamic self-adaptive adjustment of shipping business processes by building an event-driven, causal graph-based analysis mechanism.

[0008] The system includes an event perception and conversion module, which is responsible for receiving raw data from different information sources and converting the raw data into structured domain events according to a pre-set shipping process event ontology;

[0009] an event bus and persistence module, which is configured to receive the domain events, record them in chronological order into an immutable data store, and provide event subscription and distribution services to other modules within the system, and which constitutes the only source of truth for the system;

[0010] a dynamic graph construction and analysis module, which is configured to subscribe to the domain events, and construct and maintain an event causal graph in real time based on the event stream, which describes the causal relationships between events;

[0011] a state projection and query service module, which is configured to subscribe to the domain events, and generate state views for different business scenarios by aggregating and computing a series of events, which can be queried efficiently;

[0012] a unified application and interaction module, which is configured to present the state views provided by the state projection and query service module to users, and receive user operation instructions, and convert the instructions into commands;

[0013] an intelligent process adaptation engine module, which is configured to utilize the event causal graph generated by the dynamic graph construction and analysis module to perform forward-looking analysis and impact prediction.

[0014] the commands generated by the unified application and interaction module, and the commands generated by the intelligent process adaptation engine module, are both sent to the event perception and conversion module, which converts them into new domain events and publishes them to the event bus and persistence module, thereby forming a complete data and control loop.

[0015] Preferably, the dynamic graph construction and analysis module includes a graph updater and a causal relationship reasoner. The graph updater is configured to create vertices corresponding to the domain events in the event causal graph. The causal relationship reasoner is configured to identify and quantify the causal relationships between events, and create corresponding edges in the event causal graph, which are attached with conditional probability and impact weight attributes.

[0016] In a specific embodiment, the causal relationship reasoner calculates the attributes by performing statistical analysis on historical business process instances recorded in the immutable data store. Specifically, it calculates the conditional probability between two event types based on the number of times that a result event type occurs within a preset time window after a cause event type occurs in historical instances. Furthermore, the causal relationship reasoner calculates the impact weight in combination with the calculated conditional probability and a predefined business severity function value for the result event type.

[0017] Preferably, the intelligent process adaptation engine module comprises a context perceiver, an impact predictor, and a decision and command generator.

[0018] The context perceiver continuously monitors the event stream in the event bus and persistence module to identify specific trigger contexts that require triggering a forward-looking analysis, and outputs one or a set of initial events.

[0019] In one embodiment, the context perceiver monitors through a rule engine, and the trigger context can be defined as a preset single key event, or a complex event pattern (CEP) composed of multiple events combined according to specific logics (e.g. temporal, causal, aggregation).

[0020] The impact predictor receives the initial events delivered by the context perceiver, and queries the event causal graph based on the initial events to perform a forward-looking analysis, and calculates a quantified cumulative impact index.

[0021] In one embodiment, the impact predictor implements the analysis process through a graph traversal algorithm. The algorithm takes the graph vertex corresponding to the initial event as the starting node, and performs a forward search along the directed causal relationship edges in the graph, and the search is only performed along the edges with conditional probability attribute values greater than a preset probability threshold. The set of all event vertices reachable through such a restricted traversal is defined as a potential impact chain. Subsequently, the impact predictor calculates the cumulative impact index by weighted sum of the business severity of all events in the potential impact chain and the probability of path occurrence.

[0022] The decision and command generator receives the cumulative impact index, and evaluates according to a configurable decision rule set to generate the command for intervening in the business process.

[0023] In one embodiment, the decision and command generator compares the calculated cumulative impact index with one or more preset action thresholds. When the value of the cumulative impact index is greater than the action threshold, the unit generates an intervention type command, which aims to actively change the business process to avoid the predicted risk.

[0024] Preferably, the event perception and conversion module further comprises an interface adapter unit for providing connection capability with multiple external information sources, which can include API adapter, message queue adapter or database change data capture (CDC) adapter; and an event generator unit for constructing the domain event into a structured data structure according to the verified data, which structure includes event unique identifier, event type, timestamp, payload and metadata.

[0025] Preferably, the event bus and persistence module comprises: a distributed event bus for decoupling and asynchronous transmission of the domain events between modules within the system; and an event store subscribing to the distributed event bus and persistently recording all domain events in an append-only write mode to ensure the completeness and tamper-proof of the business history record.

[0026] Preferably, the state projection and query service module comprises: one or more projectors and a read model database. The projector is a consumer of the distributed event bus and subscribes to a specific set of event types of interest. The projector works based on a state transition function that defines how a new event changes the current state. The projector aggregates the received new events using the function to update or construct the state view and stores it in the read model database for querying by the unified application and interaction module.

[0027] The present application provides a smart shipping management system. It has the following advantages:

[0028] 1. The present application realizes the forward-looking prediction of business risks by setting up an intelligent process adaptive engine module and using event causal association graphs. The adaptive engine can calculate the cumulative impact index of the current initial event based on the graph and automatically generate intervention commands when the index exceeds the preset action threshold. This mechanism changes the system from traditional post-response to pre-avoidance, improving the risk control capability and process resilience of shipping management.

[0029] 2. The present application sets up a dynamic graph construction and analysis module, especially the built-in causal relationship reasoner, which can automatically learn from historical business process instances in immutable data storage. By calculating the conditional probability and impact weight between events, it explicitly turns implicit business causal relationships into event causal association graphs, which not only provides decision-making basis for the intelligent process adaptive engine, but also provides deep insights into complex business processes for management personnel, improving the scientific nature of decision-making.

[0030] 3. The present application adopts an event-driven architecture, separates the write side (command processing) and the read side (query) through the event bus and persistence module, and realizes deep decoupling and asynchronous communication between modules by making the state projection and query service module, the dynamic graph construction and analysis module, etc. as independent subscribers of the event bus, improving the overall throughput and fault tolerance of the system. At the same time, the append-only immutable data storage ensures the completeness and traceability of business data, making the system easy to expand and maintain. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1System framework structure diagram of the present application.

[0032] Among them, 100, event awareness and conversion module; 200, event bus and persistence module; 300, dynamic graph construction and analysis module; 400, intelligent process adaptive engine module; 500, state projection and query service module; 600, unified application and interaction module. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the specification of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0034] Please refer to Figure 1 The present application provides a smart shipping management system, which comprises: an event awareness and conversion module 100, an event bus and persistence module 200, a dynamic graph construction and analysis module 300, an intelligent process adaptive engine module 400, a state projection and query service module 500, and a unified application and interaction module 600.

[0035] The event awareness and conversion module 100 is used to receive raw data from different information sources, and convert the raw data into structured domain events according to a preset shipping process event ontology.

[0036] The event bus and persistence module 200 is used to receive domain events, record them in a time sequence into an immutable data storage, and provide event subscription and distribution services for other modules in the system, thereby constituting the only fact source of the system.

[0037] The dynamic graph construction and analysis module 300 is used to subscribe to domain events, and construct and maintain an event causal correlation graph for describing the causal relationship between events in real time based on the event stream.

[0038] The intelligent process adaptive engine module 400 is used to utilize the event causal correlation graph generated by the dynamic graph construction and analysis module 300 to perform prospective analysis and impact prediction, and generate a command for adjusting the business process when the preset business rules are met.

[0039] The state projection and query service module 500 is used to subscribe to domain events, aggregate and calculate a series of events to generate state views for different business scenarios, and provide data query services to the outside.

[0040] Unified application and interaction module 600, which functions to present the state view provided by state projection and query service module 500 to the user, and to receive the user's operation instructions, and convert the instructions into commands to drive the subsequent business process.

[0041] The modules of the system work together to form a complete data and control loop. A typical processing flow is as follows: Unified application and interaction module 600 receives user operations and generates commands. The commands drive event perception and conversion module 100 to generate one or more domain events. The domain events are recorded and distributed via the event bus and persistence module 200. Dynamic graph construction and analysis module 300 and state projection and query service module 500 subscribe to the events respectively to update the event causal correlation graph and the corresponding state view. At the same time, intelligent process adaptive engine module 400 can analyze based on the updated graph and generate new commands when the preset conditions are met, and the new commands drive the generation of events again, thus forming a control loop.

[0042] The following is a detailed description of each module of the system.

[0043] Event perception and conversion module 100, as the input interface of the system, converts raw data from different types of information sources with different structures into standardized domain events that follow the preset shipping process event ontology and are uniform within the system.

[0044] To achieve the above functions, event perception and conversion module 100 can specifically include an interface adapter unit, a data parsing and verification unit, and an event generator unit.

[0045] The interface adapter unit functions to provide connection capabilities with multiple external information sources. The interface adapter unit can include a set of configurable adapters for specific data sources. Specific lower-level implementations can include: an API adapter for handling synchronous requests initiated by unified application and interaction module 600 or other external systems through HTTP / HTTPS protocol; a message queue adapter for subscribing to specific topics in a message middleware to receive asynchronous messages published by other business systems; a data stream adapter for handling continuous data streams from devices such as Automatic Identification System (AIS) or Vessel Traffic Service (VTS); and a database change data capture (CDC) adapter for capturing data changes in traditional relational databases.

[0046] A data parsing and verification unit, which is configured to process raw data received by the interface adapter unit. The data parsing and verification unit converts the raw data (e.g. JSON text, XML document, NMEA message or binary data) into an internal unified data structure according to the type of the data source adapter and using a corresponding parsing method. After parsing, the data parsing and verification unit verifies the integrity, format and validity of the business logic of the data according to the preset verification rules.

[0047] An event generator unit, which is configured to construct a structured domain event according to the verified data. In the present application, a domain event is formally defined as a five-tuple:

[0048] ;

[0049] Among them:

[0050] represents the globally unique identifier of the event, which can be generated using a universally unique identifier (UUID).

[0051] represents the type of the event, which is constrained by the pre-defined shipping process event ontology (S-PEO).

[0052] represents the precise timestamp of the event occurrence, which is in the Coordinated Universal Time (UTC) and follows the ISO8601 format.

[0053] represents the payload of the event, which is a data structure containing specific business data related to the type of the event.

[0054] represents the metadata of the event, which contains the context information of the event, such as the event source system identifier, the associated business process instance ID, etc.

[0055] A specific embodiment is used to illustrate the working process of the event sensing and conversion module 100: when a user submits a new freight order through the unified application and interaction module 600, the API adapter receives an HTTP POST request containing order information. The data parsing and verification unit parses the JSON data in the request body and verifies whether the required fields (such as the sender, the receiver, the cargo information) exist and are correctly formatted. After verification, the event generator unit generates a new domain event :

[0056] The corresponding is a newly generated UUID. ​is "ShipmentOrder Created"; is the current server time; is the validated order information; contains the user ID that triggered this operation. Thereafter, the generated new domain event is output to the event bus and persistence module 200.

[0057] The event bus and persistence module 200, as the only source of truth for the system, reliably transports and immutably stores the domain events generated by the event perception and transformation module 100, and provides data distribution for other modules within the system that need to consume events.

[0058] To achieve the above functions, the event bus and persistence module 200 can specifically include a distributed event bus and an event store.

[0059] The distributed event bus can achieve decoupling and asynchronous transmission of domain events between modules within the system. The distributed event bus can be implemented using a distributed message queue system. To manage events by category, multiple topics can be set up in the event bus.

[0060] A specific implementation is to divide the topics according to business domain or aggregate root, for example, a "Voyage Events" topic can be set up to carry all events related to a voyage, or a "ContractEvents" topic can be set up to carry all events related to a contract.

[0061] To achieve high throughput and horizontal scaling, each topic can be divided into one or more partitions; when publishing events, event producers can use a business process instance ID (such as a voyage ID) as a message key to ensure that all events for the same voyage are routed to the same partition, thereby ensuring the event order within a single business process instance. To ensure the reliability of data transmission, multiple replicas of each partition can be configured and synchronized to different broker nodes. For specific implementation of the distributed message queue system, those skilled in the art can use technologies such as Apache Kafka, and the specific deployment and configuration thereof are known in the art, and will not be described here.

[0062] The event store persistently records all domain events in the system, subscribes to all topics in the distributed event bus, and writes the received events to a dedicated data store. The data store follows an append-only write mode, i.e., any event once written is not allowed to perform modification or deletion operations, thereby ensuring the integrity and non-tamperability of business history records.

[0063] To optimize the query performance on specific business process history events, the data model of event storage can employ specific partitioning and indexing strategies. One specific implementation is to use the business process instance ID (e.g. voyage ID) as the partition key of the data table, and use the timestamp or a monotonically increasing sequence number of the event as the clustering key or ordering index within the partition. In this way, all history events of a specific voyage are physically stored contiguously and ordered by occurrence time, so that the complete event sequence of the voyage from creation to now can be efficiently retrieved. The event storage can be built based on a distributed database system that supports this data model.

[0064] The dynamic graph construction and analysis module 300 continuously processes the event stream distributed by the event bus and persistence module 200, and constructs and maintains a graph structure that describes the inherent logic and causal relationships between events in the shipping domain.

[0065] The dynamic graph construction and analysis module 300 can specifically include a graph updater and a causal relationship reasoner.

[0066] The graph updater is functionally subscribed to domain events on the distributed event bus in real time. Whenever a new domain event is received, the graph updater creates a corresponding vertex in the internally maintained event causal graph.

[0067] The causal relationship reasoner is functionally responsible for identifying and quantifying the causal relationships between events and creating corresponding edges in the event causal graph. The causal relationship reasoner can employ a hybrid reasoning mechanism, which can include:

[0068] Based on a pre-set expert rule base, causal edges are established for event pairs with a deterministic sequence or constraint relationship (e.g. the "ship departure" event must follow the "cargo loading complete" event);

[0069] In addition, based on statistical analysis of the sequence of historical events in the event storage, the probabilistic influence relationship between events is calculated, and causal edges are established accordingly. The event causal graph constructed in this way has vertices as domain events and edges representing the triggering, influencing or constraining relationships between events, and can be accompanied by quantitative attributes such as conditional probability and influence weight.

[0070] The event causal graph (E-CAG) in the present application is a directed weighted graph data structure used to formally describe the causal relationships between domain events. The causal graph Can be represented by a two-tuple:

[0071] ;

[0072] Where:

[0073] a set of vertices of the graph, each vertex in the set corresponds to a unique domain event that has been recorded in the event store . Each vertex contains at least a unique identifier of its corresponding event for indexing and association.

[0074] a set of edges of the graph, each edge in the set is a directed edge, representing a cause-effect relationship from a cause event to an effect event . The cause-effect relationship can be formally defined by a five-tuple:

[0075] ;

[0076] , respectively the cause event and the effect event in the cause-effect relationship, corresponding to the start vertex and the target vertex in the graph.

[0077] represents the type of the relationship, which can be a pre-defined enumeration value, such as Triggers for deterministic trigger relationships, or Influences for probabilistic influence relationships.

[0078] represents the conditional probability, whose value indicates the probability of the effect event occurring under the condition that the cause event occurs.

[0079] represents the influence weight, whose value quantifies the importance or the strength of the influence of the cause-effect relationship on the business process.

[0080] The function of the graph updater is to maintain the vertex set of the above-mentioned event cause-effect graph in real time . The graph updater, as a consumer of the distributed event bus, continuously subscribes to the event stream.

[0081] When a new domain event is received, the graph updater creates a new vertex in the vertex set of the graph , and associates with . To facilitate subsequent queries and analysis, may store Key attributes, such as event identifiers Event Type and timestamp .

[0082] Vertex set It can be stored using a data structure that supports fast key-value retrieval, for example, using A hash table with keys is used to achieve efficient access to any event vertex in the graph. After completing vertex creation, the graph updater will update the newly received domain events. It is passed to the causal reasoner to perform subsequent edge creation or update processing.

[0083] The function of the causal reasoning inference unit is to receive data transmitted from the graph updater. And based on this event, in edge set Create or update the corresponding causal relationship edge in the middle. The implementation of a causal reasoning engine may include a hybrid reasoning mechanism, the underlying implementation of which may include rule-based reasoning methods and statistical reasoning methods.

[0084] Rule-based reasoning methods are used to handle causal relationships with deterministic logic or inherent order in business processes. A causal reasoner can be configured with an expert rule base, which stores a series of pre-defined business rules. These rules define necessary preconditions, post-triggers, or constraints between specific event types. When a new domain event... Upon arrival, the inferencer matches it against rules in the rule base. If the event meets the triggering condition of a certain rule (for example, a Cargo Loaded event necessarily requires a Vessel Arrived At Port event as a preceding event), the inferencer then checks the graph. Find the corresponding preceding event vertex in the array and create an edge with deterministic properties between the two vertices. .

[0085] For this type of edge generated by rules, its relation type It can be set as Triggers or Constraints, and its conditional probability It can be set to 1.0.

[0086] Statistical reasoning methods are used to identify non-deterministic probabilistic relationships between events. Causal reasoning engines analyze historical event data recorded in an event store to calculate the statistical strength of associations between different event types. When a new event... Upon arrival, the inferencer can trigger an analysis task to calculate the conditional probability of this event type compared to other relevant event types. .

[0087] In a specific lower-level implementation, conditional probability The calculation can be based on the same business process instance (in... Events that co-occur in the context of the identifier. Conditional probability. This means that in the event type Under what conditions does the event occur, the event type The probability of occurrence can be calculated using the conditional probability formula, as follows:

[0088] ;

[0089] in:

[0090] This represents the historical business process examples analyzed. In the set, event types Total number of occurrences.

[0091] Representative in the analyzed instance In the collection, event type After it occurs, within the preset time window Internal, event type The number of times it happens.

[0092] This is a configurable time window parameter used to define the temporal correlation between two events.

[0093] Influence weight This weight is used to quantify the importance of the causal relationship. The weight can be combined with the calculated conditional probability. and the result event The inherent business importance is calculated. A predefined business severity function can be introduced. For each event type Assign a scalar value to represent the potential impact of the event on business processes (such as cost and time). Impact weight This can be derived from the weighting effect formula, as follows:

[0094] ;

[0095] The causal reasoning engine calculates... and After obtaining the value, it can be found in the graph. Create an edge of type Influences between the corresponding event vertices. and store these calculated values as attributes of the edge.

[0096] The intelligent process adaptation engine module 400 serves as the decision unit of the system, which utilizes the event-causal graph generated by the dynamic graph construction and analysis module 300 to conduct forward-looking analysis on the shipping business process. This analysis is used to realize the prediction of potential events, and when the preset business rules are met, it automatically generates intervention commands to realize the dynamic adjustment of the business process, thereby distinguishing from the traditional passive response mode. The intelligent process adaptation engine module 400 can specifically include a context perceiver, an impact predictor, and a decision and command generator.

[0097] The context perceiver continuously monitors the event stream in the distributed event bus to identify specific contexts that need to trigger forward-looking analysis. The lower-level implementation of the context perceiver can include a configurable rule engine that defines the trigger conditions that need to be monitored, which can include:

[0098] Single key event: that is, a specific event type that is predefined and has high business importance. For example, the "Encountered Severe Weather" or "Vessel Engine Failure" event can be defined as a single key event.

[0099] Complex event pattern: that is, a composite context composed of multiple events according to a specific logic (such as time sequence, causality, aggregation). The context perceiver can integrate a complex event processing engine to allow the definition of a time window and a logic relationship across multiple events.

[0100] As a specific example, a complex event pattern can be defined as: in the context of the same business process instance (for example, the same voyage ID), within a preset time window (for example, 2 hours), the system first detects a "Port Congestion Alert" event, and then detects a "Vessel Speed Reduced" event.

[0101] When the context perceiver detects that any preset single key event or complex event pattern is met through its rule engine or CEP engine, it will capture the context (including the triggering event and related context data) and pass it to the impact predictor for subsequent analysis.

[0102] The impact predictor is responsible for receiving the triggering context passed by the context perceiver, which is represented by one or a group of initial events . The impact predictor conducts forward-looking analysis based on the initial event Query the event causal relationship graph maintained by the dynamic graph construction and analysis module 300. To perform forward-looking analysis.

[0103] This analysis process is implemented using a graph traversal algorithm. The algorithm uses... Corresponding graph vertices As the starting node, along the graph Directed causal relationship edge in Perform a forward search, and the traversal process can use either depth-first search or breadth-first search.

[0104] To constrain the scope of the analysis and exclude low-probability associations, this graph traversal algorithm is subject to a preset probability threshold. The limitation is that the algorithm only follows its conditional probability. Attribute value greater than edge Perform a traversal. All elements traversed through this restricted traversal... Reachable event vertices The set is defined as This set Together with the edges they are associated with, they form a potential influence chain.

[0105] After identifying potential impact chains, the core function of the impact predictor is to quantify the overall impact of the chain; this quantification result is defined as the cumulative impact index. This index is calculated by weighting and summing the business severity of all future events in the potential impact chain, with the weights being the path probability of the event occurring. The cumulative impact index... The calculation formula is as follows:

[0106] ;

[0107] in:

[0108] Represents the initial event The calculated cumulative impact index value.

[0109] Represents the causal relationship graph of events From Starting point, based solely on conditional probability All event vertices reachable by the edge A set of.

[0110] This represents a scalar value calculated using a predefined business severity function in the causal reasoning inference engine, used to represent an event. Once it occurs, it will have an inherent impact on the business.

[0111] Represents the initial event Reaching the target event The probability of a specific path occurring.

[0112] Path occurrence probability The calculation is performed by considering all causal relationships along the path. conditional probability Multiplying them together, we get:

[0113] ;

[0114] in:

[0115] Representative in the diagram Middle connection and A specific directed path, which consists of a series of edges constitute.

[0116] Represents an edge on the path The attached conditional probability value;

[0117] Indicates the path All edges conditional probability value Multiply.

[0118] There are multiple routes arrive In the implementation of the path, the probabilities of all paths can be summed to calculate... The total probability of occurrence. After calculation, the influence predictor will finalize the AII. Values ​​and the set of events that constitute a potential chain of influence The output is sent to the decision and command generator.

[0119] The decision and command generator functions to receive the cumulative impact index output by the impact predictor. and the set of events in the potential impact chain. The decision and command generator then evaluates this input according to a configurable set of decision rules to generate commands for intervening in business processes.

[0120] In a specific implementation, this decision rule set can be implemented as a decision matrix or a set of conditional rules. This rule set defines... The value and one or more preset action thresholds The comparison relationship between them. For example, a warning threshold can be defined. and a high-risk action threshold .

[0121] When the value of the cumulative impact index is greater than the low-risk action threshold but less than or equal to the high-risk action threshold , the decision and command generator can generate a notification-type command (e.g., a Create Warning Notification command) with detailed information of the potential impact chain in its payload. When the value of the cumulative impact index is greater than the high-risk action threshold

[0122] , the unit can generate an intervention-type command (e.g., a Suggest Route Deviation or Trigger Contingency Plan command) that aims to proactively alter the business process to circumvent the predicted risk. The decision rule set can also make decisions based on the specific event types contained in the event set

[0123] . For example, if the event set contains an event (e.g., a “Contract Breach Projected” event) with a high business severity level , the rule can be triggered to generate a specific response command (e.g., a Notify Legal Department command) even if the cumulative impact index does not reach the low-risk action threshold . The command generated by the decision and command generator is a data structure that contains explicit operational intent. This command is sent to the event perception and transformation module 100 as a new input source. The adapter of the event perception and transformation module 100 receives this command and transforms it into a new domain event (e.g., a Contingency Plan Activated event), which is then published to the event bus and persistence module 200, thus driving the adaptive adjustment of the business process and forming a complete control loop.

[0124] The state projection and query service module 500 assumes the “read” side responsibility in the system’s Command Query Responsibility Segregation (CQRS) architecture, with the core role of subscribing to the event stream provided by the event bus and persistence module 200 and transforming this abstract, time-ordered event log into one or more materialized, business query scenario-optimized state views.

[0125]

[0126] ​​​These state views, also known as "read models", are materialized representations of the current state of the system's business, with data structures designed for efficient querying, thus completely separating the system's write data model (i.e. the event log in the event store) from the system's read data model.

[0127] To achieve the above functionality, the state projection and query service module 500 can specifically include one or more projectors, a read model database, and a query interface.

[0128] A projector, whose function is to act as one or more consumers of the distributed event bus. The system can configure a dedicated projector for each type of state view. Each projector subscribes to a specific set of event types that it is interested in for building the state view. The working principle of a projector is based on a state transition function, which can be formally represented as:

[0129] ;

[0130] Where:

[0131] represents the state of the state view at time .

[0132] represents a newly received domain event related to this business process instance by the projector.

[0133] represents the updated state of the state view after processing event .

[0134] represents the state transition function, which encapsulates the specific business logic that defines how a new event changes the current state view.

[0135] Take a "Voyage Real-Time State View" projector as an example, which can subscribe to event types such as "Voyage Created", "Vessel Departed", "Vessel Position Updated", "Port Arrived", etc. When it receives a Vessel Position Updated event, its state transition function will find the state view corresponding to the voyage ID in , and update the "current position", "speed", "heading", etc. fields in the view to the data carried in the event 's payload . When it receives a Port Arrived event, the state transition function The "Status" field of the Voyage View is then updated to "Arrived".

[0136] Take the Projector of "Contract Fulfillment View" as an example, it can subscribe to the event types of "Shipment Order Created", "Cargo Loaded", "Payment Received", "Cargo Delivered", etc. Its state transition function is Then by aggregating these events, it calculates and

[0137] Updates The "Fulfillment Progress Percentage" or "Key Milestone Status" (e.g. whether payment has been made, whether cargo has been loaded) of the contract in the middle. To ensure data consistency, the projector can include idempotency handling logic when implemented, such as by tracking the event sequence number or timestamp that has been processed for each business process instance, to avoid duplicate processing of events.

[0138] The read model database, whose function is to provide persistent storage for the multiple state views generated by the projectors Different storage technologies can be selected for different state views depending on their characteristics and query requirements. For example, for "Voyage Real-Time Status View" which requires high-throughput write and fast key-value query, an in-memory database or key-value store can be selected. For state views that require geospatial queries (e.g. querying all ships within a certain sea area), a database that supports geospatial indexing can be selected. For "Cargo Manifest View" which requires complex text search or aggregation analysis, a document database or search engine can be selected.

[0139] The query interface, whose function is to provide a standardized, high-performance data query channel for the unified application and interaction module 600. This interface exposes the state views stored in the read model database. In a specific implementation, this interface can be designed as a set of RESTful APIs or a Graph QL endpoint. These APIs are designed as read-only operations, with query parameters matching the data structure of the state views to achieve efficient data retrieval.

[0140] The unified application and interaction module 600 serves as the terminal medium for users (including shipping dispatchers, captains, management personnel, etc.) to interact with the system, and its architecture design strictly follows the principle of Command Query Responsibility Segregation (CQRS). The unified application and interaction module 600 clearly divides user operations on the system into two distinct paths: one is the "write" operation that changes the system state, and the other is the "read" operation that obtains system data, and they are processed through different technical channels, thereby achieving decoupling of business logic at the front-end level.

[0141] To implement the above architecture, the unified application and interaction module 600 can specifically include a command building and distribution unit and a view rendering and query proxy unit.

[0142] The command building and distribution unit is responsible for handling all user operations that intend to change the system state. When a user performs an operation such as "submit a new shipping order", "modify a voyage", or "confirm an alarm" on the interface, the unit does not directly initiate a traditional CRUD (Create, Read, Update, Delete) request to the backend database, but encapsulates the user's operation intention as a command object (Command Object) with clear semantics.

[0143] The command object is structured as a data structure that includes: a command type identifier (Command Type) indicating the operation intention (e.g. Create Voyage Command); a command payload (Command Payload) containing all business data required to perform the operation; and command metadata (Command Metadata) containing the operation initiator ID, operation timestamp, and client context information. After construction, the command building and distribution unit sends the command object to the interface adapter unit of the event-aware and conversion module 100 through the HTTPS protocol. In this process, the front-end application is only responsible for sending the command and receiving the confirmation of the return, and does not wait for the final processing result of the backend business logic, embodying the asynchronous processing characteristics of the system.

[0144] The view rendering and query proxy unit is responsible for handling all requests for data display. When the interface needs to display a voyage list, real-time ship position, or statistical report, the view rendering and query proxy unit does not access the event-aware and conversion module 100 or the event store, but directly communicates with the state projection and query service module 500.

[0145] The view rendering and query proxy unit obtains the pre-computed state view (Read Model) optimized for a specific UI scenario by calling the query interface provided by the state projection and query service module 500. Since the data structure of the state view has been aligned with the display requirements of the front-end interface (such as data fields, nesting levels), the view rendering and query proxy unit does not need to perform complex client-side data processing or aggregation operations after receiving the data, and can directly drive the UI component to render. For specific HTTP request library implementation and UI component rendering technology, those skilled in the art can use existing front-end development frameworks (such as React, Vue, etc.) and communication libraries (such as Axios) to implement them, which are well-known technologies in the art and will not be described here.

[0146] Through the above mechanism, the unified application and interaction module 600 realizes a one-way data flow closed loop: the user's operation enters the system (as input) through the command construction and distribution unit, after the event processing flow of the backend, the read model is updated, and finally the latest state (as output) is fed back to the user by the view rendering and query agent unit from the state projection and query service module 500.

Claims

1. An intelligent shipping management system, characterized in that, include: The event perception and conversion module is used to receive raw data and convert the raw data into structured domain events based on the preset shipping process event ontology. An event bus and persistence module are used to receive the domain events, record the domain events to an immutable data storage, and provide event subscription and distribution services. The dynamic graph construction and analysis module is used to subscribe to the domain events and build and maintain an event causal relationship graph in real time based on the event stream; The intelligent process adaptive engine module is used to perform forward-looking analysis and impact prediction using the event causal relationship graph, and when the preset business rules are met, it generates a command to adjust the business process and sends the command to the event perception and transformation module. The state projection and query service module is used to subscribe to the domain events and generate a state view oriented towards business scenarios by aggregating and calculating the domain events; The unified application and interaction module is used to present the state view to the user, receive the user's operation instructions, convert the operation instructions into commands, and send the commands to the event perception and conversion module.

2. The intelligent shipping management system according to claim 1, characterized in that, The dynamic graph construction and analysis module includes: A graph updater is used to create vertices in the event causal relationship graph corresponding to the domain events; A causal reasoner is used to create edges representing causal relationships between events in the event causal relationship graph, wherein the edges of the causal relationships between events are accompanied by conditional probability and influence weight attributes.

3. The intelligent shipping management system according to claim 2, characterized in that, The causal reasoning engine is also used for: Based on the historical business process instances in the immutable data storage, the number of times the result event type occurs within a preset time window after the cause event type occurs is obtained through the conditional probability formula; And based on the conditional probability and the business severity function value of the result event type, the influence weight is obtained through the weighted influence formula.

4. The intelligent shipping management system according to claim 1, characterized in that, The intelligent process adaptive engine module includes: The context-aware unit is used to monitor the event flow in the event bus and persistence module, identify the triggering context that triggers prospective analysis, and output the initial event. An impact predictor is used to query the causal relationship graph of the event based on the initial event and solve the cumulative impact index using the cumulative impact index calculation formula. And a decision and command generator, used to generate the commands for adjusting business processes based on the cumulative impact index.

5. The intelligent shipping management system according to claim 4, characterized in that, The context awareness device identifies the triggering context by monitoring a single key event or a complex event pattern composed of multiple events combined according to a specific logic.

6. The intelligent shipping management system according to claim 4, characterized in that, The influence predictor is used for: Using the graph vertex corresponding to the initial event as the starting node, a graph traversal is performed on the event causal relationship graph to identify the set of all event vertices reachable from the starting node and whose edge conditional probability is greater than a preset probability threshold, as a potential influence chain; The cumulative impact index is calculated by weighting and summing the business severity and path occurrence probability of all events in the potential impact chain.

7. The intelligent shipping management system according to claim 4, characterized in that, The decision and command generator is used for: The cumulative impact index is compared with one or more preset action thresholds; And when the cumulative impact index is greater than the action threshold, the command is generated.

8. The intelligent shipping management system according to claim 1, characterized in that, The event sensing and conversion module includes: An interface adapter unit is used to provide connectivity to various external information sources; The event generator unit is used to construct the domain event into a data structure that includes a unique event identifier, event type, timestamp, payload, and metadata.

9. The intelligent shipping management system according to claim 1, characterized in that, The event bus and persistence module include: A distributed event bus is used to enable asynchronous transmission of the domain events between modules; An event store is used to subscribe to the distributed event bus and persistently record the domain events through an append-only write mode, ensuring the immutability of the immutable data storage.

10. The intelligent shipping management system according to claim 1, characterized in that, The state projection and query service module includes: One or more projectors are used to subscribe to a specific set of event types and aggregate new events based on a state transition function to update the current state, thereby constructing the state view; Read the model database to persistently store the state view.

Images