Event-driven CPSPS distributed operation management and control method and system

Abstract

The invention discloses an event-driven CPSPS distributed operation management and control method and system. The method comprises the following steps: establishing a process drawing model of a processing process; establishing an RFID event model, a sensor event model and a CPS node event model containing event triggering conditions; performing semantic annotation on the sensing stream data to obtain RDF stream data; performing continuous query and monitoring on the RDF stream data by adopting a sliding time window, performing mode matching on a continuously output query result and an event triggering condition by utilizing an improved RETE mode matching algorithm, and outputting an event instance; constructing an event processing mechanism according to the event instance; a distributed production interaction and autonomous cooperation method based on CPS node self-organization driving is adopted to complete process task distribution, and a global Gantt chart is formed; and executing the production task according to the global Gantt chart. The Gantt chart is generated in a multi-CPS node distributed production social mode, and event-driven CPS node distributed production social and autonomous cooperation are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of production process operation control, and in particular to an event-driven CPSPS distributed operation control method and system. Background Art

[0002] Industry 4.0 has significantly advanced the integration of manufacturing and the internet. While the concept of Industry 4.0 remains debatable and subject to refinement, a significant practical challenge arises: a degree of ambiguity in manufacturing supply. Quantities, timing, product types, and functionalities are uncertain, often relying solely on experience. This is clearly unscientific. To address this, the Cyber-Physical Service Production System (CPSPS) has been proposed. This system leverages cyber-physical systems to systematically, connect, and computationally drive intelligent services and manufacturing. This approach addresses consumer demand, starting with data from demand-side services, then digitizing and driving production, and ultimately, providing integrated online and offline after-sales services.

[0003] However, CPSPS technology is not mature at present, and there has been no corresponding research on how to achieve distributed generation of operation control. As a result, the existing operation control methods still have problems such as poor precision, low efficiency, insufficient dynamic response capabilities, and lack of foresight in decision-making. It is difficult to meet the needs of distributed production social and autonomous collaboration in the context of fast data generation and high real-time processing requirements in the current human-machine-object interconnected manufacturing environment. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide an event-driven CPSPS distributed operation control method and system, which generates event instances based on real-time streaming data and event trigger condition detection and drives CPSPS to perform distributed operation control, and can realize event-driven CPS node distributed production social and autonomous collaboration.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] In a first aspect, the present invention provides an event-driven CPSPS distributed operation control method, the key of which is: comprising the following steps:

[0007] Define the concepts related to production events of CPSPS operation and establish a process diagram model for the processing;

[0008] Establish RFID event models, sensor event models, and CPS node event models that include event triggering conditions, and determine the corresponding application scenarios for each event model;

[0009] Establish an event ontology model, perform semantic annotation on sensor stream data, and obtain RDF stream data;

[0010] A sliding time window is used to continuously query and monitor RDF stream data. The improved RETE pattern matching algorithm is used to match the continuously output query results with event trigger conditions and output event instances.

[0011] Based on the finite automaton theory, an event processing mechanism is constructed according to event instances;

[0012] Based on the distributed production interaction and autonomous collaboration method driven by CPS node self-organization, process task allocation is completed and a global Gantt chart is formed;

[0013] Execute production tasks according to the global Gantt chart.

[0014] In a second aspect, the present invention provides an event-driven CPSPS distributed operation control system for implementing the method described in the first aspect, characterized in that it includes:

[0015] The first model building module is used to build a process diagram model of the machining process based on the production event-related concepts of the defined CPSPS operation;

[0016] The second model building module is used to build RFID event models, sensor event models and CPS node event models containing event trigger conditions, and determine the corresponding application scenarios of each event model;

[0017] The semantic annotation module is used to establish an event ontology model and perform semantic annotation on sensor stream data to obtain RDF stream data;

[0018] The event instance generation module is used to continuously query and monitor RDF stream data using a sliding time window, and uses the improved RETE pattern matching algorithm to match the continuously output query results with event trigger conditions to output event instances;

[0019] An event processing mechanism building module is used to build an event processing mechanism based on event instances using finite automata theory;

[0020] The process allocation module is used to complete process task allocation and form a global Gantt chart using a distributed production interaction and autonomous collaboration method driven by CPS node self-organization;

[0021] The execution module is used to execute production tasks according to the global Gantt chart.

[0022] The significant effects of the present invention are: the present invention aims at the new idea of distributed production social and autonomous collaboration and edge computing in the context of fast data generation speed and high real-time processing requirements in the current human-machine-object interconnected manufacturing environment. First, a process diagram model of the processing process is proposed, and the production event-related concepts of CPSPS operation are defined; secondly, an RFID event model, a sensor event model and a CPS node event model containing event triggering conditions are established respectively, and the corresponding application scenarios are described; thirdly, in order to effectively control the real-time sensor stream data of CPSPS, an event ontology model is established, the sensor stream data is semantically annotated, and RDF stream data is obtained. A sliding time window is used Continuously query and monitor RDF stream data, pattern-match the continuously output query results with event trigger conditions, output event instances, and construct an event processing mechanism based on finite automaton theory. Finally, to achieve swarm intelligence for self-organization of production equipment, a distributed production interaction and collaboration method driven by CPS node self-organization is proposed. A processing capability and processing demand model for CPS nodes and intelligent workpieces is established. A matching strategy based on ontology and constraint reasoning is used to match demand and capability, resulting in a set of candidate processing equipment. Furthermore, a gray key analysis model is used to establish an optimal equipment evaluation model, determine the optimal processing equipment, and conduct production task negotiation and contract signing to complete process task allocation. Unlike the traditional method of pre-generating a production planning Gantt chart for production task execution, this invention generates a Gantt chart through distributed production social networking among multiple CPS nodes. The Gantt chart is dynamically updated over time and is the output of the distributed operation control of the SocialF-oriented CPSPS system. This achieves event-driven distributed production social networking and autonomous collaboration among CPS nodes. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the method of the present invention;

[0024] Figure 2 It is a schematic diagram of the processing process diagram model;

[0025] Figure 3 It is a relationship diagram between the related concepts of CPSPS event model;

[0026] Figure 4 This is a schematic diagram of the CPSPS discrete event control system framework;

[0027] Figure 5 This is a schematic diagram of the event model based on the SSN ontology;

[0028] Figure 6 It is an improved Rete network construction flow chart;

[0029] Figure 7It is a schematic diagram of the CPSPS discrete event control system based on finite state machine theory;

[0030] Figure 8 It is a diagram of the "event-state-instruction" mapping relationship;

[0031] Figure 9 It is a logical flow chart of distributed production social and autonomous collaboration;

[0032] Figure 10 It is a schematic diagram of the manufacturing resource capability-demand ontology model;

[0033] Figure 11 It is a demand-capability matching flow chart based on ontology and rule reasoning;

[0034] Figure 12 It is a principle block diagram of the system of the present invention. DETAILED DESCRIPTION

[0035] The specific implementation manner and working principle of the present invention will be further described in detail below with reference to the accompanying drawings.

[0036] First, the nouns in the embodiments of the present invention are defined as follows:

[0037] Definition 1. Manufacturing Process: refers to the sum of a series of continuous operations performed by the process operator on the workpiece at a specific workstation during the CPSPS operation and control process, thereby changing the workpiece's shape, size, position, state and other attributes. A manufacturing process can be formally described as:

[0038] P id (Typ,Nm,Op,Loc,St,Et,,Info,Sub) (1)

[0039] Where: P id Represents the ID of the process; Typ, Nm, Op, and Loc represent the type, name, operator, and location of the process respectively; St and Et represent the start time and end time of the process respectively; Info represents a detailed description of the process; Sub represents the set of sub-processes, Sub = {P1, P2, ..., Pn}. A process can be decomposed into n sub-processes, and each sub-process can be further divided until the granularity requirements of operation control are met.

[0040] Definition 2: Object Status: This refers to the condition of an object in physical or cyberspace during CPSPS operation and control. It is characterized by a set of physical quantities with attribute characteristics (temperature, power, acceleration, etc.), temporal characteristics (the moment of the state), or spatial characteristics (the location of the object). For example, the power of a machine tool at time t is 253W, and the status of a workpiece at time t is "in the buffer zone."

[0041] Definition 3. Observation: Observing a physical object or phenomenon using an observation device and reading the object's state. The observation device in this invention primarily includes RFID devices for monitoring process logistics and sensors for monitoring equipment conditions. Therefore, observation here specifically refers to a single reading of the RFID device or sensor, which can be formally described as:

[0042] Obsid(Typ, Dev_ID, Prop, Time, Loc)(2)

[0043] Where: Obsid represents the observation ID; Typ represents the type of observation, including RFID device data reading and sensor data reading, where sensor data reading can be further divided according to sensor type; Dev_ID represents the observation device ID; Prop represents the attribute characteristics / physical quantity set of interest in the observation, Prop = {prop1, prop2, ..., prop n}; Time and Loc represent the time (usually a timestamp) and location where the observation occurred, respectively.

[0044] Definition 4. Streaming Data: This refers to a collection of sequentially ordered (time-stamped) observations generated by the same observation device over multiple observations of the same physical object or phenomenon. By processing and analyzing streaming data and combining it with event triggering conditions, events of interest are generated.

[0045] Definition 5. Production Event: refers to a description of the target object and its state changes. In the CPSPS control process, one or more observation results can be evaluated according to the event triggering conditions and mapped into event instances. For the sake of simplicity, the production events that appear in this embodiment are referred to as events. An event is formally described as:

[0046] E id (Typ,Dev_ID,<s1,s2> ,Prop,Cond,Time,Loc)(3)

[0047] Where, E idIndicates the ID of the event; Type indicates the type of event. In the CPSPS operation and management process, event types include RFID events, sensor events, and CPS node events, that is, Typ∈{E_RFID,E_Sensor,E_CPSN}; Dev_ID indicates the device ID that generates the event, that is, the event source (Event of Source), corresponding to the event type. Event sources include RFID devices, sensors, and CPS nodes. Events generated by RFID devices and sensors rely on a single data source and are atomic events, while events generated by CPS nodes rely on multiple sensor data sources and are composite events of atomic events.<s1,s2> Indicates the change of object state, that is, the transition from state s1 to state s2. The occurrence of an event can be determined by the change of object state; Prop represents the attribute set of the event, Prop = {prop1, prop2, ..., prop n}; Cond, Time, and Loc represent the conditions, time, and location of the event respectively.

[0048] Based on the time of their occurrence, events can be categorized as instantaneous events and interval events. An instantaneous event occurs at a specific moment, i.e., Time = t0. An interval event occurs within a specific time interval, i.e., Time = [t1, t2]: the event begins at t1 and continues until it ends at t2.

[0049] Definition 6. Event Condition: This condition is used to determine whether the state transition of an observed object satisfies a certain condition. It is a function of a set of event attributes. When the attribute values in the information space meet certain conditions, it can be determined that the state of the object in the physical space has changed, and an event has occurred.

[0050] Definition 7, Event Detection: Process the observed stream data and match it with event trigger conditions. If the match is successful, an event instance is generated.

[0051] Definition 8, RFID Event: This refers to the RFID device (RFID antenna / reader) continuously scanning the detection space at a user-defined sensing frequency to sense whether a tagged object (such as a smart workpiece) enters the detection space of the RFID antenna to determine whether an event is triggered.

[0052] Definition 9. Sensor Event: A sensor continuously collects attribute characteristic data of a physical object at a user-defined sampling frequency, processes the collected stream data, and extracts relevant indicators to determine whether the predefined event conditions are met. If so, a sensor event occurs.

[0053] Definition 10. CPS node event (CPSN Event): refers to a composite event formed by synthesizing several sensor events according to certain rules, used to describe more complex object state changes.

[0054] Definition 11. RDF stream data (RDF stream): uniquely identified by a uniform resource locator, consisting of several RDF stream tuples that imply a time series relationship, where each stream tuple contains an RDF triple and a corresponding timestamp t.

[0055] Definition 12: Sharing Degree of Atomic Condition: refers to the ratio of the number of atomic conditions in an event trigger rule referenced by rules in the rule base to the total number of rules.

[0056] Definition 13. Sharing Degree of Pattern: Since a pattern is composed of several atomic conditions, its sharing degree is the sum of the sharing degrees of each atom.

[0057] Definition 14, Machining Capability (MC): defines the type of machining methods, machining characteristics, machining materials, machining dimensions, machining performance, etc. that can be used by a machining equipment resource to process parts, and describes the capacity, scope, and achievable performance quality of the machining equipment resource.

[0058] Definition 15: Machining Requirement (MR) describes the machining method type, machining characteristics, machining materials, machining dimensions, machining quality, and other requirements that the workpiece process should meet during machining.

[0059] Example 1:

[0060] like Figure 1 As shown, this embodiment proposes an event-driven CPSPS distributed operation control method, the specific steps are as follows:

[0061] Step 1: Define the concepts related to the production events of CPSPS operation and establish a process diagram model of the processing process;

[0062] The processing procedure is the core concept of production operation within SocialF production. From the perspective of production logic, the processing tasks put into the workshop are decomposed into a collection of a batch of processes, and the CPSPS under SocialF is composed of multiple CPS nodes with equipment as the core. Each node performs decentralized interaction and collaboration around the manufacturing task centered on the process, thereby completing the production of the task. From the perspective of data collection and association, the IDS for SocialF is constructed by collecting dynamic process-related data (process logistics data, equipment operating data, and process quality data) and associating it with static resource data. In particular, the processing procedure is the most important type of all manufacturing process types. This embodiment divides a single processing procedure into several stages according to key time points, such as Figure 2 shown.

[0063] CPSPS operation control can be divided into key time points (granularity) of the process according to actual needs. The state of the workpiece can be regarded as the state of the process at this granularity. However, during the operation of CPSPS, the state of objects such as equipment, workpieces, and processes will not always remain in a fixed state. Therefore, an event mechanism is required to model state changes and specify how CPS nodes respond to these changes.

[0064] Based on the physical meaning of each time node within the machining process, this embodiment deploys three RFID readers in the machine tool's entry buffer, machining area, and exit buffer, respectively. These readers collect time and status data for workpieces entering and leaving the relevant areas, enabling modeling and detection of RFID events, and thus enabling logistics monitoring within and between processes. Furthermore, during the machining process, the operating conditions of machine tools, cutting tools, and other components will change over time, significantly impacting workpiece quality and equipment production efficiency. Therefore, relevant sensors can be deployed to collect equipment operating condition data, enabling modeling and detection of sensor events, and thus enabling monitoring of the operating conditions of CPS nodes.

[0065] Step 2: Establish RFID event models, sensor event models, and CPS node event models that include event triggering conditions, and determine the corresponding application scenarios for each event model;

[0066] CPSPS operational control is implemented with the process at its core. During implementation, it collects process logistics data (RFID data) and equipment operating data (sensor data) from the processing process. This collected (streaming) data is processed and matched with event triggering conditions to generate events. The generated event instances are bound to different nodes in the process diagram model and generate corresponding production instructions. The CPS nodes are then responsible for parsing and executing the production instructions, controlling the production process of the process tasks. Through the closed-loop control process of "task execution → triggering production events → generating production instructions → controlling task execution," event-driven CPSPS distributed production social and autonomous operational control is achieved.

[0067] In the previous section, we defined the concepts related to the event model in the CPSPS operation and control process. The relationship between these concepts is as follows: Figure 3 As shown in Figure 1 . Physical objects monitored for events include RFID-tagged workpieces, machine tools, manipulators, transport carts, conveyor belts, and the like. Their states are characterized by relevant attribute characteristics (such as temperature, power, and position), and the values of these attribute characteristics are perceived by observation devices (RFID devices and sensors), representing a digital mapping of physical quantities in information space. RFID devices or sensors continuously monitor and generate stream data with timestamps. By processing this stream data, physical state changes that meet event conditions are extracted, generating RFID events or sensor events. The RFID events and sensor events referred to in this embodiment are both generated by a single observation device and are generated by evaluating stream data with a single attribute variable. These events are referred to as atomic events in this embodiment. Since an atomic event is simply a description of a physical object's state change at a specific moment by a single observation device, and a CPS node has multiple observation devices, the various types of stream data generated by these different observation devices will be evaluated to generate multiple atomic events. These atomic events are then "combined" according to certain rules to produce a composite event. The CPS node event in this embodiment is a composite event.

[0068] Therefore, the establishment process and application scenarios of the three event models in this example are as follows:

[0069] The process of establishing the RFID event model is as follows: collecting process logistics data to form process flow data with timestamps; processing the process flow data, extracting physical state changes that meet event conditions, and matching them with event trigger conditions to generate RFID events.

[0070] In the RFID-driven process flow, RFID events are used to monitor the flow within the processing process and the flow between processes. Four typical RFID application scenarios are:

[0071] (1) Scenario 1: A fixed RFID reader / antenna forms a fixed detection space pattern, and the RFID-tagged object "enters / leaves" the detection space;

[0072] (2) Scenario 2: The reader / antenna is fixed on a movable object (such as a transport vehicle) to form a mobile detection space mode. The vehicle-mounted RFID antenna / reader detects the RFID-tagged object "entering / leaving" the detection space;

[0073] (3) Scenario 3: A fixed RFID reader / antenna forms an access control mode, and the RFID-tagged object "passes" through the access control;

[0074] (4) Scenario 4: The handheld reader (including antenna) forms a random detection spatial pattern, and the operator uses the handheld terminal to randomly track the status of the RFID-tagged object.

[0075] By instantiating the above four RFID application scenarios in CPSPS operation and control, we can see that RFID events include entry events Ein, exit events Eout, and pass events Epass. Since the triggering principle of RFID events is that the RFID device senses whether the tagged object is within the detection space, the triggering conditions of RFID events based on stream data can be expressed as:

[0076] (0 <AVG(WP.flag)<1)∧WP.nextProcess=c∧WP.state=s (4)

[0077] Where: flag represents the flag bit of each data record in the RFID stream data of a labeled object (in this embodiment, the smart workpiece), flag∈{0,1}, “0” means that the smart workpiece is not sensed at the current moment, that is, E in Has not yet occurred or E out "1" means that the intelligent artifact has been sensed at the current moment, that is, E in Has occurred or E out Not happened yet; 0 <AVG Δt(flag) < 1 means that within the Δt time period, the average value of the flag bits of all data records in the RFID stream data is greater than 0 and less than 1. Its physical meaning is that within the Δt time period, the SW is detected by the RFID device at some times and not by the RFID device at other times, indicating that the smart workpiece has experienced two states, that is, the state has changed, indicating that an RFID event has occurred; WP.nextProcess = c indicates whether there is a next process for the current process of the smart workpiece, c∈{0,1}, "0" indicates that there is no next process, and "1" indicates that there is a next process; Object.state = s indicates the state of the smart workpiece in the RFID system, including nine states such as "in the buffer zone", "out of the buffer zone", and "in the processing area". The state transition relationship of SW is shown in the case study section below.

[0078] The process of establishing the sensor event model is as follows: the sensor continuously collects attribute feature data of the physical object at a user-defined sampling frequency to form sensor stream data with a timestamp; the collected sensor stream data is processed and relevant indicators are extracted to determine whether the predefined event trigger conditions are met. If so, a sensor event is generated.

[0079] Sensor events are used to monitor the working conditions of machine tools, cutting tools, and other equipment involved in the cutting process within the machining process, as well as the production environment. These include monitoring working conditions such as machine power, spindle speed, and tool cutting temperature, as well as monitoring production environment conditions such as noise, temperature, and humidity. The triggering conditions for sensor events can be expressed as:

[0080] C1OPC2OP...OPC n (5)

[0081]

[0082] Where OP represents the logical operators AND(∧), OR(∨) and C i The i-th sub-condition of the event trigger condition Cond, each sub-condition can be recursively defined and combined by each atomic expression in formula (6); α, β represent numerical constants; represents the comparison operator, represents an arithmetic operator, S Obs It represents the streaming data with time stamps formed by multiple observations of the same physical object / phenomenon by the same observation device; f(S Obs ) represents the convection data S Obs The processing operation function can be a simple aggregation function, such as the sum function SUM(S Obs ), average value function AVG(SObs ), find the maximum value function MAX(S Obs ), find the minimum function MIN(S Obs ) and the function COUNT(S Obs ), in addition, the operation function can also be other complex functions for stream data processing, such as signal processing functions; trend(S Obs ) indicates the judgment flow data S Obs The value of the trend (increase or decrease), trend (S Obs )∈{up,down}; Indicates the extraction of stream data S Obs duration and compare it with Δt.

[0083] The establishment process of the CPS node event model is as follows: synthesizing several sensor events into a composite event according to certain rules; judging whether the composite event meets the predefined event triggering conditions, and if so, generating a CPS node event.

[0084] The triggering conditions of CPS node events are the synthesis rules of atomic events, which can be described as:

[0085] E CPSN =f(E1,E2,...,E n )(7)

[0086]

[0087] Where: E i represents the i-th atomic event, E i ∈{E RFID ,E Sensor}; f represents the composition function, which is used to describe the composition rules of atomic events and can be recursively defined by formula (8); And Δt (E1, E2) means that events E1 and E2 both occur within the period Δt; Or Δt (E1, E2) means that either event E1 or E2 can occur within the Δt period; Not Δt (E) indicates that event E does not occur within the Δt period; Before Δt (E1, E2) means that event E1 occurs before E2 by time Δt; After Δt (E1, E2) indicates that event E1 occurs Δt later than E2.

[0088] Step 3: Establish an event ontology model, perform semantic annotation on the sensor stream data, and obtain RDF stream data;

[0089] This embodiment proposes a CPSPS discrete event control system framework for RDF stream data processing. Figure 4 ,The framework mainly includes four aspects: (1) preprocessing of raw sensor stream data; (2) continuous query of RDF stream data based on sliding time window; (3) event instance generation based on RETE pattern matching algorithm; (4) event processing based on automata theory.

[0090] Based on the above four aspects, the specific process of this step is as follows:

[0091] Step 301: Construct an event ontology model between production event-related concepts such as sensors, observations, events, event conditions, and event attributes;

[0092] In order to realize event detection and efficient reasoning based on sensor stream data, an ontology model between concepts such as sensor, observation, event, event condition, and event attribute is constructed, such as Figure 6 The event ontology model constructed in this embodiment is a modification and extension of the Semantic Sensor Network (SSN) ontology model, a W3C recommended standard. Based on SSN, it adds core classes such as Event, EventType, Production Instruction, CPS Node (CPSN), and Event Trigger Condition, and designs the object attribute association relationships between them.

[0093] In the event model proposed in this embodiment, events are generated by stream data formed by observations, and each observation is obtained through a sensor. The sensors in the SSN include the RFID devices and sensors proposed in this embodiment. Properties are used to characterize the physical quantities of interest to sensors, observations, and events. Results describe the circumstances of each observation, including the observed value, time, and unit. Events generate instructions, and each instruction is executed by a CPS node, which is also responsible for evaluating event triggering conditions. The event type indicates that each generated event instance must belong to an RFID event, a sensor event, or a CPS node event.

[0094] Step 302: After constructing the event-centered ontology model, determine the relevant sensors and the data they collect by using ontology semantic queries;

[0095] Step 303: Use the semantic annotation tool Jena to annotate the original stream data, thereby obtaining RDF stream data.

[0096] In practice, stream data is RFID data or sensor data that is continuously generated over time. Therefore, theoretically, stream data has temporal relationships and is infinite. Therefore, it is necessary to continuously query and monitor stream data to extract stream data that meets specific time or time periods. In light of the above analysis, this embodiment uses a sliding time window operator to implement continuous querying of RDF stream data. RDF stream data has the following form:

[0097]

[0098] More generally, an RDF stream is formally described as:

[0099] S IRI ={(<s,p,o> ,t)|<s,p,o> ∈((I∪B)×I×(I∪B∪L)),t∈T} (10)

[0100] Where: S IRI Represents an RDF stream data, associated with a unique identifier IRI, usually a stream data access address consisting of an IP address and a port number; (<s,p,o> ,t) represents an RDF triple with a timestamp, referred to as an RDF stream tuple, a triple<s,p,o> The definition of refers to Chapter 3, t is the timestamp of the triple; T represents the set of infinite timestamps. It should be noted that the timestamps in the stream tuples have a monotonically non-decreasing relationship, that is, t i ≤t i+1 , the meaning of the equal sign is that any number of stream tuples can have the same timestamp, indicating that these triplets "occur" at the same time.

[0101] Considering the infinite nature of RDF stream data in the time domain, a sliding time window operator is introduced to enable continuous query and monitoring of stream data. The purpose of the sliding time window on RDF stream data is to extract the latest stream tuples from the stream data at a specified frequency by matching the given graph pattern. Therefore, a time window is defined as:

[0102] W type ::=(Range,Slide,Frequency) (11)

[0103] Where: W typeThe time window type is divided into logical time windows (LW) and physical time windows (PW) according to the different ways of extracting stream tuples. Range is the length of the time window. If the time window type is LW, Range is a specified time interval. The units can be milliseconds (ms), seconds (s), minutes (m), hours (h), days (d), etc. Slide is the step size of the time window sliding forward. Frequency is the frequency of performing evaluation operations on the sliding time window, including aggregation operations on stream data (average, maximum, minimum, etc.) and constructing new RDF triples.

[0104] According to the above definition, a function acting on the stream data S IRI The LW operator on is:

[0105]

[0106] Where: l, δ, η represent the length, step size and evaluation frequency of the logical time window respectively; t s ,t e They represent the start and end time of the time window to extract the stream data, usually t e =Now means the end time is the current time, which is used to extract the latest stream data. The length of the time window l = |t e -t s |.

[0107] A function that acts on the stream data S IRI The PW operator on is:

[0108]

[0109] Where: l, δ, η represent the length, step size and evaluation frequency of PW respectively; N = Count(S IRI ,t s ,t e ) represents the latest N stream tuples extracted by PW on the stream data, where Count(S IRI ,t s ,t e ) represents the time interval (t s ,t e ] is the number of stream tuples on .

[0110] In some preferred embodiments, the C-SPARQL semantic stream data query engine is used to implement continuous querying and monitoring of RDF stream data using time window operators. This engine is an extension of the SPARQL 1.1 standard, supporting continuous querying of RDF stream data, as well as aggregation operations (SUM, AVG, MAX, MIN, and COUNT), grouping operations (GROUP), and filtering operations (FILTER) on stream data.

[0111] Step 4: Use a sliding time window to continuously query and monitor the RDF stream data, and use the improved RETE pattern matching algorithm to match the continuously output query results with the event trigger conditions to output event instances;

[0112] Event instances are obtained based on state changes, and event triggering conditions are the characterization of state changes. This embodiment uses an improved Rete algorithm to perform pattern matching on the continuous query results of RDF stream data with the event triggering rule library, thereby realizing event detection and generating event instances. The Rete algorithm is a forward chain rule matching algorithm for production systems proposed by Dr. Charles Forgy. However, in the construction of traditional Rete networks, the different orders of atomic conditions in the rules will lead to the compilation of Rete networks with different structures, and the complexity of these networks is also different. How to build an efficient matching Rete network with lower storage cost and faster matching speed is an urgent problem to be solved.

[0113] To address the above issues, a basic approach is proposed: since many rules may share some of the same patterns, the performance of the rule engine can be improved by compressing the number of nodes in the shared memory of the nodes in the Rete network; patterns with more general or stronger constraints are matched earlier as much as possible to increase their sharing level; and rules that are easily changed are matched later as much as possible to reduce the changes in the rule base caused by rule changes. To this end, this embodiment proposes an improved Rete algorithm that introduces the concept of node sharing and sorts nodes according to their sharing level when compiling event trigger rules, thereby improving the node sharing level in the Rete network and reducing the number of redundant nodes.

[0114] Assume that there are n rules in the rule base, that is, R = {r1, r2, ..., r n}, atomic condition c i Sharing Indicates that

[0115]

[0116] Through the above definition, the atomic condition sharing degree in the rule can be calculated and the atomic conditions can be sorted according to the sharing degree. The higher the atomic condition sharing degree, the higher the order in the rule that references the atomic condition. Since there are cases where the atomic conditions have the same sharing degree in a rule, the different order of the atomic conditions with the same sharing degree will also have a significant impact on the sharing performance of the nodes in the Rete network. Therefore, the concept of pattern sharing degree is introduced. Assume that the pattern P j Contains m atomic conditions, using represents the mode sharing degree, then

[0117]

[0118] In the case of the same atomic condition sharing degree, the atomic conditions can be further sorted using the pattern sharing degree. Assuming that the atomic condition c i Appears in k patterns (P1, P2, ..., P k ) in the Represents the atomic condition c i The sum of the shared degrees in k patterns is

[0119]

[0120] because Reflects the effects of other atomic conditions on the atomic condition c i Dependence The larger the value, the higher the degree of dependence and the higher the degree of sharing in the pattern. In the same case, according to Sort atomic conditions and build a highly shared Rete network.

[0121] Improved Rete network construction process Figure 6 As shown, it mainly includes four steps: (1) According to the proposed sharing degree model, calculate the pattern sharing degree of the atomic conditions in the event trigger rule set, and rearrange the order of the atomic conditions in the rule accordingly, and output the sorted rule set; (2) Generate an Alpha network based on the sorted rule set; (3) Classify the nodes in the Alpha network, sort the Alpha classifications in descending order according to the Alpha classification sharing degree and average sharing degree, and output the sorted Alpha network; (4) Based on the sorted Alpha network, combine Beta nodes and finally output the improved Rete network. When inserting the Beta node, the traditional Rete algorithm only considers the logical AND relationship between the two atomic conditions. In fact, in the event model proposed in this embodiment, the atomic conditions of the event trigger rule can be in a logical AND relationship. Therefore, when constructing the Beta node, this embodiment also considers the logical AND relationship so that it can adapt to the proposed event model.

[0122] After the Rete network is built, the sliding time window can be continuously queried and the pre-processed RDF triple facts can be input into the Rete network for pattern matching. The event trigger pattern that meets the event trigger pattern indicates that an event instance has been generated. The specific process of the improved Rete network matching is as follows:

[0123] Step 401: traverse the atomic condition set and the rule set, and calculate the sharing degree of each atomic condition;

[0124] Step 402: Calculate the sharing degree of each pattern; Calculate the pattern sharing degree of each atomic condition;

[0125] Step 403: sort the atomic conditions in the rule set in descending order according to their pattern sharing degrees; and output the sorted rule set;

[0126] Step 404: Create a root node;

[0127] Step 405: traverse the rule set and extract the rule;

[0128] Step 406: traverse the sub-condition set and extract the sub-condition;

[0129] Step 407: Check the parameter type of sub-condition j. If it is a new type, add a node of this type.

[0130] Step 408: If sub-condition j does not exist in the Alpha network, insert sub-condition j as an Alpha node into the Alpha network and create an Alpha memory table.

[0131] Step 409: Determine whether all sub-conditions in the current rule have been traversed. If so, proceed to step 409; otherwise, return to step 406.

[0132] Step 4010: Determine whether all rules in the rule set have been traversed. If so, output the Alpha network; otherwise, return to step 405.

[0133] Step 4011: traverse the Alpha network and calculate the sharing degree of the Alpha categories; sort the Alpha categories in descending order according to the sharing degree; sort the Alpha categories with the same sharing degree in descending order according to the average sharing degree; and output the sorted Alpha network.

[0134] Step 4012: traverse the rule set and extract the rule;

[0135] Step 4013: traverse the sub-condition set and extract the sub-condition;

[0136] Step 4014: Combine the Beta nodes: Beta(2)'s left input node is Alpha(1), and its right input node is Alpha(2); Beta(i)'s left input node is Beta(i-1), and its right input node is Alpha(i);

[0137] Step 4015: Inline the memory tables of the two parent nodes into one's own memory table;

[0138] Step 4016: Determine whether all sub-conditions in the current rule have been traversed. If so, proceed to step 4017; otherwise, return to step 4013.

[0139] Step 4017: Encapsulate the conclusion of the current rule into a terminal node as the output node of Beta(n);

[0140] Step 4018: Determine whether all rules in the rule set have been traversed. If so, output the Rete network; otherwise, return to step 4012.

[0141] Step 5: Based on the finite automaton theory, construct an event processing mechanism according to the event instance;

[0142] The state changes of physical objects generate event instances, which in turn generate production instructions to control the physical objects. This process can be represented by a finite state machine (FSM). The finite state machine can transition the physical object from one state to another based on event triggering conditions, represented by a five-tuple:

[0143] (S,Σ,θ,s0,S F )(18)

[0144] Where: S represents a finite non-empty state set, S={s1,s2,...,s m}; Σ represents the set of events that may occur in the system, Σ={E1,E2,...,E n}; θ represents the state transition function of the system, i.e., θ:S×Σ→S. In this embodiment, the state transition function is the event processing function, which is a complex event response process, and its output is the production instruction; s0 represents the initial state of the system, s0∈S; S F Represents the acceptance state set, which can be recorded in the system’s identification state set, S F ∈S.

[0145] The CPSPS discrete event control system based on finite state machine theory is a closed-loop control process with state feedback, such as Figure 7 As shown. When the system state is changed from s i-1 Transformed into s i When event Ei Trigger. At this time, the event processing function θ(E i ,s i ) is called to handle event E i Respond and process, generate a production instruction set and act on the system to create a new state.

[0146] Event processing function θ(E i ,s i ) takes event instances and the current state of the system as input and outputs a set of production instructions. Therefore, in a CPSPS control system constructed by a finite automaton, the essence of the event processing function is to construct a mapping relationship between “state-event-instruction”, such as Figure 8 As shown, first, according to the production monitoring requirements and scenarios, the state set S, event set Σ and production instruction set I={I1,I2,...,I k}, then the instruction set is reorganized into instruction subsets And build a "state-event-instruction" mapping relationship.

[0147] Step 6: Based on the distributed production interaction and autonomous collaboration method driven by CPS node self-organization, complete the process task allocation and form a global Gantt chart;

[0148] The above discrete event control system modeling based on stream data processing solves the problem of CPSPS distributed operation control. However, the ultimate goal of building CPSPS is to allocate process tasks in the form of autonomous collaboration of CPS nodes and generate a task Gantt chart. This embodiment proposes a method for realizing the self-matching problem of processing equipment and processes through production social interaction and autonomous collaboration between CPS nodes, that is, using ontology-based and rule-based reasoning to achieve preliminary demand-capability matching, identify the solution space of candidate processing equipment, and then obtain the optimal processing equipment based on the grey correlation analysis method, and accordingly realize the distributed production social interaction and autonomous collaboration of multiple CPS nodes around process tasks and driven by events. Its logical process is as follows: Figure 9 As shown, it includes four stages. Specifically:

[0149] Step 601: Establish a processing capability and processing requirement ontology model: Based on the configuration of the machine tool CPS node, formally describe its functional attributes and processing capabilities. At the same time, formally describe the SW process processing requirements, and ensure that each process can be processed by a machine tool CPS node;

[0150] The processing capability can be formally described by the following octet:

[0151] MC::={C_id,C_nam,C_des,C_typ,C_mat,C_fea,C_dim,C_per} (19)

[0152] Where: C_id represents the machining capability ID; C_nam represents the machining capability name; C_des represents the basic description of the machining capability; C_typ represents the set of machinable types, including turning, milling, drilling, boring, etc.; C_mat represents the set of machinable materials, including stainless steel, cast iron, aluminum, etc.; C_fea represents the set of machinable part features, including faces, holes, bosses, steps, grooves, etc.; C_dim represents the machinable size range, including length, width, height, diameter, etc.; C_per represents the achievable machining performance, which describes the degree to which a machining equipment can perform, including dimensional accuracy, shape accuracy, position accuracy, roughness, etc.

[0153] Similarly, a processing requirement can be formally described by the following octet:

[0154] MR::={R_id,R_nam,R_des,R_typ,R_mat,R_fea,R_dim,R_per} (20)

[0155] The elements represent the ID, name, basic description, optional processing type, material, workpiece characteristics, workpiece size, required geometric tolerances and surface roughness and other processing qualities of the process processing requirements.

[0156] According to the above processing capability and processing requirement information model, an ontology model based on the “capability-requirement” matching perspective is established, such as Figure 10 As shown in the figure, machining capability describes the functionality and performance of machining equipment such as machine tools and can be broken down into six subcategories: machining type, machining characteristics, machining materials, machining dimensions, machining accuracy, and surface quality. Each subcategory can be further subdivided. Similarly, machining requirements describe the machining requirements of a process, encompassing six aspects: machining type, machining characteristics, and machining materials, corresponding one-to-one to the subcategories of machining capability.

[0157] Step 602: Demand-Capability Matching: Each SW entering the processing site broadcasts a targeted message to all machine tool CPS nodes in the network, asking whether any machine tool can process the current process. The machine tool CPS node uses a matching strategy based on ontology and constraint reasoning to determine whether to accept the task, combining its current state with the task queue to be processed, and then sends feedback to the SW.

[0158] In this embodiment, the matching strategy of ontology and constraint reasoning is used to perform demand-capability matching to obtain the candidate set of machine tool CPS nodes that complete the process. Since the processing capabilities and requirements have different dimensions and importance, the matching methods are also different. Figure 11 As shown, the specific process is as follows:

[0159] Step 6021: Exact matching: Exact matching is used to achieve matching of equal or identical classes and attributes. MR i ≡MC j , then the matching of MR and MC is called exact matching. This embodiment uses the description logic inference engine Racer to achieve exact matching of processing materials.

[0160] Step 6022: Inclusive matching: Inclusive matching is used to match processing requirements within the scope of processing capabilities. Subsum(MC j ,MR i ), the matching of MR and MC is called inclusive matching. This embodiment uses the description logic inference engine Racer to achieve inclusive matching of processing dimensions.

[0161] Step 6023, similarity calculation: Under the premise of satisfying exact matching and inclusive matching, similarity calculation is used to perform conceptual similarity matching of processing types and processing features.

[0162] Considering the complex hierarchical relationship between the two classes of processing type and manufacturing feature, and the difficulty of fully and strictly defining them in the ontology, this embodiment proposes a comprehensive similarity matching algorithm in an incomplete ontology. The similarity between two concepts X and Y can be calculated using the following formula:

[0163] sim(X,Y)=α·sim Dis_Dep (X,Y)+β·sim Sup_Sub (X,Y) (21)

[0164]

[0165] Where: sim(X,Y) is the comprehensive similarity between concepts X and Y; sim Dis_Dep (X, Y) is the similarity between concepts X and Y under the influence of concept depth and semantic distance; sim Sup_Sub (X, Y) is the similarity between concepts X and Y calculated based on the superclass-subclass relationship. The meanings of other symbols are shown in Table 1.

[0166] Table 1 Definition of related symbols in similarity calculation formula

[0167]

[0168]

[0169] Step 6024: perform constraint reasoning. The CPS node comprehensively determines whether to accept the task and provides feedback based on its current state and the task queue to be processed, thereby achieving demand-capability matching.

[0170] This embodiment considers that machining accuracy and surface quality lack physical meaning when used alone and are generally used in combination with other classes or attributes. Therefore, this embodiment uses a semantic web rule language for constraint reasoning. Semantic Web Rule Language (SWRL), a W3C specification for representing rules in a semantic manner, can be used to describe rules in the IF-THEN format and is highly convenient for reasoning using the inference engine JESS.

[0171] Step 603: Each process has a set of candidate machine tool CPS nodes. Therefore, the purpose of this step is to use grey relational analysis to select the optimal machine tool CPS node from a solution space. Therefore, this embodiment constructs an evaluation index system that reflects the cost, efficiency, and quality of process processing, uses grey relational analysis to establish an optimal equipment evaluation model, and uses this optimal equipment evaluation model to determine the optimal processing equipment from the candidate processing equipment set.

[0172] Step 6031, indicator system construction: Construct an evaluation indicator system to reflect the cost, efficiency and quality of process processing, including: process processing cost (C1), process quality pass rate (C2), process processing time (C3), process on-time completion rate (C4), equipment failure rate (C5). Where: C1 = C total / N total , that is, the average cost of a single process is the total cost of all processes of processing a certain type of parts by the equipment C total Its total number of historical processes N total The ratio of C2 = (N total -N waste ) / N total , that is, the process quality qualification rate is the number of all quality qualified processes in the equipment processing a certain type of parts (N total -N waste ) and the total number of processes N total The ratio of That is, the average processing time of a process is the sum of the processing time of all processes in the equipment processing a certain type of parts. and the total number of processes N total The ratio of and They represent the process cache exit time and cache entry time monitored by RFID equipment respectively; C4=N on-time / N total , that is, the on-time completion rate of the process is the number of all on-time completed processes N in the equipment processing a certain type of parts on-time N of the total number of processes total Ratio; C5 = t failue / t working , that is, the failure rate of a single processing equipment is the downtime of a single equipment failure tfailue The load time of a single device t working ratio.

[0173] The above indicators are calculated based on the historical production records of the processing equipment.

[0174] Step 6032: Grey correlation analysis:

[0175] The grey relational analysis is to obtain the optimal processing equipment for the matching process by using the grey relational analysis evaluation method on the basis of the construction of the index system.

[0176] Step A1: Assume that m equipment evaluation indicators are constructed and n candidate processing equipment are obtained through demand-capacity matching. Based on the historical production records of the n candidate processing equipment, the m evaluation indicators are calculated to obtain n data sequences, forming the following matrix:

[0177]

[0178] Where: X i ' represents the i-th data sequence, X i '=(x i '(1),x i '(2),…,x i '(m)) T ,i=1,2,…,n。

[0179] Step A2: Determine the optimal value of each indicator based on the process processing requirement information and evaluation indicator type of the intelligent workpiece to form an ideal indicator sequence X'0, which is recorded as:

[0180] X'0=(x'0(1),x'0(2),…,x'0(m)) T (25)

[0181] Step A3: Obtain the converted evaluation index sequence and ideal index sequence through dimensionless conversion, and calculate the absolute difference between the corresponding elements of the evaluation index sequence and the ideal index sequence of each candidate processing equipment one by one;

[0182] Since the physical meanings of the indicators of the processing equipment are different, the dimensions of the data are also different. In order to facilitate comparison, the evaluation matrix (X1', X'2,..., X' n ) and the ideal index sequence X'0 are dimensionless. The dimensionless conversion method is as follows:

[0183]

[0184] Where: represents the minimum value of the kth indicator; Represents the maximum value of the kth index, where k = 1, 2, …, m and i = 0, 1, …, n.

[0185] By non-dimensionalization, the converted evaluation matrix and ideal index sequence are obtained, and the absolute difference between the corresponding elements of the evaluation index sequence of each processing equipment and the ideal index sequence is calculated one by one, that is, |x0(k)-x j (k)|, on this basis, the correlation coefficient ζ is obtained i (k), denoted as:

[0186]

[0187] Where: i (k) represents the correlation coefficient between the kth evaluation index of the i-th processing equipment and the ideal index; ρ represents the discrimination coefficient, ρ∈(0,1). The smaller ρ is, the greater the difference between the correlation coefficients and the stronger the discrimination ability. Usually ρ is taken as 0.5.

[0188] Step A4: Calculate the weighted average of each indicator for each processing equipment based on the degree of influence of each indicator on the selection of the optimal processing equipment:

[0189]

[0190] Where: r i Indicates the matching degree between the i-th processing equipment and the process. The larger the value, the higher the matching degree. k Represents the weight of the kth indicator, the weight vector W=(w1,w2,…,w m The value of ) can be obtained through the classic Analytic Hierarchy Process (AHP), which will not be described in detail here.

[0191] Step A5: According to the weighted average r i The matching degree between the candidate processing equipment set and the process is sorted from high to low according to the size of , and the processing equipment with the highest ranking is selected as the optimal processing equipment for the process.

[0192] Step 604: After the optimal machine tool CPS node is selected, the process flows according to the RFID events of the process diagram model. During the processing phase of the diagram model, the working condition data is continuously monitored to monitor the equipment working status driven by sensor events and CPS node events. By continuously summarizing the matching results between processes and machine tools and adding them to the process task Gantt chart, the Gantt chart is dynamically updated, ultimately forming a global Gantt chart to support the processing progress monitoring of order tasks. The process of autonomous production communication and negotiation is as follows:

[0193] (1) SW enters the production site through the access control, triggers the RFID access control event, and publishes the processing requirement information of the current process to all machine tool CPS nodes;

[0194] (2) All machine tool CPS nodes receive the process processing requirement information from the RFID tag of the SW, match it with their own processing capabilities, and evaluate whether they meet the processing requirements;

[0195] (3) All machine tool CPS nodes feed back the evaluation results to the SW. In addition to information on whether the process can be processed, the results also include various evaluation indicators required for grey relational analysis and other historical production records.

[0196] (4) SW receives all matching results, forms a set of candidate machine tool CPS nodes, and obtains a sequence of machine tool CPS nodes with a decreasing matching degree based on grey correlation analysis;

[0197] (5) SW sends a contract request to the machine tool CPS node with the highest matching degree and assigns the current process to the machine tool CPS node;

[0198] (6) The target machine tool CPS node sends a confirmation or rejection message to the SW: if the contract is signed, it enters the second stage for process processing; if the contract fails, the SW selects the next machine tool CPS node from the candidate machine tool CPS nodes according to the matching degree from high to low to sign the contract.

[0199] Step 7: Execute production tasks according to the global Gantt chart.

[0200] Example 2:

[0201] like Figure 12 As shown, this embodiment proposes an event-driven CPSPS distributed operation control system for implementing the method described in Example 1, including:

[0202] The first model building module is used to build a process diagram model of the machining process based on the production event-related concepts of the defined CPSPS operation;

[0203] The second model building module is used to build RFID event models, sensor event models and CPS node event models containing event trigger conditions, and determine the corresponding application scenarios of each event model;

[0204] The semantic annotation module is used to establish an event ontology model and perform semantic annotation on sensor stream data to obtain RDF stream data;

[0205] The event instance generation module is used to continuously query and monitor RDF stream data using a sliding time window, and uses the improved RETE pattern matching algorithm to match the continuously output query results with event trigger conditions to output event instances;

[0206] An event processing mechanism building module is used to build an event processing mechanism based on event instances using finite automata theory;

[0207] The process allocation module is used to complete process task allocation and form a global Gantt chart using a distributed production interaction and autonomous collaboration method driven by CPS node self-organization;

[0208] The execution module is used to execute production tasks according to the global Gantt chart.

[0209] Example 3:

[0210] This embodiment provides a computer terminal, including: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the above-mentioned event-driven CPSPS distributed operation control method is implemented.

[0211] Example 4:

[0212] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the above-mentioned event-driven CPSPS distributed operation control method is implemented.

[0213] In summary, the present invention aims at the new idea of distributed production social and autonomous collaboration and edge computing in the context of fast data generation speed and high real-time processing requirements in the current human-machine-object interconnected manufacturing environment. First, a process diagram model of the processing process is proposed, and the production event-related concepts of CPSPS operation are defined; secondly, an RFID event model, a sensor event model and a CPS node event model containing event triggering conditions are established respectively, and the corresponding application scenarios are described; thirdly, in order to effectively control the real-time sensor stream data of CPSPS, an event ontology model is established, the sensor stream data is semantically annotated, and RDF stream data is obtained. A sliding time window is used for R DF stream data is continuously queried and monitored, and the continuously output query results are pattern-matched with event trigger conditions to output event instances. An event processing mechanism is then constructed based on finite automata theory. Finally, to achieve swarm intelligence for self-organization of production equipment, a distributed production interaction and collaboration method driven by CPS node self-organization is proposed. A processing capability and processing requirement model for CPS nodes and intelligent workpieces is established. A matching strategy based on ontology and constraint reasoning is used to perform demand-capability matching, resulting in a set of candidate processing equipment. Furthermore, a gray key analysis model is used to establish an optimal equipment evaluation model, determine the optimal processing equipment, and conduct production task negotiation and contract signing to complete process task allocation. Unlike the traditional method of pre-generating a production planning Gantt chart for production task execution, the present invention generates a Gantt chart through distributed production social networking among multiple CPS nodes. The Gantt chart is dynamically updated over time and is the output of the distributed operation control of the SocialF-oriented CPSPS system. This achieves event-driven distributed production social networking and autonomous collaboration among CPS nodes.

[0214] The technical solution provided by the present invention is described in detail above. In this embodiment, specific examples are used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified in a number of ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. An event-driven CPSPS distributed operation control method, characterized in that: The steps include: Define the concepts related to production events of CPSPS operation and establish a process diagram model for the processing; Establish RFID event models, sensor event models, and CPS node event models that include event triggering conditions, and determine the corresponding application scenarios for each event model; Establish an event ontology model, perform semantic annotation on sensor stream data, and obtain RDF stream data; A sliding time window is used to continuously query and monitor RDF stream data. The improved RETE pattern matching algorithm is used to match the continuously output query results with event trigger conditions and output event instances. Based on the finite automaton theory, an event processing mechanism is constructed according to event instances; Based on the distributed production interaction and autonomous collaboration method driven by CPS node self-organization, process task allocation is completed and a global Gantt chart is formed; Execute production tasks according to the global Gantt chart.

2. The event-driven CPSPS distributed operation control method according to claim 1 is characterized in that: The production event-related concepts of the CPSPS operation include manufacturing process, object status, observation, stream data generated by observation, production event, event condition, and event detection.

3. The event-driven CPSPS distributed operation control method according to claim 1 is characterized in that: The process of establishing the RFID event model is as follows: Collect process logistics data to form process flow data with timestamps; Process process flow data, extract physical state changes that meet event conditions, match them with event trigger conditions, and generate RFID events; The process of establishing the sensor event model is as follows: The sensor continuously collects attribute feature data of physical objects at a user-defined sampling frequency, forming sensor stream data with time stamps; Process the collected sensor stream data and extract relevant indicators to determine whether the predefined event trigger conditions are met. If so, a sensor event is generated; The process of establishing the CPS node event model is as follows: Combining several sensor events into composite events according to certain rules; Determine whether the composite event meets the predefined event triggering conditions. If so, generate a CPS node event.

4. The event-driven CPSPS distributed operation control method according to claim 1, characterized in that: The specific process of establishing an event ontology model, semantically annotating sensor stream data, and obtaining RDF stream data is as follows: Construct an event ontology model between production event-related concepts such as sensors, observations, events, event conditions, and event attributes; Use ontology semantic queries to identify relevant sensors and the data they collect; The semantic annotation tool Jena is used to annotate the original stream data to obtain RDF stream data.

5. The event-driven CPSPS distributed operation control method according to claim 1 is characterized in that: The specific process of the improved RETE pattern matching algorithm is as follows: Establishing an improved Rete network; The query results of continuous query output of RDF stream data using sliding time window are input into the improved Rete network for pattern matching; Determine whether the event trigger mode is met, and if so, generate an event instance.

6. The event-driven CPSPS distributed operation control method according to claim 5 is characterized in that: The improved Rete network establishment process is as follows: Establish a sharing degree model, calculate the pattern sharing degree of the atomic conditions in the event trigger rule set, and rearrange the order of the atomic conditions in the event trigger rule set based on this, and output the sorted rule set; Generate an Alpha network based on the sorted rule set; Classify the nodes in the Alpha network, sort the Alpha classifications in descending order according to the Alpha classification sharing degree and the average sharing degree, and output the sorted Alpha network; Based on the sorted Alpha network, the Beta nodes are combined to output the improved Rete network.

7. The event-driven CPSPS distributed operation control method according to claim 1 is characterized in that: The process of the distributed production interaction and autonomous collaboration method based on CPS node self-organization is as follows: Establish processing capability and processing requirement ontology model; The matching strategy of ontology and constraint reasoning is used to perform demand-capability matching and obtain the candidate processing equipment set; Construct an evaluation index system to reflect the cost, efficiency and quality of process processing, establish an optimal equipment evaluation model using grey correlation analysis, and use the optimal equipment evaluation model to determine the optimal processing equipment from the candidate processing equipment set; According to the RFID events of the process diagram model, the matching results of the process and the machine tool are continuously summarized to negotiate and sign the production task, and added to the process task Gantt chart to complete the process task allocation, and finally form a global Gantt chart.

8. The event-driven CPSPS distributed operation control method according to claim 7 is characterized in that: The specific process of using the ontology and constraint reasoning matching strategy to perform demand-capability matching is as follows: The description logic inference engine Racer is used to achieve accurate matching of processing materials; The description logic inference engine Racer is used to achieve inclusive matching of processing dimensions; Under the premise of satisfying exact matching and inclusive matching, similarity calculation is used to perform conceptual similarity matching of processing types and processing features; By performing constraint reasoning, the CPS node comprehensively determines whether to undertake the task and provides feedback based on its current status and the task queue to be processed, thus achieving demand-capability matching.

9. The event-driven CPSPS distributed operation control method according to claim 7, characterized in that: The specific process of establishing the optimal equipment evaluation model using the grey correlation analysis method is as follows: Calculating each evaluation indicator in the evaluation indicator system based on historical production records of each candidate processing equipment in the candidate processing equipment set; According to the process processing requirement information and evaluation index type of the intelligent workpiece, the optimal value of each evaluation index is determined to form an ideal index sequence; By non-dimensionalization, the converted evaluation index sequence and ideal index sequence are obtained, and the absolute difference between the corresponding elements of the evaluation index sequence and the ideal index sequence of each candidate processing equipment is calculated one by one; According to the degree of influence of each indicator on the selection of the optimal processing equipment, the weighted average value of each evaluation indicator of each candidate processing equipment is calculated; According to the size of the weighted average, the matching degree between the candidate processing equipment set and the process is sorted from high to low, and the processing equipment with the highest ranking is selected as the optimal processing equipment for the process.

10. An event-driven CPSPS distributed operation control system for implementing the method according to any one of claims 1 to 9, characterized in that: include: The first model building module is used to build a process diagram model of the machining process based on the production event-related concepts of the defined CPSPS operation; The second model building module is used to build RFID event models, sensor event models and CPS node event models containing event trigger conditions, and determine the corresponding application scenarios of each event model; The semantic annotation module is used to establish an event ontology model and perform semantic annotation on sensor stream data to obtain RDF stream data; The event instance generation module is used to continuously query and monitor RDF stream data using a sliding time window, and uses the improved RETE pattern matching algorithm to match the continuously output query results with event trigger conditions to output event instances; An event processing mechanism building module is used to build an event processing mechanism based on event instances using finite automata theory; The process allocation module is used to complete process task allocation and form a global Gantt chart using a distributed production interaction and autonomous collaboration method driven by CPS node self-organization; The execution module is used to execute production tasks according to the global Gantt chart.

Images