CN118348918A

CN118348918A - Multi-source data sensing, fusion and predictive analysis method and system for aircraft assembly production line

Info

Publication number: CN118348918A
Application number: CN202410229879.0A
Authority: CN
Inventors: 郭飞燕; 张硕; 石金康; 马健; 鲍强伟
Original assignee: University of Science and Technology Beijing USTB
Current assignee: University of Science and Technology Beijing USTB
Priority date: 2024-02-29
Filing date: 2024-02-29
Publication date: 2024-07-16

Abstract

The invention provides a method and a system for multi-source data perception, fusion and predictive analysis of an aircraft assembly production line, wherein the method comprises the following steps: s1: analyzing data acquisition objects and information acquisition requirements of an aircraft assembly production line; s2: constructing a multi-source data acquisition system of an aircraft assembly production line by adopting an OPC production cascade technology; s3: characterizing the multi-source data entity of the production line; s4: constructing a multi-source data ontology model of an aircraft assembly production line; s5: instantiating a description production line multisource data entity; s6: designing a long-short-period memory network model considering the time sequence characteristics of the assembly data; s7: designing an analysis prediction model training method based on actually measured multi-source data; s8: and constructing a multisource assembly data analysis and prediction model with time sequence characteristics. The invention realizes the application of production line operation monitoring, equipment state detection and assembly quality improvement of the aircraft assembly production line, and provides a high-quality data basis for virtual efficient operation of the aircraft assembly production line.

Description

Multi-source data sensing, fusion and predictive analysis method and system for aircraft assembly production line

[ Field of technology ]

The invention relates to the field of intelligent manufacturing, in particular to a method and a system for multi-source data sensing, fusion and predictive analysis of an aircraft assembly production line.

[ Background Art ]

With the development, aircraft manufacturing processes are continually evolving towards intelligent, informative modes. The aircraft assembly production line is used as a key link of an aircraft manufacturing system, and is an important factor for ensuring high quality and high efficiency of aircraft manufacturing. In the aircraft assembly production process, various parts are gradually installed through each station according to an assembly process flow in a certain sequence, and a complete product is formed after all the stations are passed through, wherein the process comprises a series of operations such as posture adjustment, butt joint and measurement of various large parts of an aircraft. In the traditional aircraft assembly production line, a large number of fixed tool type frames are used, and the method ensures the stability and accuracy of assembly to a certain extent. However, as the market demand for aircraft increases, enterprises are facing urgent demands for improved assembly production efficiency. In order to meet market demands, advanced aircraft digital assembly production lines gradually become mainstream, and a large number of digital and flexible hardware process equipment and digital control systems, such as industrial robots, numerical control positioners, automatic drilling and riveting machines and the like, are adopted, so that the assembly quality and efficiency of the aircraft are greatly improved.

The aircraft can be used for various tasks such as transportation, sightseeing, search and rescue, military operations and the like, and in order to meet the requirements of different tasks, the aircraft is suitable for the aircraft assembling processes of different models and configurations, different types of assembling execution equipment and detection equipment are required to be equipped in a production line, namely, in a mixed line production mode, quick adjustment and adaptation capability are required. Meanwhile, the aircraft has a plurality of models, a linear manufacturing mode is difficult to form, the equipment acquisition data of an assembly production site is distributed, and a production line management system capable of centrally managing data is needed. With the wide application of the Saobo-physical system in manufacturing industry, the real production line operation process can be simulated by constructing an information model of an assembly production line in a virtual space, so that an effective solution is provided for the aspects of production line operation optimization, equipment state management, assembly quality improvement and the like in assembly production. In the virtual operation process of the assembly production line, a large amount of multi-source heterogeneous data such as production management and control data, assembly scheduling results, logistics distribution information, component quality parameters, tool equipment use parameters, quality detection results and the like exist, and the problems of various production line data structures, non-uniform description forms, low utilization rate and the like exist, so that data acquisition is difficult and insufficient, data fusion efficiency is low, an information set conforming to uniform description grammar is difficult to form, and comprehensive and effective data support cannot be provided for production line state analysis and continuous optimization of multiple assembly stages.

Therefore, it is necessary to study a method and a system for sensing, fusing and predicting multi-source data in an assembly line, so as to realize acquisition of multi-source data, unified characterization of data format, prediction processing analysis of data training and the like in the operation process of the assembly line, thereby providing a data basis for digital twin construction and virtual efficient operation of the assembly line, and solving or alleviating one or more of the problems in the prior art.

[ Invention ]

In view of the above, the invention provides a method and a system for sensing, fusing and predicting and analyzing multi-source data of an aircraft assembly line, which combine the characteristics of measured data and assembly processes of all acquisition devices of the product assembly line, and can realize the application of the multi-source data sensing and fusing of the assembly line in the aspects of production line operation monitoring, device state detection and assembly quality improvement by combining the multi-source data acquisition system, the fusion technology and the construction of an assembly data analysis and prediction model of the assembly line, thereby providing a data basis for the digital twin construction and virtual efficient operation of the aircraft assembly line.

On one hand, the invention provides a multi-source data sensing, fusion and predictive analysis method for an aircraft assembly production line, which is used for the application of the aspects of production line running state analysis, equipment use state monitoring, assembly quality improvement and the like, and comprises the following steps:

S1: analyzing data acquisition objects and information acquisition requirements of an aircraft assembly production line;

S2: constructing a multi-source data acquisition system of an aircraft assembly production line by adopting an OPC production cascade technology;

s3: characterizing the multi-source data entity of the production line by adopting a semantic description method based on the ontology;

s4: constructing a multi-source data ontology model of an aircraft assembly production line;

s5: instantiation description of the production line multisource data entity by adopting JSON language;

S6: designing a long-short-period memory network model considering the time sequence characteristics of the assembly data;

s7: designing an analysis prediction model training method based on actually measured multi-source data;

s8: and constructing a multisource assembly data analysis and prediction model with time sequence characteristics.

In the aspect and any possible implementation manner as described above, there is further provided an implementation manner, where the S1 specifically includes:

S11: analyzing the process of an aircraft assembly production line, namely, component assembly, body riveting and whole assembly, wherein in the aspect of data acquisition, the data acquired by the production line mainly comprise two types of monitoring data and quality control data from the use process of assembly equipment on the production line;

S12: and analyzing the specific types of equipment execution equipment and quality detection equipment related in the aircraft assembly production line, defining attribute description of related data of the equipment, and defining equipment information required to be acquired in the running process of the production line by combining production practice.

In the aspect and any possible implementation manner as described above, there is further provided an implementation manner, where the device information to be collected during the operation of the production line in S12 mainly includes: the equipment running state information (such as start-stop state, rotation speed of each shaft and the like in the intersection point hole gantry finish machining system), the actual values of production process parameters (such as adjustment quantity and gesture adjustment track of each motion shaft and the like in the machine body assembly digital butt joint assembly system), the alarm information (such as fault and abnormal data of equipment), the equipment working mode and the like.

In the aspect and any possible implementation manner as described above, there is further provided an implementation manner, where the S2 specifically includes:

S21: designing data interfaces on assembly execution equipment and an assembly detection unit in a production line, and collecting data of all the equipment;

s22: aiming at the related data which are communicated and acquired by the equipment to be tested in the acquisition layer, the OPC protocol is adopted to transmit the data to a monitoring system or other application programs, so that the data are transmitted among different equipment and systems;

S23: KEP SERVER EX is selected as a data acquisition OPC server, is connected with various devices, converts acquired data into standard communication protocols such as OPC UA, OPC Client and the like, loads an OPC information model, and externally distributes the device information model and the data in an OPC UA protocol format.

In the aspect and any possible implementation manner as described above, there is further provided an implementation manner, where the S3 specifically includes:

s31: analyzing the functions of all the devices on the production line, collecting data and assembling process information, adopting an ontology semantic description method to enumerate important concepts in a multi-source data ontology model of the production line, and defining hierarchical relations of concept classes;

S32: and analyzing object data and data attributes in the assembly process information, taking modeling of assembly process information as a main study object, defining object attributes and data attributes, and assisting in modeling of various description attributes of assembly tasks and assembly equipment and association relations of various concept classes.

In the aspect and any possible implementation manner as described above, there is further provided an implementation manner, where the assembly line multisource data in S31 is mainly divided into two types of assembly equipment and assembly process; the assembly equipment mainly comprises assembly execution equipment and quality detection equipment, and the assembly process information generally comprises assembly tasks and assembly procedures.

In the foregoing aspect and any possible implementation manner, there is further provided an implementation manner, where the S4 specifically includes:

S41: creating a new ontology project in ontology modeling software, setting the identification of the multi-source data ontology model of the aircraft assembly line, creating concept classes and concept class hierarchical relations according to the concept classes in the multi-source data ontology model of the aircraft assembly line defined in S3, adding examples under the concept classes, and filling specific objects under each class.

S42: and S3, defining object attributes in the multi-source data ontology model of the aircraft assembly production line and data attributes of assembly procedures in the multi-source data ontology model of the aircraft assembly production line, creating concept classes, object attributes and data attributes, logically reasoning the ontology by utilizing an reasoning mechanism of ontology modeling software, verifying that the logical structure of the ontology is correct, storing ontology items, and completing the construction of the ontology model.

In the foregoing aspect and any possible implementation manner, there is further provided an implementation manner, where the S5 specifically includes:

S51: and analyzing a multisource data ontology model of the aircraft assembly production line, wherein the multisource data ontology model comprises information descriptions of entities, attributes, relations and the like.

S52: and (3) instantiating and setting up a multisource data ontology model of the aircraft assembly production line by adopting a JSON language format, and generating JSON data packets of each execution device and each detection device.

In the foregoing aspect and any possible implementation manner, there is further provided an implementation manner, where the S6 specifically includes:

S61: the architecture of the long-term and short-term memory network model is clarified, the architecture comprises an input layer, an LSTM layer and an output layer, and the input layer and the output layer of the model architecture are determined by considering the time sequence characteristic of the assembly process data.

S62: and analyzing the assembly production line process and the LSTM layer architecture, and designing the LSTM layer from three aspects of an activation function, a loss function and a network training method.

In the foregoing aspect and any possible implementation manner, there is further provided an implementation manner, where the S7 specifically includes:

S71: analyzing and acquiring a time sequence sample set under the execution time of each procedure from the acquired data, dividing the time sequence sample set into a training set and a testing set, and extracting different subsets from the training set, wherein each subset contains a part of training set features, and the features in each subset can be repeated.

S72: selecting each subset by adopting L1 regularization features to obtain a plurality of feature subsets, and solving the intersection of the feature subsets to obtain key influence factors in the finally selected input layer data; aiming at the assembly quality problem prediction of model output, the quality problem is converted into a binary vector form which is easy to process by a computer according to the type of the quality problem by adopting single thermal coding, and the binary vector form is used for prediction model training.

S73: and taking the processed data as sample data, optimizing parameters by using an Adam algorithm, and dividing the time sequence sample data into a training set and a testing set for training the constructed production line data analysis prediction model. If the change trend of the loss function of the model test set is the same as that of the loss function of the training set, the model test set and the training set are respectively shown as the loss function which is rapidly reduced and gradually converged to the lowest loss value in the initial stage, and the LSTM prediction model can be reasonably constructed.

In the foregoing aspect and any possible implementation manner, there is further provided an implementation manner, where the S8 specifically includes:

s81: according to the time sequence characteristic of the aircraft assembly process data, data related to aircraft assembly are collected, wherein the data comprise assembly time, assembly procedures, assembly equipment information and the like;

S82: dividing the data into a plurality of sequence samples according to the size of a time window, wherein each sample contains continuous data in a period of time, and constructing an aircraft assembly process data analysis prediction model based on a long-period memory network.

S83: the method comprises the steps of determining input and output of a prediction model, wherein an input layer is actual measurement data such as assembly time, assembly working procedures, assembly equipment information and the like, an output layer is performance index prediction results such as assembly qualification rate, working procedure execution rate, equipment utilization rate and the like of the current assembly production line operation, training and optimizing the model by adopting a network training method, and constructing a multi-source data analysis prediction model based on LSTM.

In accordance with aspects and any one of the possible implementations described above, there is further provided an assembly line multisource data sensing, fusion and predictive analysis system, the data sensing, fusion and predictive analysis method including the steps of:

The object and demand analysis module is used for analyzing data acquisition objects and information acquisition demands of the aircraft assembly production line;

The system construction module is used for constructing a multi-source data acquisition system of an aircraft assembly production line by adopting an OPC production cascade technology according to data acquisition objects and information acquisition requirements;

The entity characterization module is used for characterizing the entity in the production line multi-source data system by adopting a semantic description method based on the ontology;

the body model construction module is used for constructing a multi-source data body model of the aircraft assembly production line through the entity characterization result;

The instantiation description module is used for instantiating and describing the multisource data ontology model by adopting a JSON language;

The long-term memory network model construction module is used for constructing a long-term memory network model considering the time sequence characteristics of the assembly data according to the instantiation description result;

the training method determining module is used for designing an analysis prediction model training method based on the actually measured multi-source data according to the long-term memory network model;

The multi-source assembly data analysis and prediction model construction module is used for designing a multi-source assembly data analysis and prediction model, training and optimizing according to an analysis and prediction model training method, and obtaining the multi-source assembly data analysis and prediction model with time sequence characteristics.

Compared with the prior art, the invention can obtain the following technical effects:

1) The data acquisition system of the invention covers the whole scene of an assembly line and special assembly equipment used for the structural characteristics of an airplane: for example, a machine body assembly butting device, a solid rivet fastener automatic drilling and riveting device, a component quality detection unit, a water tightness intelligent test device, a multi-machine body section butting intersection point hole gantry finish machining device and the like are adopted, and a unified industrial control system communication interface standard OPC technology is adopted to more efficiently and fully acquire real-time data of key assembly steps of an assembly production line;

2) According to the invention, body model construction is carried out on assembly equipment and assembly process data related in an aircraft assembly production line, six aspects of entity position information, state information, time information, behavior information, event information and history information are used for describing the multi-source data entity of the aircraft assembly production line, and then a prot g software is adopted for constructing the multi-source data body model of the aircraft assembly production line, so that the model has a good hierarchical structure and clear logic relationship, the defect of a single data system can be overcome, and data from different sources can be integrated to form a comprehensive and accurate information set, so that a sufficient data basis is provided for digital twin construction and virtual operation of the aircraft assembly production line;

3) In view of the time sequence characteristics of the aircraft assembly stations, a data analysis prediction model for integrating all assembly process time periods is constructed, and performance indexes such as the next stage assembly productivity, assembly qualification rate, process execution rate, equipment utilization rate and the like can be more accurately predicted.

Of course, it is not necessary for any of the products embodying the invention to achieve all of the technical effects described above at the same time.

[ Description of the drawings ]

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of a method and a system for sensing, fusing and predictive analysis of multi-source data in an assembly line according to the present invention;

FIG. 2 is a schematic diagram of a multi-source data fusion process for an aircraft assembly line according to the present invention;

FIG. 3 is a schematic diagram illustrating the attributes of specific equipment and its associated data in an aircraft assembly line in accordance with the present invention;

FIG. 4 is a schematic diagram of an aircraft assembly line operational data acquisition system architecture according to the present invention;

FIG. 5 is a schematic view of the overall structure of a data acquisition layer in an aircraft assembly line according to the present invention;

FIG. 6 is a diagram illustrating a KEP SERVER EX server architecture according to the present invention;

FIG. 7 is a schematic diagram of a model of an aircraft assembly line multisource data ontology constructed using prot g software in accordance with the present invention;

FIG. 8 is a schematic diagram of the process of constructing an analytical predictive model of aircraft assembly process data according to the present invention;

FIG. 9 is a schematic illustration of a production line operation monitoring application of an aircraft assembly line according to the present invention;

FIG. 10 is a schematic diagram of an apparatus status detection application flow for an aircraft assembly line according to the present invention;

Fig. 11 is a schematic diagram of an assembly quality improvement application flow of an aircraft assembly line according to the present invention.

[ Detailed description ] of the invention

For a better understanding of the technical solution of the present invention, the following detailed description of the embodiments of the present invention refers to the accompanying drawings.

It should be understood that the described embodiments are merely some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

The invention solves the technical problems by adopting the following general ideas:

Firstly, analyzing data acquisition objects, data acquisition equipment and data acquisition service architecture in an assembly production line, and constructing an aircraft assembly production line multi-source data acquisition system based on an OPC production product union technology; (OPC is OLE for Process Control in the present invention, OLE for process control is an industry standard); furthermore, a semantic description method based on an ontology is adopted to describe the multi-source data entity of the aircraft production line from six aspects of position information class, state information class, time information class, behavior information class, event information class and history information class of the entity, a multi-source data ontology model of the aircraft assembly production line is constructed, and entity instantiation description is carried out; secondly, taking the fact that the aircraft assembly process data has time sequence characteristics into consideration, constructing a multi-source data analysis and prediction model framework based on a long-period memory network, and realizing analysis and prediction of multi-source data in an assembly production line through model training and verification of various model parameters and collected data; subsequently, by combining the actual measurement data and the assembly process characteristics of all the acquisition devices of the aircraft assembly production line, and constructing a multisource data acquisition system, a fusion technology and an assembly data analysis prediction model of the aircraft assembly production line, a solution is provided for the application of the aircraft assembly production line in the aspects of production line operation monitoring, device state detection and assembly quality improvement.

The invention provides an assembly production line multisource data sensing, fusing and predictive analysis method, which is applied to the aspects of production line running state analysis, equipment use state monitoring, assembly quality improvement and the like, and comprises the following steps:

s1: analyzing data acquisition object and information acquisition requirement of aircraft assembly production line

Specifically, the main components and the operation sequence of the product assembly line are first clarified. The aircraft assembly line is mainly divided into three main parts including component assembly, fuselage riveting and full aircraft final assembly. At first, the components of the aircraft are assembled, and operators accurately assemble the components such as the rotor, the tail boom, the frame of the aircraft body and the like by means of tooling equipment on a production line according to design drawings and technological requirements so as to ensure the assembly precision of each component. Then, the aircraft body is riveted and assembled, the aircraft body mainly comprises a front aircraft body, a middle aircraft body and a tail structure, when the front aircraft body and the middle aircraft body are assembled, an operator usually adopts a frame beam riveting mode, and the tail structure and the middle aircraft body adopt a flange screwed mode, so that a complete aircraft body structure is assembled. And finally, performing full aircraft final assembly of the aircraft, and integrating all parts of the assembled parts and the riveted parts of the aircraft body by operators through tooling equipment to form a complete aircraft.

Further, through the process analysis of the aircraft assembly production line, the production line is known to comprise a plurality of production links, each link relates to a large number of tool execution devices and detection devices, two types of data including process monitoring data and quality control data are required to be collected from assembly devices on the production line, debugging and testing are carried out on each device of the aircraft, and normal operation of each device is ensured.

Further, in connection with production practice, the main specific types of equipment executing devices, quality detecting devices involved in an aircraft assembly line are analyzed, and a description of attributes of specific devices and their associated data in the line are listed as shown in fig. 3. In the aspect of monitoring data in the use process of the assembly equipment, the data monitoring record of the assembly process is carried out by aiming at the use process of the butt joint, finish machining and drilling and riveting equipment in the production line; in terms of quality control data, quality information of a product assembly process in the running process of a production line is defined by acquiring quality detection results of a part quality intelligent detection unit and air tightness intelligent test equipment.

Further, according to the description of the equipment attribute information, the equipment information required to be collected in the running process of the clear production line mainly comprises: the equipment running state information (such as start-stop state, rotation speed of each shaft and the like in the intersection point hole gantry finish machining system), the actual values of production process parameters (such as adjustment quantity and gesture adjustment track of each motion shaft and the like in the machine body assembly digital butt joint assembly system), the alarm information (such as fault and abnormal data of equipment), the equipment working mode and the like.

S2: construction of multi-source data acquisition system of aircraft assembly line by OPC production cascade technology

Specifically, a unified industrial control system communication interface standard OPC is adopted, an aircraft data acquisition network system is constructed from a data acquisition layer, a data transmission layer and a service layer, and the data acquisition system architecture is shown in figure 4. And the data acquisition layer is mainly used for carrying out data acquisition on all the devices by assembling the execution device and the data interface on the assembly detection unit in the production line. The equipment comprises a main body assembly butt joint device, automatic drilling and riveting equipment for solid rivet fasteners, a component quality detection unit, a water tightness intelligent test device, a multi-fuselage section butt joint intersection point hole gantry finishing equipment and other stations which are provided with intelligent equipment, wherein the intelligent equipment is provided with SCADA systems, and production data are read from equipment interfaces and uploaded to a data acquisition service system through the SCADA systems arranged on equipment station machines.

Further, for individual devices that do not have an integrated SCADA system, but still require data acquisition, external data acquisition devices may be used to acquire real-time data of the device, such as through sensors, data acquisition modules, or other hardware, and may be connected to a separate SCADA system or monitoring platform. The data acquisition layer is mainly connected with industrial equipment directly or indirectly through the interface module, data acquisition is carried out, after the data acquisition is completed, the data acquisition layer is checked and packaged through the processing module and is sent to the data transmission layer according to a certain communication protocol, and the overall structure of the data acquisition layer is shown in fig. 5.

Further, according to the related data which are communicated and acquired by the equipment to be tested in the acquisition layer, the data are transmitted to a monitoring system or other application programs by adopting an OPC protocol, the acquired data are converted into standard communication protocols such as OPC UA, OPC Client and the like, and the communication protocols are accessed to a subsequent server, so that the data are transmitted between different equipment and systems.

Further, for the characteristics of the aircraft assembly production line, KEP SERVER EX developed by KEP ware Technologies company is selected as a data acquisition OPC server, various devices such as a fuselage assembly docking device, a component quality detection unit and the like are connected, the acquired data are converted into standard communication protocols such as OPC UA and OPC Client, the standard communication protocols are loaded into an OPC information model, the device information model and the data are externally released in an OPC UA protocol format, and the architecture of the KEP SERVER EX server is shown in fig. 6. In KEP SERVER EX servers, server objects are managed in a grouped manner, and each server contains one or more group objects, while preserving its own setting and status information. The group object contains one or more item objects, also holding its own setting and status information. The item objects act as connections between the server and the device data sources, managed and operated upon by the group object. In addition, the KEP SERVER EX server is also provided with a data storage plug-in (Data Logger) for defining variables and collection periods of data collection, and an interface for accessing an SQL database is provided, so that the collected data can be automatically written into the database in batches for reading and analysis statistics of a host system.

S3: representing the multi-source data entity of the production line by adopting semantic description method based on the ontology

Specifically, an ontology semantic description method is adopted to enumerate important concepts in the production line multi-source data ontology model, and hierarchical relations of concept classes are defined. The method comprises the steps of analyzing the functions of all equipment on a production line, collecting data and assembling process information, wherein the object described by the multi-source data ontology model of the aircraft assembling production line is entity, the multi-source data of the assembling production line is mainly divided into two types of assembling equipment and assembling process, the assembling equipment mainly comprises assembling execution equipment and quality detection equipment, and the assembling process information generally comprises assembling tasks and assembling procedures. The assembly process in the assembly process information carries most of information description tasks, including modeling of elements such as assembly equipment, assembly tasks of each process, assembly actions and the like.

Further, the modeling of the assembly process information is taken as a main research object, and the multi-source data entity of the aircraft production line is described from six aspects of position information class, state information class, time information class, behavior information class, event information class and history information class of the entity. The position information is used for describing the position of the equipment entity in the production line, and comprises relative station positions and absolute coordinates in an assembly workshop; the state information is used for describing the working state of equipment entities in the production line, including various aspects such as a standby state, a normal running state and the like; the action information class is used for describing the operation content of the execution device in the entity in the assembly process, and the operation information can control the operation of the execution device; event information classes are used to collect and deliver notifications of related events during the assembly process of the device entity. The historical information class is used to record the specific time and operations performed by the equipment entity during the assembly process, and the summary of the concept class in the multi-source data ontology model of the aircraft assembly line is shown in table 1.

Table 1 concept classes in multisource data ontology model for aircraft assembly line

Further, attributes of the concept class are defined, including object attributes and data attributes. In the process of building an ontology model, an object attribute is a key concept and is used for describing relationships among different concept class instances. The object properties present in the multi-source data ontology model of the aircraft assembly line are shown in table 2. Object attributes hasEquipment, hasLocation, etc. represent that an instance in the entity class possesses related information such as equipment, location information class, etc. observe shows that the historical information class has the time, events, and actions to be performed by the observation and recording device entity during the assembly process, etc.

Table 2 object properties in multisource data ontology model for aircraft assembly line

In the process of building the body model, the data attribute is used for describing specific properties such as weight, position, date and the like of a concept class or an individual, and generally relates to specific data types such as numerical values, texts, date and the like, and the data attribute of an assembly procedure in the multi-source data body model of the aircraft assembly production line is shown in table 3. Examples of concept classes that can be derived from the fields and values defined in Table 3 include a variety of data types. The input and output parameters of the motion are action_input and action_output respectively, which are of character string type, and JSON character string design is adopted here so as to express different motion parameters. And extracting and analyzing input parameters in the action operation, and obtaining and converting the result into a JSON character string for output after the processing is completed.

Table 3 data attributes of assembly procedures in multisource data ontology model for aircraft assembly line

S4: constructing a multisource data ontology model for an aircraft assembly line

Specifically, a new ontology project is created in ontology modeling software, an ontology project is created in the software, the identification of an aircraft assembly line multi-source data ontology model is set, the Entities options in the software are opened according to the concept class in the aircraft assembly line multi-source data ontology model defined in table 1, the CLASS ENTITIES options in the software are selected to create the concept class and the concept class hierarchical relationship, and an instance is added under the concept class to fill the specific object under each class.

Further, a Entities option in the software is opened, object attributes in the multi-source Data ontology model of the aircraft assembly line and Data attributes of assembly procedures in the multi-source Data ontology model of the aircraft assembly line are defined according to tables 2 and 3, object Properties and Data Properties options in the software are selected to create concept classes, object attributes and Data attributes, the constructed multi-source Data ontology model of the aircraft assembly line can be shown by running Onto _Graf, logical reasoning is carried out on the ontology by utilizing an reasoning mechanism of the ontology modeling software, the logical structure of the ontology is verified to be correct, and ontology items are saved, so that the construction of the ontology model is completed. The conceptual class hierarchy of the multi-source data ontology model of the aircraft assembly line is shown in fig. 7, in which solid lines represent inheritance relationships between conceptual classes, and dotted lines represent object properties between the conceptual classes.

S5: describing production line multisource data entity by using JSON language instantiation

Specifically, analyzing a multisource data ontology model of an aircraft assembly production line, wherein entities refer to six types of assembly equipment, including digital butt-joint assembly of a fuselage assembly, gantry finish machining of an intersection point hole, automatic drilling and riveting, automatic riveting, intelligent detection of component quality, intelligent water tightness testing equipment, and information description of attributes, relations and the like.

Further, a JSON format in a Python programming language is adopted to instantiate and describe six aspects of a position information class, a state information class, a time information class, a behavior information class, an event information class and a history information class for equipment entities in the production line multisource data ontology model.

S6: long-short-term memory network model with design consideration of time sequence characteristics of assembly data

Specifically, the architecture of the transparent long-term and short-term memory network model mainly comprises an input layer, an LSTM layer and an output layer. Considering that the assembly process data has a time sequence characteristic, the data is divided into a plurality of sequence samples according to the size of a time window, and each sample contains continuous data in a period of time. And determining input and output of a prediction model, wherein an input layer is serial measured data such as assembly time, assembly working procedures, assembly equipment information and the like, and an output layer is a performance index prediction result such as assembly qualification rate, working procedure execution rate, equipment utilization rate and the like of the current assembly production line operation.

Further, LSTM layer design is performed from three aspects of activation function, loss function, and network training method. In view of the complexity of assembly line data and the ability of the ELU activation function to effectively avoid gradient vanishing problems and improve model convergence, the present model selects the ELU function as the activation function for the LSTM quality prediction model. Considering that the analysis and prediction problem of the assembly data of the assembly production line belongs to the regression class problem, the root mean square error is selected as the construction basis of the subsequent model loss function. For complex systems such as aircraft assembly lines, different parameters may have different gradients, and in order to effectively adjust the update step size of each parameter and improve the training efficiency, the model adopts Adam algorithm with adaptive learning rate characteristics.

S7: analysis prediction model training method based on actually measured multi-source data

Specifically, a time sequence sample set under the execution time of each procedure is obtained from the collected data through analysis, the time sequence sample set is divided into a training set and a test set, different subsets are extracted from the training set, each subset comprises a part of training set features, and the features in each subset can be repeated.

Further, L1 regularization feature selection is carried out on each subset to obtain a plurality of feature subsets, and intersection sets of the feature subsets are obtained, namely key influence factors in the input layer are finally selected. In the aspect of model output assembly quality problem prediction, the quality problems in the aircraft assembly process mainly include assembly step difference, assembly clearance, overlarge assembly stress value and the like between assembly bodies. In order to enable the LSTM data analysis prediction model to perform better learning training, the quality problem is converted into a binary vector form which is easy to process by a computer according to the type of the LSTM data analysis prediction model by adopting single thermal coding, for example, the assembly step difference between assemblies is excessively large 0001, the assembly gap is excessively large 0010 and the like.

And further, taking the data after the data processing as sample data, optimizing parameters by using an Adam algorithm, and dividing the time sequence sample data into a training set and a test set for constructing the training of the production line data analysis prediction model. If the loss function of the model is rapidly reduced and tends to converge, and meanwhile, the data change trend of the test set and the training set is similar, the LSTM prediction model can be reasonably constructed.

S8: constructing a multisource assembly data analysis prediction model with time sequence characteristics

Specifically, the aircraft assembly process is analyzed, and data related to aircraft assembly including assembly time, assembly procedures, assembly equipment information, and the like is collected in consideration of the time sequence characteristics of the aircraft assembly process data.

Further, the data is divided into a plurality of sequence samples according to the size of a time window, each sample comprises continuous data in a period of time, and an aircraft assembly process data analysis prediction model based on a long-period memory network is constructed.

Further, input and output of a prediction model are determined, an input layer is actual measurement data such as assembly time, assembly working procedures, assembly equipment information and the like, an output layer is a performance index prediction result such as assembly qualification rate, working procedure execution rate, equipment utilization rate and the like of the current assembly production line operation, necessary characteristic engineering and data preprocessing are carried out on the data of the input layer, effective information in the data is extracted, influence of noise and abnormal values on model learning is eliminated, and therefore the model can learn the data effectively. The model is trained and optimized by continuously adjusting model parameters and adopting a network training method, so that the prediction result of the model on training data gradually approaches to an actual value, a final multi-source data analysis prediction model based on LSTM is constructed, and the construction process of the data analysis prediction model in the aircraft assembly process is shown in figure 8.

The invention also provides a multi-source data sensing, fusing and predictive analysis system of the assembly production line, wherein the data sensing, fusing and predictive analysis method comprises the following steps:

The multi-source assembly data analysis and prediction model construction module is used for designing a multi-source assembly data analysis and prediction model, training and optimizing according to an analysis and prediction model training method, and obtaining the multi-source assembly data analysis and prediction model with time sequence characteristics

Example 1:

In this embodiment, a production line operation monitoring application of a certain type of aircraft assembly line is taken as an example. The production line operation monitoring application is mainly used for predicting the equipment utilization rate and the production line beat-productivity and visually displaying the equipment utilization rate and the production line beat-productivity through a histogram and a line drawing. Taking the production line operation monitoring application of the model aircraft assembly production line as an example, the implementation steps of the multi-source data sensing, fusion and predictive analysis method of the assembly production line in the production line operation monitoring application are described.

The operational monitoring application steps in the aircraft assembly line are shown in fig. 9.

Firstly, according to the characteristics of an aircraft assembly production line, the assembly equipment and information acquisition requirements related in the production line are defined, and a multisource data acquisition system of the aircraft assembly production line is adopted to acquire and store relevant measured data. And (3) performing fusion processing on the actual measurement data of equipment such as assembly butt joint assembly, component quality detection and the like of the sensing and acquisition fuselage assembly by utilizing a multi-source data fusion technology, and generating a standard JSON data packet related to assembly execution equipment data and quality detection equipment data. And secondly, analyzing the prediction model according to the assembly data to obtain the processing progress data of each time period of the production line, and calculating the equipment utilization rate and the beat-productivity of the production line. And finally, performing visual display by adopting a bar graph and a line graph in the aspects of ECharts plug-in creation equipment utilization rate, production line beat-productivity and the like, so as to realize the production line operation monitoring application of a certain aircraft assembly production line.

The invention provides an assembly line multisource data sensing, fusion and predictive analysis method applied to monitoring of production line operation, in particular to a method for predicting equipment utilization rate and production line beat-productivity and visually displaying the equipment utilization rate and the production line beat-productivity through a histogram and a line diagram, wherein the assembly line multisource data sensing, fusion and predictive analysis method applied to monitoring of production line operation comprises the following steps:

Specifically, according to the characteristics of the aircraft assembly production line, the assembly equipment involved in the production line is defined, and comprises equipment such as a fuselage assembly butt joint device, an intersection point hole finish machining device, an automatic drilling and riveting device, a component quality intelligent detection unit, a water tightness intelligent test device and the like. From the angles of working conditions, quality and production line running state, the information acquisition requirements of the assembly site are analyzed, and actual measurement data such as execution data, processing progress and the like of all devices on the assembly production line under all time points are obtained and need to be perceived. The data comprise the starting and stopping state and the rotating speed of each shaft of equipment in the hole gantry finish machining system, the adjustment quantity and the gesture adjustment track of each motion shaft in the machine body assembly digital butt joint assembly system and the like, and information such as order task numbers, assembly part names, planned output, machining start time, planned completion time, actual machining time and the like.

Furthermore, by adopting a multisource data acquisition system of an aircraft assembly production line, the actual measurement data of the processing progress of equipment such as the assembly of the assembly butt joint of the airframe, the processing of the intersection point hole gantry machine, the quality detection of the parts and the like are acquired and stored. And cleaning and fusing the measured data by adopting a multisource data fusion technology of an aircraft assembly production line, generating JSON data packets of each execution device and each detection device, and organically fusing various data into a complete information set.

Furthermore, by combining with an aircraft assembly process and adopting an LSTM assembly data analysis and prediction model, the fused JSON data of each device is analyzed and predicted, and the assembly progress related data of each time period of the production line is accurately obtained. Meanwhile, a statistical process analysis (SPC) method is applied, a set of analysis mechanism of the utilization rate and the beat-productivity of the aircraft assembly production line is established, and the utilization rate of the equipment and the beat-productivity of the production line are accurately calculated through calculation of the number of completed procedures per hour and the output per hour and the working time.

Further, analyzing the equipment utilization rate and the beat-productivity of the production line in each time period, summarizing to form an evolution trend of the running state of the production line, creating a histogram and a line graph of the equipment utilization rate, the beat-productivity of the production line and the like by adopting ECharts plug-ins, and presenting analysis results in an intuitive mode to realize the running monitoring of the aircraft assembly production line.

Example 2:

In the present embodiment, an example of the application of the equipment status detection of a certain type of aircraft assembly line is given. The equipment state detection application is mainly used for predicting the running state and equipment faults of the aircraft assembly equipment and carrying out visual display through a visual monitoring platform. Taking the production line operation monitoring application of the model aircraft assembly production line as an example, the implementation steps of the multi-source data sensing, fusion and predictive analysis method of the assembly production line in the production line operation monitoring application are described.

The equipment status detection application steps in an aircraft assembly line are shown in fig. 10.

Firstly, analyzing the characteristics of an aircraft assembly process and an assembly process, determining the assembly equipment and information acquisition requirements related in a production line, acquiring the running state information of each equipment by means of a constructed data acquisition system, storing and utilizing a multi-source data fusion technology to perform fusion processing, and generating a standard JSON data packet related to assembly execution equipment data and quality detection equipment data. And secondly, constructing an equipment operation data analysis prediction model, determining that the model is input into equipment real-time operation parameters and equipment processing parameters, and outputting the model into equipment operation states including states of shutdown, standby, no-load, normal operation and the like. Finally, analyzing and predicting the fused equipment data, accurately obtaining the equipment operation state related information of each time period of the production line, and transmitting the information to a visual monitoring platform to show that the method for sensing, fusing and predicting the multi-source data of the assembly production line is applied to equipment state detection, particularly, predicting the operation state and equipment faults of the aircraft assembly equipment, and visually displaying the operation state and the equipment faults through the visual monitoring platform, wherein the method for sensing, fusing and predicting the multi-source data of the assembly production line is applied to the equipment state detection and comprises the following steps:

Specifically, the assembly equipment involved in the clear production line comprises equipment such as machine body assembly butt joint, intersection point hole finish machining, automatic drilling and riveting, intelligent detection units for component quality, intelligent water tightness test and the like. From the angles of working conditions, production line running states and prediction model construction, the information acquisition requirements of the assembly site are analyzed, and real-time running state information of all devices on the assembly production line under all time points is obtained. Such information includes information on the current state of the device, run time, program information, assembly stress, tooling coordinates, power, etc.

Furthermore, through the multisource data acquisition system of the aircraft assembly production line, real-time running state information of equipment such as the assembly of the assembly butt joint of the airframe, the gantry machining of the intersection point hole, the quality detection of the parts and the like is acquired and stored. And cleaning and fusing the measured data by adopting a multisource data fusion technology of an aircraft assembly production line, generating JSON data packets of each execution device and each detection device, and organically fusing various data into a complete information set.

Further, the characteristics of the aircraft assembly process and the assembly process are analyzed, an LSTM assembly data analysis prediction model is constructed, the model input is determined to be the real-time operation parameters of the equipment, the equipment processing parameters are determined, and the model output is the equipment operation state, including the states of shutdown, standby, no-load, normal operation and the like.

Further, the assembly data analysis and prediction model is adopted to analyze and predict the JSON data of each fused device, so that the relevant information of the running state of the device in each time period of the production line is accurately obtained, and the information is transmitted to the visual monitoring platform for display. And the subsequent common faults and solutions can be arranged through the detection result of the equipment state of the aircraft assembly production line, and a fault diagnosis knowledge base is constructed to assist maintenance personnel to rapidly process the faults, so that the processing efficiency is improved.

Example 3:

In this embodiment, an example of an assembly quality improvement application of an aircraft assembly line of a certain type is given. The core of the application of the assembly quality improvement is assembly quality prediction, and the assembly quality in the aircraft assembly process is predicted by utilizing an assembly quality prediction model, so that the quality improvement is realized by adopting a related strategy. Taking the assembly quality improvement application of the model aircraft assembly production line as an example, the implementation steps of the multi-source data sensing, fusion and predictive analysis method of the assembly production line in the production line operation monitoring application are described.

The assembly quality improvement application step in an aircraft assembly line is shown in fig. 11.

Firstly, according to the characteristic that the aircraft assembly quality data has time sequence, the aircraft assembly process is analyzed, meanwhile, the common assembly quality problem in the aircraft assembly process is analyzed, and an LSTM-based aircraft assembly process quality data analysis and prediction model is constructed. And secondly, determining an input layer and an output layer of a prediction model, acquiring related actual measurement data by using a multi-source data acquisition system of an aircraft assembly production line, and carrying out fusion processing on the perceived and acquired actual measurement data by using a multi-source data fusion technology to obtain a complete information set. And further, analyzing and acquiring a time sequence sample set under the execution time of each procedure according to a large amount of acquired data, and performing model training on the basis. Meanwhile, an assembly quality problem diagnosis knowledge base is established by combining the aircraft assembly quality data, expert experience and a big data analysis method. And finally, controlling the assembly process of the aircraft according to the output result of the quality prediction model and the production control decision of the assembly quality problem diagnosis knowledge base.

The invention provides an application of an assembly line multisource data sensing, fusing and predictive analysis method in assembly quality improvement, in particular to an application of the assembly line multisource data sensing, fusing and predictive analysis method in production line operation monitoring by predicting the assembly quality in the aircraft assembly process through an assembly quality prediction model and further adopting a related strategy to realize quality improvement, wherein the application comprises the following steps:

Specifically, according to the characteristics of time sequence and multi-stage assembly of the aircraft, the aircraft assembly process is analyzed, the related characteristics are summarized, meanwhile, the common assembly quality problems in the aircraft assembly process are analyzed, and an LSTM-based aircraft assembly process quality data analysis and prediction model is constructed. The input layer of the prediction model is actually measured data such as assembly time, assembly working procedure, assembly equipment information and the like, and the output layer of the prediction model is the quality problem of the aircraft assembly process and comprises assembly steps, assembly gaps, overlarge assembly stress values and the like among the assemblies.

Further, the related measured data obtained through analysis are collected by utilizing a multi-source data collection system of the aircraft assembly production line, and the sensed and obtained measured data are fused through a multi-source data fusion technology, so that a complete information set is obtained.

Further, based on a large amount of quality detection data sets acquired and fused in the aircraft assembly process, sample data of key quality characteristics are selected, and an Adam algorithm is utilized to find optimal parameters, so that the construction of a model is completed. And then, predicting through the sample data of the key quality characteristics, judging whether the quality meets the standard, determining the type of unqualified, and outputting a quality problem prediction result in the assembly process.

Further, according to the assembly quality prediction result, corresponding measures are adopted to improve and optimize the situation of abnormal assembly quality. When the quality problem in the aircraft assembly process is predicted, the influence degree of each element data on the assembly quality is analyzed, and quality problem data and various list data corresponding to the problem data are formed. And (3) combining aircraft assembly quality data, expert experience and a big data analysis method to establish an aircraft assembly quality problem diagnosis knowledge base which comprises specific reasons of quality problems, emergency solutions and targeted treatment measures. The quality control system for the aircraft assembly process takes the prediction result of the assembly quality prediction system and the diagnosis knowledge base of the assembly quality problem as inputs, clearly presents factors and treatment measures related to the problem, and controls the aircraft assembly process.

The method and the system for sensing, fusing and predicting the multi-source data of the assembly production line provided by the embodiment of the application are described in detail. The above description of embodiments is only for aiding in the understanding of the method of the present application and its core ideas; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present application, the present description should not be construed as limiting the present application in view of the above.

Certain terms are used throughout the description and claims to refer to particular components. Those of skill in the art will appreciate that a hardware manufacturer may refer to the same component by different names. The description and claims do not take the form of an element differentiated by name, but rather by functionality. As referred to throughout the specification and claims, the terms "comprising," including, "and" includes "are intended to be interpreted as" including/comprising, but not limited to. By "substantially" is meant that within an acceptable error range, a person skilled in the art is able to solve the technical problem within a certain error range, substantially achieving the technical effect. The description hereinafter sets forth a preferred embodiment for practicing the application, but is not intended to limit the scope of the application, as the description is given for the purpose of illustrating the general principles of the application. The scope of the application is defined by the appended claims.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a product or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such product or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a commodity or system comprising such elements.

It should be understood that the term "and/or" as used herein is merely one relationship describing the association of the associated objects, meaning that there may be three relationships, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.

While the foregoing description illustrates and describes the preferred embodiments of the present application, it is to be understood that the application is not limited to the forms disclosed herein, but is not to be construed as limited to other embodiments, and is capable of numerous other combinations, modifications and environments and is capable of changes or modifications within the scope of the inventive concept as expressed herein, either as a result of the foregoing teachings or as a result of the knowledge or technology of the relevant art. And that modifications and variations which do not depart from the spirit and scope of the application are intended to be within the scope of the appended claims.

Claims

1. The multi-source data sensing, fusion and predictive analysis method for the aircraft assembly production line is used for production line running state analysis, equipment use state monitoring and assembly quality improvement in the aircraft intelligent assembly process, and is characterized by comprising the following steps of:

s2: constructing a multi-source data acquisition system of an aircraft assembly production line by adopting an OPC production cascade technology according to data acquisition objects and information acquisition requirements;

S3: adopting a semantic description method based on an ontology to characterize entities in a production line multi-source data system;

s4: constructing a multi-source data ontology model of the aircraft assembly line through the entity characterization result;

S5: using JSON language to instantiate and describe a multisource data ontology model;

s6: constructing a long-term and short-term memory network model considering the time sequence characteristics of the assembly data according to the instantiation description result;

S7: designing an analysis prediction model training method based on measured multi-source data according to the long-term and short-term memory network model;

S8: designing a multisource assembly data analysis and prediction model, and training and optimizing according to a training method of the analysis and prediction model to obtain the multisource assembly data analysis and prediction model with time sequence characteristics.

2. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein the S1 specifically comprises:

S11: analyzing the process of an aircraft assembly production line, wherein the process comprises component assembly, body riveting and whole assembly, and the data acquisition object comprises assembly equipment use process monitoring data and quality control data on the production line;

S12: analyzing types of equipment execution equipment and quality detection equipment related in an aircraft assembly production line, defining attribute description of relevant data of the equipment, and determining equipment information required to be collected in the production line operation process according to production practice, wherein the equipment information required to be collected in the production line operation process comprises, but is not limited to: equipment running state information, actual values of production process parameters, alarm information and equipment working modes.

3. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein the step S2 specifically comprises:

s21: designing a data interface on assembly execution equipment and an assembly detection unit in the production line, and collecting data;

s22: aiming at the related data which are communicated and acquired by the equipment to be tested in the acquisition layer, the OPC protocol is adopted to transmit the data to the monitoring system, so that the data are transmitted among different equipment and systems;

S23: KEP SERVER EX is selected as a data acquisition OPC server, is connected with various devices, converts acquired data into standard OPC UA and/or OPC Client communication protocols, loads an OPC information model, and externally distributes the device information model and the data in an OPC UA protocol format.

4. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein the step S3 specifically comprises:

S32: analyzing object data and data attributes in the assembly process information, taking modeling of assembly process information as a main study object, defining object attributes and data attributes, and assisting in modeling of various description attributes of assembly tasks and assembly equipment and association relations of various concept classes;

the assembly equipment comprises assembly execution equipment and quality detection equipment, and the assembly process information comprises assembly tasks and assembly procedures.

5. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein S4 specifically comprises:

S41: creating a new ontology project in ontology modeling software, setting the identification of an aircraft assembly line multi-source data ontology model, creating concept classes and concept class hierarchical relations according to concept classes in the aircraft assembly line multi-source data ontology model defined in S3, adding examples under the concept classes, and filling specific objects under each class;

S42: and (3) defining object attributes in the multi-source data ontology model of the aircraft assembly production line and data attributes of assembly procedures in the multi-source data ontology model of the aircraft assembly production line, creating concept classes, object attributes and data attributes, logically reasoning the ontology by utilizing an reasoning mechanism of ontology modeling software, verifying that the logical structure of the ontology is correct, storing ontology items, and completing the construction of the ontology model.

6. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein S5 specifically comprises:

s51: analyzing a multisource data ontology model of an aircraft assembly line, wherein the multisource data ontology model comprises information description of entities, attributes and relations;

7. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein S6 specifically comprises:

S61: the method comprises the steps of defining the architecture of a long-term and short-term memory network model, comprising an input layer, an LSTM layer and an output layer, and determining the input layer and the output layer of the model architecture by considering the characteristic that the data in the assembly process has time sequence;

8. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein S7 specifically comprises:

S71: analyzing and acquiring a time sequence sample set under the execution time of each procedure from the acquired data, dividing the time sequence sample set into a training set and a testing set, and extracting different subsets from the training set, wherein each subset comprises a part of training set characteristics, and the characteristics in each subset can be repeated;

S72: selecting each subset by adopting L1 regularization features to obtain a plurality of feature subsets, solving the intersection of the feature subsets to obtain key influence factors in the finally selected input layer data, predicting the assembly quality problem output by a model, and converting the quality problem into a binary vector form which is easy to process by a computer according to the type of the single-heat encoding by adopting single-heat encoding for predicting model training;

S73: and optimizing parameters by using an Adam algorithm by taking the processed data as sample data, dividing the time sequence sample data into a training set and a testing set, and training a data analysis and prediction model of the constructed production line, wherein if the loss function of the model is rapidly reduced and tends to be converged, and meanwhile, the data change trend of the testing set and the data change trend of the training set are the same and/or similar, the LSTM prediction model is judged to be reasonably constructed.

9. The multi-source data sensing, fusion and predictive analysis method according to claim 1, wherein S8 specifically comprises:

S81: collecting aircraft assembly-related data including, but not limited to, assembly time, assembly procedures, and assembly equipment information, based on the aircraft assembly process data having a time-sequential nature;

S82: dividing data into a plurality of sequence samples according to the size of a time window, wherein each sample contains continuous data in a period of time, and constructing an aircraft assembly process data analysis prediction model based on a long-period memory network;

S83: the method comprises the steps of determining input and output of a prediction model, wherein an input layer is actual measurement data of assembly time, assembly working procedures and assembly equipment information, an output layer is a performance index prediction result of assembly qualification rate, working procedure execution rate and equipment utilization rate of current assembly production line operation, and training and optimizing the model by adopting a network training method to realize construction of an LSTM-based multi-source data analysis prediction model.

10. The multi-source data sensing, fusing and predictive analysis system of the assembly production line is characterized in that the data sensing, fusing and predictive analysis method comprises the following steps: