Intelligent gateway fault diagnosis method and system based on transfer learning

Abstract

The invention relates to the technical field of intelligent gateways of the Internet of Things, in particular to an intelligent gateway fault diagnosis method and system based on transfer learning, and the method comprises the steps: completing the data collection and format conversion of a multi-rate multi-protocol device through the implementation of a first-stage data collection stage, and achieving the real-time unified processing of a communication protocol; in the data analysis stage, structured analysis is carried out on heterogeneous data, and a modular system with bidirectional migration capability is constructed; in the intelligent diagnosis stage, a fault feature migration model is established, the problem of structural inconsistency of multi-source heterogeneous data is solved through a data feature alignment technology, and a real-time diagnosis algorithm supporting a multi-rate sensing channel is developed particularly for rotary mechanical devices such as a water turbine and a wind power gear set. The transfer learning technology is innovatively embedded into an intelligent gateway architecture, and the limitation of a traditional single-machine diagnosis mode is broken through, so that the equipment synchronously completes feature learning, model transfer and online diagnosis in a data transmission process, and parallel processing of triple functions of transmission, learning and diagnosis is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things intelligent gateways, and particularly to an intelligent gateway fault diagnosis method and system based on transfer learning. Background Art

[0002] Manufacturing is the main body of the national economy. Currently, the global manufacturing industry is accelerating towards the digital and intelligent era, and intelligent manufacturing has an increasingly greater impact on the competitiveness of the manufacturing industry. In the intelligent manufacturing system, equipment is the core and guarantee of manufacturing production.

[0003] During the data acquisition process, in the actual equipment operation status monitoring, the amount of early minor fault samples collected is small, and the sample quality is low due to multi-rate sampling of sensors, random packet loss, diversity of sensor types, and different storage forms of the collected information. The small number and low quality of fault samples will inevitably affect the accuracy of feature extraction of the fault diagnosis model based on deep learning, and cannot guarantee the effectiveness of the fault diagnosis method based on deep learning. Therefore, a more effective processing method is urgently needed. Summary of the Invention

[0004] In view of the problems existing in the above-mentioned prior art, the present invention is proposed.

[0005] Therefore, the technical problem to be solved by the present invention is: aiming at the actual needs of today's industry and manufacturing, the present invention proposes an intelligent gateway architecture with transfer learning fault diagnosis function equipped with a real-time operating system. This intelligent gateway can achieve unified communication format and transfer learning in a multi-rate and multi-protocol environment.

[0006] To solve the above technical problems, the present invention provides the following technical solutions. The intelligent gateway fault diagnosis method based on transfer learning includes: primary data acquisition and format conversion; parsing and processing of the primary data; constructing a primary module to achieve bidirectional transfer; and realizing real-time fault diagnosis.

[0007] As a preferred solution of the intelligent gateway fault diagnosis method based on transfer learning according to the present invention, wherein: the format conversion includes converting the collected data into a digital signal recognizable by a computer, and being able to display and record at the sending end.

[0008] As a preferred solution of the intelligent gateway fault diagnosis method based on transfer learning according to the present invention, wherein: the parsing and processing includes protocol conversion of the collected data, and parsing and processing by using the upper-layer interface and lower-layer protocol parsing, and converting it into a unified format for transmission.

[0009] As a preferred solution of the intelligent gateway fault diagnosis method based on transfer learning according to the present invention, wherein: the realization of bidirectional transfer includes forward transfer and reverse transfer;

[0010] Among them, forward migration includes migrating from an incomplete structure to a complete model;

[0011] Reverse migration includes migrating from a complete model to an incomplete structure.

[0012] As a preferred solution of the intelligent gateway fault diagnosis method based on transfer learning according to the present invention, wherein: the fault diagnosis further constructs and implements fault diagnosis according to bidirectional transfer learning, including implementing fault detection in the online stage and fault detection in the offline stage.

[0013] As a preferred solution of the intelligent gateway fault diagnosis method based on transfer learning according to the present invention, wherein: the primary data includes multi-protocol data and multi-rate data, and the gateway serial port uses a communication module to implement data communication between the network layer and the sensing device, and collects data of the sensing device in real time through the established communication connection.

[0014] As a preferred solution of the intelligent gateway fault diagnosis method based on transfer learning according to the present invention, wherein: the primary module includes a learning and diagnosis module, which uses transfer learning to perform bidirectional migration of complete and incomplete data collected at multiple rates, thereby realizing real-time fault diagnosis.

[0015] Another object of the present invention is to provide an intelligent gateway fault diagnosis system based on transfer learning, which realizes collaborative diagnosis of incomplete samples and complete samples in a multi-rate sampling scenario by constructing a bidirectional dynamic parameter transfer mechanism, and uses edge computing capabilities to synchronously complete data feature extraction, online model optimization, and fault type discrimination, so as to break through the dependence of traditional diagnosis methods on data integrity and consistency and effectively improve the diagnosis accuracy of key equipment such as hydroturbines / wind turbine gearboxes in the early stage of minor faults on the premise of ensuring industrial-level real-time performance.

[0016] To solve the above technical problems, the present invention provides the following technical solution: an intelligent gateway fault diagnosis system based on transfer learning, including: a data acquisition module, a protocol conversion module, and a learning and diagnosis module;

[0017] The data acquisition module converts the signal output by the sensor into a digital signal recognizable by a computer, sends it to the receiving end for recording, and monitors physical quantities;

[0018] The protocol conversion module consists of an upper-layer interface and a lower-layer protocol parser, which parses and processes data of different protocols and converts them into a unified format for transmission;

[0019] The learning and diagnosis module uses transfer learning to perform bidirectional migration of complete and incomplete data collected at multiple rates, thereby constructing a real-time fault diagnosis system and realizing online fault diagnosis of samples with inconsistent structures.

[0020] A computer device includes a memory and a processor. The memory stores a computer program. It is characterized in that when the processor executes the computer program, the steps of the intelligent gateway fault diagnosis method based on transfer learning as described above are implemented.

[0021] A computer-readable storage medium stores a computer program thereon. It is characterized in that when the computer program is executed by a processor, the steps of the intelligent gateway fault diagnosis method based on transfer learning as described above are implemented.

[0022] Advantages of the present invention: The intelligent gateway can achieve unified communication format and transfer learning in multi-rate and multi-protocol environments; data collection and unified protocol, converting the data uploaded by devices with different sampling rates and different communication protocols into a unified protocol and transmitting it to the upper layer; for fault detection of gear sets such as hydroturbines / wind turbines, data learning and diagnosis are carried out, aligning different structured data by using the inconsistency of the data structures collected by the gateway, establishing the transfer between deep learning models, and constructing a real-time fault diagnosis system under multi-rate sampling, enabling the intelligent gateway to have the ability to transmit, learn, and diagnose simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 It is a flowchart of the intelligent gateway fault diagnosis method based on transfer learning provided by an embodiment of the present invention.

[0025] Figure 2 It is a gateway hardware structure diagram of the intelligent gateway fault diagnosis method based on transfer learning provided by an embodiment of the present invention.

[0026] Figure 3 It is a gateway software architecture diagram of the intelligent gateway fault diagnosis method based on transfer learning provided by an embodiment of the present invention.

[0027] Figure 4 It is a schematic diagram of the two-way transfer mechanism of the intelligent gateway fault diagnosis method based on transfer learning provided by an embodiment of the present invention.

[0028] Figure 5 It is a fault diagnosis flowchart of the intelligent gateway fault diagnosis method based on transfer learning provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0030] Example 1, referring to Figure 1 - Figure 2 , which is an embodiment of the present invention. This embodiment provides an intelligent gateway fault diagnosis method based on transfer learning, including:

[0031] S1: Primary data acquisition and format conversion.

[0032] It should be noted that, as Figure 1 shown in S1, the data uploaded by devices with different sampling rates and different communication protocols is converted into a unified protocol and transmitted to the upper layer.

[0033] Furthermore, in the hardware architecture of the intelligent gateway, the primary data includes multi-protocol data and multi-rate data. The gateway serial port uses a communication module to realize data communication between the network layer and the sensing devices, and collects the data of the sensing devices in real time through the established communication connection;

[0034] Even further, the hardware architecture of the intelligent gateway includes a storage module, a power module, a peripheral circuit module, a communication module, and a human-machine interface; among them, the storage module includes the functions of program storage and data storage, including an Ethernet port, a serial port, and a USB port. The Ethernet port is mainly responsible for the transmission connection between the gateway and the upper-layer server. The serial port is mainly responsible for the connection of data collection from on-site devices and instruments. The USB port is mainly responsible for gateway parameter configuration and version upgrade download; the human-machine interface realizes functions such as information interaction between humans and the gateway and related parameter settings.

[0035] In the embodiment of the present application, the primary data is heterogeneous data collected through a multi-protocol communication module controlled by the MediaTek MT76828AN main control chip in the hardware architecture of the intelligent gateway. Among them, the Ethernet port (HR91105A chip) receives Modbus TCP / Profinet protocol data, the serial port (SP3232EEN+MAX13487EESA+ chipset) collects RS485 / RS232 protocol data, and the USB interface (GL850G chip) obtains local configuration parameters; the original timing data of multi-rate sampling includes high-frequency vibration signals (50 - 1000Hz) and low-frequency temperature / pressure signals (1 - 10Hz) of hydraulic turbines / wind power gearboxes.

[0036] In an alternative embodiment, the primary data can also be implemented in other ways. For example, a wireless communication module can be used to replace some of the wired interfaces for data acquisition. Specifically, the gateway can integrate a Wi-Fi 4 (802.11n) communication module to replace some of the serial communication functions and directly receive sensor data that supports wireless protocols in AP mode. The RS485 interface of the field device can be wirelessly transformed by connecting an external wireless transparent transmission module (such as a Zigbee-to-serial device), and the gateway receives data through the 2.4GHz band.

[0037] In an alternative embodiment, the primary data can also be implemented in other ways. For example, a low-cost general-purpose main control chip can be used to build the hardware architecture. The gateway can use an STM32F407 series ARM Cortex-M4 chip to replace the MT76828AN in the document and connect an external independent PHY chip (such as DP83848) to achieve Ethernet communication. Data acquisition communicates with the device by simulating the serial port protocol through GPIO (such as the UART half-duplex mode), reducing the hardware cost. The power module can use a common LDO linear voltage regulator (such as AMS1117), and the human-machine interface is based on LED status indicators. The physical interaction interface is replaced by a mobile phone APP, and parameter configuration depends on an external intelligent device.

[0038] In the embodiment of the present application, the hardware architecture of the intelligent gateway is to build a multi-module hardware system, deploy a communication interface module, and expand corresponding functional modules and interfaces through the MT76828AN SOC of MediaTek. The solution adopted by the gateway is that a main control chip with a MIPS architecture controls the network port and serial port to collect data of industrial field devices, and then realizes multiple network access methods through external modules and conducts data communication with the cloud server in the application layer. Specifically, such as Figure 2As shown, the gateway network interface is the connector HR91105A with RJ45. Since the main controller MT7628 already integrates a wired network port, no additional chip support is required. HR91105A fully complies with the IEEE802.3 standard. Its physical layer protocol interface fully supports the use of Category 3, 4, and 5 unshielded twisted pairs at 10MBps and Category 5 unshielded twisted pair at 100MBps, and can adaptively and automatically coordinate the configuration to maximize the network bandwidth; the gateway power supply module provides a stable working voltage for each hardware module of the system, mainly including the 5V power supply and the 3.3V power supply obtained by the MP1482 chip; the gateway storage module consists of a FLASH memory and an SDRA memory; the gateway serial communication module uses the SP3232EEN and MAX13487EESA+ control chips to realize data communication between the network layer and the sensing device, and collects the data of the sensing device in real time through the established communication connection; the gateway USB module uses the GL850G chip. GL850G is an advanced version of the GenesysLogic hub solution and fully complies with the Universal Serial Bus specification; GL850G has proven compatibility, lower power consumption, and a better cost structure, and is widely used.

[0039] In an alternative embodiment, the hardware architecture of the intelligent gateway can also be implemented in other ways. For example, an industrial-grade main control chip based on the ARM Cortex-A7 architecture (such as NXP i.MX6ULL) is used; a dual-port Ethernet MAC controller is integrated, and an external PHY chip (such as Realtek RTL8201CP) is required to achieve 10 / 100M adaptive Ethernet communication. An additional circuit area of 22mm×14mm is required, and the coordination timing between the PHY chip and the main control chip needs to be compensated by software; the storage module can use an eMMC5.1 chip (such as Kingston EMMC04G-M627) to simplify the storage structure; the serial communication module can select the MAX3232CSE chip; the USB module is implemented using the FTDI FT231XS-R chip, and an external 12MHz crystal oscillator is required and only supports the USB2.0 full-speed mode (12Mbps).

[0040] In an optional embodiment, the hardware architecture of the intelligent gateway can also be implemented in other ways. A solution can also be adopted that uses a RISC-V open-source architecture master chip (such as GD32VF103CBT6) in cooperation with an externally extended Ethernet controller (such as DM9000AEP). It is necessary to connect the DM9000AEP chip through the FSMC bus to achieve 10M Ethernet communication, and the MAC address needs to be stored in an external EEPROM. The power module can adopt a scheme of an AP2112K-3.3V linear voltage regulator combined with an AMS1117-5.0V. The storage module is changed to a single-chip solution of a NOR Flash with an SPI interface (such as W25Q128JVSIQ). The serial communication module uses an ADM3202ARNZ chip. The USB interface adopts a CP2102-GM bridge chip, which depends on the Windows system driver for configuration, restricting the cross-platform application ability.

[0041] Embodiment 2, refer to Figure 3 - Figure 4 , which is an embodiment of the present invention. This embodiment provides an intelligent gateway fault diagnosis method based on transfer learning, including:

[0042] S2: Parse and process the primary data.

[0043] As Figure 1 shown in S2, the parsing and processing include protocol conversion of the collected data, and parsing and processing are performed using the upper-layer interface and lower-layer protocol parsing, and then converted into a unified format for transmission.

[0044] S3: Construct a primary module to achieve bidirectional transfer.

[0045] It should be noted that, as Figure 1 shown in S3, achieving bidirectional transfer includes forward transfer and reverse transfer;

[0046] Among them, the forward transfer includes transferring from an incomplete structure to a complete structure model; the reverse transfer includes transferring from a complete structure model to an incomplete structure model; the primary module includes a learning and diagnosis module, which uses transfer learning to perform bidirectional transfer on the complete and incomplete data collected at multiple rates, thereby realizing real-time fault diagnosis.

[0047] Furthermore, as Figure 3 shown, transferring from an incomplete DNN model to a complete DNN model includes classifying samples according to the missing values of the samples, and respectively establishing and training a fault diagnosis model DNN for incomplete structure samples for each type of missing data s , establishing a fault diagnosis model DNN for complete structure samples c , realizing the fault diagnosis model DNN of incomplete structure samples s to the fault diagnosis model DNN of complete structure samples cMigration and training and optimizing the model DNN with a small number of complete-structured samples c ;

[0048] Classify the samples according to the missing values in the samples, and divide the sample set into a complete-structured sample set X c and an incomplete-structured sample set X s , and according to the missing values, divide X s specifically into n categories; X c represents that the samples in this sample set simultaneously contain the values of several sensors, that is, complete-structured samples; X si represents that the samples in this sample set contain the values from sensor 1 to sensor n-i+1, where i = 1, 2,..., n; assume that the complete-structured samples contain P variables. If the value of the p-th variable of the sample x m is missing, then C pm (x m ) is 0, otherwise it is 1; C m (x m ) represents the missing status of the sample x m . When C m (x m ) = 1, it means that the sample x m is a complete-structured sample; when , it means that the sample x m is an incomplete-structured sample, that is, missing data; when C pi (x i ) = C pi (x j ), is true, then the sample x i and the sample x j belong to the same category of samples, that is, they both belong to the complete sample data set or a certain category of missing sample data sets;

[0049] Establish and train a fault diagnosis model DNN for incomplete-structured samples for each type of missing data s It is expressed as:

[0050] DNN s = Feedforwwad(h s1 ,h s2 ,…,h sN ) ⑴

[0051]

[0052] As shown in formula ⑴, taking the i-th type of missing data as an example, establish a fault diagnosis model DNN si , which is composed of N stacked autoencoders, h m1 ,h m2 ……h mNThey are the number of neurons in the first hidden layer, the second hidden layer, and the Nth hidden layer of the DNN respectively; si

[0053] As shown in formula (2), train the DNN network model of typical significant faults, where X s is the training sample set of typical significant faults; θ s =[{θ s ,θ s1 ,…,θ s2}] represents the set of initial network parameters encoded by N autoencoders, where θ sN =[{W sk ,b sk}] represents the set of parameter matrices of the weight and bias between the input layer and the hidden layer of the kth autoencoder AE sk in the DNN s , and the parameters will be randomly initialized; k represents the set of initial parameters of the decoding network of N autoencoders, represents the set of parameter matrices of the weight and bias between the hidden layer and the output layer of the kth autoencoder AE s in the DNN k , and these parameters will also be randomly initialized; Train and update the DNN model parameters layer by layer to obtain the updated encoded network parameters θ' s and the decoding network parameters s Obtain the abstract features, expressed as: H =[σ(W' sN …σ(W' sN (σ(W' s2 X s1 +b′ s )+b′ s1 )+…+b′ s2 )]; sN

[0054] Among them, H sN is passed as the input to the Softmax classifier to update and obtain the Softmax parameter θ' ss ;

[0055] Establish a fault diagnosis model DNN for structurally complete samples, expressed as: c

[0056] DNN i =[Feesforwad(h i1 ,h i2 ,…,h iN ) ⑶

[0057] ​​​​

[0058] As shown in formulas ⑶ and ⑷, a deep neural network model DNN for atypical faults is established. i , this model is similar to DNN s The network structure is the same, and the number of neurons in each layer is also the same;

[0059] Implementing a DNN fault diagnosis model for incomplete sample structures s Fault Diagnosis Model DNN for Structured Complete Samples c The migration includes: Since the dimensions of the structurally complete samples and the structurally incomplete samples are different, when migrating the fault diagnosis model based on the incomplete structure to the fault diagnosis model based on the complete structure, the network structure of the deep neural network needs to be modified. The migration process is as follows: Figure 3 As shown;

[0060] Using a small number of structurally complete samples to train and optimize the model DNN c It is expressed as:

[0061]

[0062] As shown in formula ⑸, the training model DNN ci For example, take a complete sample dataset X with only a few samples c As the training set, the fault diagnosis model DNN of the incomplete sample structure is implemented s Fault Diagnosis Model DNN for Structured Complete Samples c The network parameters θ obtained by migration i and As initial parameters, train the atypical fault diagnosis model DNN i , get the updated encoding network parameters θ' i and decode network parameters Get the abstract feature parameters, expressed as: H iN =σ(W' iN …σ(W' i2 (σ(W' i1 X i +b' i1 )+b' i2 ))+…+b' iN );H iN As input data, the structured incomplete sample fault diagnosis model DNN is implemented in steps s Fault Diagnosis Model DNN for Structured Complete Samples c The migration of θ is As the initial parameters of the atypical fault Softmax classifier, train the atypical fault Softmax classifier, update and obtain the Softmax parameter θ' is ; Use supervised back propagation algorithm to train DNNi Parameter Fine-tune and optimize to obtain the parameters of the micro-fault diagnosis model

[0063] Furthermore, migrating from a structurally complete DNN model to a structurally complete DNN model includes migrating from a structurally complete fault diagnosis model to a structurally incomplete fault diagnosis model and training and optimizing the model DNN with structurally incomplete samples s ;

[0064] The migration from a structurally complete fault diagnosis model to a structurally incomplete fault diagnosis model is based on DNN ci Migrate to DNN si Taking... as an example, the migration of the first autoencoder encoding network parameters is expressed as:

[0065]

[0066] The migration of the first autoencoder decoding network parameters is expressed as:

[0067]

[0068] The migration of the remaining autoencoder encoding network parameters is expressed as:

[0069]

[0070] The migration of the remaining autoencoder decoding network parameters is expressed as:

[0071]

[0072] The migration of the Softmax layer parameters is expressed as:

[0073] θ SS =θ″ cS ⑽

[0074] As shown in formula ⑻, assign the encoding network parameters θ″ of the second to the Nth layer of the autoencoders of the DNN trained through forward migration ci to the encoding network parameters θ of the corresponding layers of DNN c2 ,……θ″ cN in sequence si ,……θ m2 ,……θ mN ,DNN si The initial parameter set θ of the encoding network si ={θ m1 ,θ m2 ,…,θ mN};

[0075] As shown in formula ⑼, the DNN trained through forward migrationci The encoding network parameters of the autoencoders from the second layer to the Nth layer Are sequentially assigned to the DNN si The encoding network parameters of the corresponding layer Set of initial encoding network parameters

[0076] As shown in Equation (10), through the transfer parameter θ' c ' S Obtain the initial parameter θ of the Softmax classifier for the structurally incomplete samples SS , Similarly, transfer the DNN ci Respectively to the DNN s1 , DNN s2 , …… DNN sN Model;

[0077] Train and optimize the model DNN with structurally incomplete samples s It is expressed as:

[0078]

[0079] The network parameters obtained by migrating from the structurally complete fault diagnosis model to the structurally incomplete fault diagnosis model are used as initial parameters to train and optimize multiple structurally incomplete DNNs s1 , DNN s2 ,... DNN sn Model, train the model DNN si , The training and optimization process is the same as that in the forward transfer, where a fault diagnosis model DNN for structurally incomplete samples is established and trained separately for each type of missing data s ; In this way, DNN si And DNN ci Are alternately trained, optimized, and migrated until a fault diagnosis model with the highest accuracy is reached, and finally the parameters of the structurally complete fault diagnosis model and the set of parameters of the structurally incomplete fault diagnosis model T c And the set of parameters of the structurally incomplete fault diagnosis model T s ={T s1 , T s2 , ……, T sN}, T si Represents the parameters of the structurally incomplete model DNN si .

[0080] S4: Implement real-time fault diagnosis.

[0081] It should be noted that, as Figure 1As shown in S4, fault diagnosis is further constructed based on bidirectional transfer learning to achieve fault diagnosis, including fault detection in the online stage and fault detection in the offline stage.

[0082] Furthermore, during the data acquisition process, aiming at the problem of inconsistent data sample structures at different times under multi-rate sampling, a fault diagnosis method based on transfer learning under multi-rate sampling is proposed. The purpose is to establish a bidirectional transfer mechanism from part to global and from global to part by making full use of data samples with inconsistent structures, and to achieve innovation at the level of real-time improvement transfer mechanism for fault diagnosis.

[0083] Even further, since the samples with incomplete structures lack the values of different sensors and cannot be used for fault diagnosis with the same deep neural network; in the offline stage, corresponding fault diagnosis models DNN s1 , DNN s2 , …… DNN sn and DNN c The flowchart of fault diagnosis in the online diagnosis stage is as Figure 4 shown.

[0084] The above is a schematic solution of a fault diagnosis method for an intelligent gateway based on transfer learning in this embodiment. It should be noted that the technical solution of the system of the fault diagnosis method for the intelligent gateway based on transfer learning belongs to the same concept as the above technical solution of the fault diagnosis method for the intelligent gateway based on transfer learning. For the details not described in detail in the technical solution of the fault diagnosis system for the intelligent gateway based on transfer learning in this embodiment, reference can be made to the description of the technical solution of the fault diagnosis method for the intelligent gateway based on transfer learning.

[0085] Example 3, referring to Figure 5 , is an embodiment of the present invention, which provides a fault diagnosis system for an intelligent gateway based on transfer learning, including: a data acquisition module, a protocol conversion module, and a learning and diagnosis module;

[0086] The data acquisition module converts the signals output by the sensors into digital signals recognizable by the computer, sends them to the receiving end for recording, and monitors the physical quantities;

[0087] The protocol conversion module consists of two parts: an upper-layer interface and a lower-layer protocol parsing, and parses and processes data with different protocols and converts them into a unified format for transmission;

[0088] The learning and diagnosis module uses transfer learning to perform bidirectional transfer on the complete and incomplete data collected at multiple rates, thereby constructing a real-time fault diagnosis system and realizing online fault diagnosis of samples with inconsistent structures.

[0089] Specifically, the system also consists of gateway software and cloud platform software. Among them, the gateway is based on an embedded real-time operating system and mainly consists of a data acquisition module, a protocol conversion module, and a learning and diagnosis module, mainly realizing functions such as system data acquisition, transmission, transfer learning, and fault diagnosis; the cloud platform selects a commonly used free platform, which is simple to deploy and convenient to use, mainly realizing functions such as data display, storage, monitoring, and training; among them, the embedded real-time operating system gateway adopts the SylixOS system of Yihui, and SylixOS provides a complete development platform integrating design, development, debugging, simulation, deployment, and testing, facilitating system development and debugging.

[0090] This embodiment also provides a computing device applicable to the situation of the intelligent gateway fault diagnosis method based on transfer learning, including:

[0091] A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent gateway fault diagnosis method based on transfer learning as proposed in the above embodiment.

[0092] This embodiment also provides a storage medium on which a computer program is stored, and when the program is executed by a processor, it implements the intelligent gateway fault diagnosis method based on transfer learning as proposed in the above embodiment.

[0093] The storage medium proposed in this embodiment and the intelligent gateway fault diagnosis method based on transfer learning proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0094] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the essence of the technical solution of the present invention, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs and other various media that can store program codes.

[0095] Logic and / or steps described otherwise herein, for example, can be considered as a definite ordered list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0096] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), and the like.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An intelligent gateway fault diagnosis method based on transfer learning, characterized in that: Including: Primary data acquisition and format conversion; Parsing and processing the primary data; Constructing a primary module to achieve two-way migration; Implementing real-time fault diagnosis.

2. The intelligent gateway fault diagnosis method based on transfer learning according to claim 1, wherein: The format conversion includes converting the acquired data into digital signals recognizable by a computer, and being able to display and record at the sending end.

3. The intelligent gateway fault diagnosis method based on transfer learning according to claim 2, wherein: The parsing and processing includes performing protocol conversion on the acquired data, and parsing and processing using the upper-layer interface and lower-layer protocol parsing, and converting it into a unified format for transmission.

4. The intelligent gateway fault diagnosis method based on transfer learning according to claim 3, wherein: The implementation of two-way migration includes forward migration and reverse migration; Among them, forward migration includes migrating from an incomplete structure to a complete model; Reverse migration includes migrating from a complete model to an incomplete model.

5. The intelligent gateway fault diagnosis method based on transfer learning according to claim 4, wherein: The fault diagnosis is further constructed based on two-way transfer learning to achieve fault diagnosis, including fault detection in the online stage and fault detection in the offline stage.

6. The intelligent gateway fault diagnosis method based on transfer learning according to claim 5, characterized in that: The primary data includes multi-protocol data and multi-rate data. The gateway serial port uses a communication module to realize data communication between the network layer and the sensing device, and collects the data of the sensing device in real time through the established communication connection.

7. The intelligent gateway fault diagnosis method based on transfer learning according to claim 6, characterized in that: The primary module includes a learning and diagnosis module, which uses transfer learning to perform two-way migration on the complete and incomplete data collected at multiple rates, and then realizes real-time fault diagnosis.

8. A system for an intelligent gateway fault diagnosis method based on transfer learning according to any one of claims 1-7, characterized in that: Including: A data acquisition module, a protocol conversion module, and a learning and diagnosis module; The data acquisition module converts the signal output by the sensor into a digital signal recognizable by a computer, sends it to the receiving end for recording, and monitors the physical quantity; The protocol conversion module consists of two parts: upper-layer interface and lower-layer protocol parsing, and parses and processes data with different protocols and converts it into a unified format for transmission; The learning and diagnosis module uses transfer learning to perform two-way migration on the complete and incomplete data collected at multiple rates, thereby constructing a real-time fault diagnosis system and realizing online fault diagnosis of samples with inconsistent structures.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for fault diagnosis of an intelligent gateway based on transfer learning according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for fault diagnosis of an intelligent gateway based on transfer learning according to any one of claims 1 to 7.

Images