CN112164153A - AI edge calculation fault diagnosis device - Google Patents

AI edge calculation fault diagnosis device Download PDF

Info

Publication number
CN112164153A
CN112164153A CN202011002985.3A CN202011002985A CN112164153A CN 112164153 A CN112164153 A CN 112164153A CN 202011002985 A CN202011002985 A CN 202011002985A CN 112164153 A CN112164153 A CN 112164153A
Authority
CN
China
Prior art keywords
module
interface
data
neural network
fault
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011002985.3A
Other languages
Chinese (zh)
Inventor
彭炜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiamen Dewei Zhi Lian Technology Co ltd
Original Assignee
Xiamen Dewei Zhi Lian Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xiamen Dewei Zhi Lian Technology Co ltd filed Critical Xiamen Dewei Zhi Lian Technology Co ltd
Priority to CN202011002985.3A priority Critical patent/CN112164153A/en
Publication of CN112164153A publication Critical patent/CN112164153A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G07—CHECKING-DEVICES
    • G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
    • G07C3/00—Registering or indicating the condition or the working of machines or other apparatus, other than vehicles
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/18—Multiprotocol handlers, e.g. single devices capable of handling multiple protocols
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/22—Parsing or analysis of headers

Abstract

The invention discloses an AI edge calculation fault diagnostor, which comprises: the system comprises a platform architecture, a software system, a multi-protocol support module, an AI support module, a supervision learning module and a differentiation adaptation module; the multi-protocol support module is used for downwards butting the field equipment and acquiring operation data, upwards butting the upper computer to receive a control instruction or transmit data, matching a communication protocol of the field equipment and analyzing the data; the AI supporting module is used for preprocessing the acquired data; the supervised learning module is used for carrying out neural network model training on line to obtain a fault diagnosis model and deploying the trained fault diagnosis model for fault diagnosis; the differentiated adaptation module is used for saving and configuring parameters of the supervised learning module so as to adapt to the field device. The invention can effectively diagnose the unexpected mode of the equipment by combining the edge calculation and the artificial intelligence; and the privacy, the rapidity and the safety of data transmission are ensured by adopting localized deployment.

Description

AI edge calculation fault diagnosis device
Technical Field
The invention relates to the field of intelligent manufacturing, in particular to an AI edge calculation fault diagnoser.
Background
With the continuous development of various industries, the demand of mechanical equipment is increasing, and in order to ensure the normal operation of the mechanical equipment, a diagnostic device needs to perform fault diagnosis on the mechanical equipment in real time or periodically so as to maintain the mechanical equipment in time, however, most of the existing diagnostic devices cannot diagnose the possible unexpected behavior patterns of the equipment and cannot describe the industrial process by using an accurate mathematical model, or the fault diagnosis calculation is delivered to a cloud or a remote server to be completed, so that the problems of instantaneity and unsafe private data are caused.
Disclosure of Invention
In view of the above-mentioned drawbacks of the prior art, it is an object of the present invention to provide a fault diagnosis apparatus that can effectively diagnose an unexpected mode and is suitable for localized deployment.
To achieve the above object, the present invention provides an AI-edge calculation fault diagnoser, including:
the system comprises a platform architecture, a software system, a multi-protocol support module, an AI support module, a supervision learning module and a differentiation adaptation module;
the platform architecture and the software system are used for constructing an edge computing platform;
the multi-protocol support module comprises at least one bus interface or Ethernet interface, at least one protocol packet and at least one control module, wherein the at least one bus interface or Ethernet interface is used for downwards butting the field equipment and acquiring the operating data of the field equipment;
the AI supporting module is used for preprocessing the acquired data;
the supervised learning module is used for carrying out neural network model training through the preprocessed data acquired on line to acquire a fault diagnosis model; deploying a trained fault diagnosis model, and performing fault diagnosis according to the preprocessed data acquired on line;
the differentiation adapting module is used for saving and configuring parameters of the supervision learning module so as to adapt to field equipment.
The AI edge calculation fault diagnosis device is an independent device and is arranged on the side of field equipment, the operation parameters and the operation state of one or more pieces of equipment are monitored in real time by using the mode of combining edge calculation and artificial intelligence, and online fault diagnosis model training and fault diagnosis are carried out, so that the problems of real-time performance of fault diagnosis, unsafe private data and the like are solved. Meanwhile, the difference between the equipment and the automatic adjustment equipment is made up through a differentiation adaptation module, so that a perfect training effect is achieved.
Further, the platform architecture is an ARM platform architecture or an X86 platform architecture.
Further, the software system is a Linux operating system.
Further, the protocol packet includes multiple protocols among an equipment networking protocol, an industrial field networking protocol, and an application layer networking protocol.
Further, the bus interface at least comprises one of an RS-485 interface and a CAN interface.
Further, the multi-protocol support module at least comprises two Ethernet interfaces, an RS-485 interface and a CAN interface.
The system further comprises a multi-sensor support module used for directly inputting data collected by the sensors, and the interface types of the multi-sensor support module comprise a digital quantity input and output interface, an analog quantity input interface, an RS-232 serial interface and an RS-422 serial interface.
Further, the AI support module includes an AI hardware acceleration module and an algorithm library, and/or includes an AI software library.
Further, the neural network model deployed by the supervised learning module includes but is not limited to: BP neural network, quantum neural network of multilayer excitation function and RBF neural network. Different technical problems are solved by deploying different neural network models in the supervised learning module.
The technical effects are as follows:
the AI edge calculation fault diagnotor of the invention analyzes and infers real-time data by collecting real-time data such as operation parameters, operation states and the like of field equipment and sensors in a mode of combining edge calculation and artificial intelligence, applies a bottom layer numerical model of a neural network, takes the neural network as a representation and processing mode of a knowledge source and fuses with other inference mechanisms, realizes multi-mode inference and can effectively diagnose an unexpected mode of equipment; and the privacy, the rapidity and the safety of data transmission are ensured by adopting localized deployment.
Drawings
FIG. 1 is a system block diagram of an AI edge calculation fault diagnoser of the present invention;
FIG. 2 is a schematic interface diagram of an AI edge calculation fault diagnoser of the present invention;
FIG. 3 is a schematic diagram of the connections of a typical neural network model;
fig. 4 is an application block diagram of the BP neural network in fault diagnosis.
Detailed Description
To further illustrate the various embodiments, the invention provides the accompanying drawings. The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the embodiments. Those skilled in the art will appreciate still other possible embodiments and advantages of the present invention with reference to these figures. Elements in the figures are not drawn to scale and like reference numerals are generally used to indicate like elements.
The invention will now be further described with reference to the accompanying drawings and detailed description.
As shown in fig. 1 and 2, the present invention discloses an AI edge calculation fault diagnoser, which is a software and hardware integrated product integrating data acquisition and neural network model training. The method mainly comprises the following steps: the system comprises a platform framework, a software system, a multi-protocol support module, a multi-sensor support module, an AI support module and a supervised learning module, and is realized through the AI support module and the supervised learning module: data processing, model training, model deployment, fault diagnosis and the like.
1. Platform architecture and software system
The platform architecture and the software system construct an edge computing platform, wherein the platform architecture can be an ARM platform architecture or a similar platform architecture with higher comprehensive cost performance, or a general x86 platform architecture or a similar architecture. The software system generally adopts operating systems such as Linux and the like, and is suitable for industrial field application.
2. Multi-protocol support module
The multi-protocol support module comprises at least one bus interface or Ethernet interface used for downward butt joint of the field device and obtaining device operation data and/or performing device control, at least one Ethernet interface used for upward butt joint of an upper computer for receiving control instructions or transmitting data, and a multi-protocol software module used for matching communication protocols and data interaction of the field device.
In order to meet the requirements of network connection mode and distance, the type of ethernet interface can be RJ-45 interface, and also can be optical fiber interface and other types of interfaces. The bus interface comprises at least one of an RS-485 interface and a CAN interface. The fault diagnostor CAN be in bus connection through RS-485 and CAN, one-to-many connection between the fault diagnostor and the diagnosed equipment is realized, and the operation data of each equipment is acquired and/or the equipment is controlled.
The protocol support comprises three types of equipment networking protocols, industrial field networking protocols and application layer networking protocols. The industrial field protocol includes such protocols as Profinet, Ethernet/IP, Ethercat, CC-LINK, Canopen, Devicenet, etc.; equipment protocols include control protocols such as flanker CNC, hidehan CNC, siemens CNC, mitsubishi CNC, injection molding machine, etc.; the application layer protocols comprise protocols such as OPC/OPC UA, MQTT, MTConnect and the like, the multi-protocol support module is provided with a protocol packet, the protocol packet comprises a plurality of protocols, and the protocols are deployed according to application requirements and can be maintained and updated in the application process; the AI edge computing fault diagnoser accesses operational data and/or performs device control for various field devices by one or more of these protocols by docking it to the host computer and receives control commands or passes (computed) data to the host computer.
3. Multi-sensor support module
The sensors include a variety of sensors, such as pressure, temperature, position, etc., for obtaining operational status data of the field device and/or status data of the field environment. The sensors are complex and various, the interfaces are various, and parts of the sensors exist independently, are not integrated in equipment, and cannot be obtained through direct access equipment, so that the interfaces of the multi-sensor support module need to be configured according to application in order to realize data acquisition of each sensor.
And the multi-sensor support module comprises one or more of interfaces such as RS-232, RS-422, I/O (digital quantity signal input/output interface), AD (analog quantity signal acquisition interface) and the like, and is used for directly acquiring and/or controlling data of the sensor.
The RS-485 and CAN interfaces are bus interfaces, CAN be simultaneously connected with a plurality of sensors/equipment, and have wide applicability, and preferably, the multi-sensor support module at least comprises one of the RS-485 and CAN interfaces.
4. AI support module
The AI module is used for preprocessing data, so that the data can meet the requirement of neural network model training, and preprocessing processes such as data cleaning, feature extraction, normalization, feature association and the like are performed. Depending on the field device, the AI edge computing fault diagnoser may or may not optionally support graphical data.
When the graphical data is supported, for example, image recognition is carried out, the AI support module prefers hardware, the AI support module comprises an AI hardware acceleration module and an algorithm library, and the difficulty of loading of the AI model and the difficulty of service integration can be simplified by deploying the AI hardware acceleration module and the algorithm library.
For the occasions of non-graphic input, such as diagnosis of sensor signals, the AI support module can select software, use a C + + AI software library and complete local custom design, training and deployment based on the neural network model based on the processing capacity of the CPU, and does not depend on an upper computer to carry out learning and training of the neural network model. Moreover, due to the integration of localized training and deployment, the input and the output can be used as the training set again, so that the shortage of the early training set or the change of the matching machine caused by long-time operation can be made up.
5. Supervision learning module
The supervised learning module is used for training the neural network model, deploying the trained neural network diagnosis model, and diagnosing faults according to the collected operation data or the operation data processed by the AI support module.
The input data and the output data can be simultaneously acquired through the multi-sensor support module and the multi-protocol support module, the acquired data is input to the neural network model each time, the output of each time is also a mark (with prior knowledge), and the data and the inner meaning of the mark are analyzed to generate a neural network model example with an inference function. As shown in FIG. 3, the model which is obtained by supervised learning of the 2-3 layers of fully-connected neural network is adopted for data input, and error comparison is carried out on the model and actual data, so that the calculation amount is small and the calculation is accurate.
A typical neural network model is shown in fig. 3, with neurons laid out as the following layers. The leftmost layer is an input layer and is responsible for receiving input data; the rightmost layer is the output layer from which we can obtain neural network output data. The layers between the input layer and the output layer are hidden layers because they are not visible to the outside. The absence of connections between neurons in the same layer but with connections between adjacent layers is called full connections (each connection has a weight).
Common neural network models include: BP neural network, quantum neural network of multilayer excitation function, RBF neural network, etc. One skilled in the art can deploy different neural network models in the supervised learning module according to the needs of a specific application to solve different technical problems.
Examples are as follows:
(1) BP neural network model
In the aspect of fault diagnosis of a diesel engine of a certain field device, a BP neural network model is adopted. The model adopts a 3-layer BP neural network model, consists of an output layer, a hidden layer and an output layer, and has self-learning and self-adaptive capabilities; the method has good nonlinear mapping capability and is suitable for solving the problem of complex internal mechanism. Firstly, extracting characteristic data from the existing equipment characteristic signal, preprocessing the characteristic data to be used as the input of a neural network, outputting the known fault result data to be used as the neural network, constructing a BP neural network, training the constructed BP neural network by using a training sample set formed by the existing characteristic data and the known fault result data and performing network self-learning, so that the corresponding relation between the weight and the threshold of the BP neural network and the known fault result is realized to achieve the expected fault result output. And (3) utilizing the trained model to carry out fault diagnosis:
1) the fault samples are input to the nodes of the input layer and are also output by the neurons of the layer.
2) By hidden layer node
Figure BDA0002694964650000071
Is an input. The output of the hidden layer neurons is found and used as the input to the output layer.
3) By
Figure BDA0002694964650000072
The output of the input layer is determined.
4) The final result of the output layer neurons is determined by the threshold function.
The specific implementation steps of the BP neural network are shown in fig. 4.
Taking four example faults in the diesel engine as an example, each fault sample has 5 fault characteristic values, so that the input node of the network is selected to be 5; each output node represents a fault type, so that 4 fault types plus a normal state need to correspond to 5 output nodes; the number of hidden nodes can be adjusted and set to be 60. And then setting a system error, a learning rate, a maximum iteration number, an activation function from an input layer to a hidden layer, an activation function from the hidden layer to an output layer and an output end threshold value.
(2) Quantum neural network of multilayer excitation function
The quantum neural network of multilayer excitation functions improves the conventional BP neural network, is provided based on the idea of quantum state superposition in quantum theory, the excitation functions of hidden quantum neurons adopt superposition of a plurality of traditional excitation functions,the network has inherent ambiguity, and the method can reasonably distribute decision uncertainty data to each fault mode, reduce uncertainty of mode identification and improve accuracy of mode identification. Only the node of the hidden layer with the number of u needs to be output by the function br＝f(WTX- θ), (r ═ 1, 2.., u) instead
Figure BDA0002694964650000081
Wherein theta issIs a quantum spacing; s is the quantum interval number, and the selection of the size of the quantum interval number is the same as the number of the fault modes to be diagnosed; beta is a steepness factor.
(3) RBF neural network
Fault diagnosis on the motor of a field device. The RBF neural network is a three-layer forward local approximation network, is superior to the BP neural network in the aspects of approximation capability, classification capability, learning speed and the like, and has the capability of approximating any function with any precision. The input layer of the RBF neural network is equivalent to an independent variable. The hidden layer selects a base function as an activation function (usually adopting a Gaussian base function), generates nonlinearity on input through a radial (Gaussian) base function, and outputs
Figure BDA0002694964650000082
-wherein
Figure BDA0002694964650000083
The output of the jth unit of the hidden layer is an Euclidean norm; x is input quantity in dimension of M multiplied by Q (M is input layer node, Q is sample number); c. CjIs the center of the jth unit basis function of the hidden layer; sigmajIs the jth perceived variable; n is the number of hidden nodes. The output layer performs linear weighted combination on the output of the hidden layer, and the output of the output layer is
Figure BDA0002694964650000084
-wherein ykIs the output of the kth cell of the output layer; w is ajkThe weight from the jth unit of the hidden layer to the kth unit of the output layer; p is an output layer node.
Such as diagnosing electrical and mechanical faults of a motor. Adopting 10 characteristic frequencies of corresponding faults as characteristic vectors to output, and determining the number of output nodes to be 10; if the electrical fault and the mechanical fault have two main problems respectively, determining the number of output nodes to be 5 by adding a normal state; inputting 40 training samples, and then carrying out RBF neural network training.
6. Differentiated adaptation module
These differences are unavoidable for different types of mechanical equipment, even if of the same type, or all with a difference of varying degrees. Parameters of the supervised learning module are saved and configured through the differentiated adaptation module so as to automatically adapt to different field devices.
If the model is not in a graphic input occasion, due to the integration of design, training and deployment, input and output data can be collected to serve as new training data, after monitoring data are trained for a long time, different training data can be introduced in real time to generate different effects, the parameters are stored in a differentiated adaptation module and are transmitted into a neural network model to be trained to optimize the output of the model, finally, although the types of the neural networks of the same type of equipment are the same, the trained model is most matched with the equipment, and weight parameters between each neural network layer and each layer are the optimal parameters of the machine, so that the difference between the equipment is compensated and automatically adjusted, and the perfect training effect and the optimized output are achieved.
Example 1:
the embodiment provides an application of a fault diagnosis device in fault diagnosis of a certain machine tool.
The fault diagnosis device comprises an edge computing platform constructed by an ARM platform architecture and a Linux operating system, and comprises two Ethernet interfaces including a LAN1, a LAN2, 1 RS-485 interface and 1 CAN interface, wherein one Ethernet interface CAN be used for being connected with an upper computer and interacting with the upper computer, and the other Ethernet interface CAN be used for being connected with a diagnosed field device. The RS-485 and CAN interfaces CAN be connected with devices, facilities and sensors with corresponding bus interfaces, and a plurality of devices, facilities and sensors CAN be simultaneously connected through the 485 bus and/or the CAN bus to carry out data acquisition. The fault diagnosis device is of a small-size structure and has a very simple interface. To facilitate the connection of multiple devices, sensors, the fault diagnoser includes at least one of an RS-485 interface and a CAN interface. Preferably, in order to satisfy the requirement of protocol support as much as possible on the basis of extremely simple and small interface, the fault diagnosis device is provided with two Ethernet interfaces, an RS-485 interface and a CAN interface.
In the embodiment, the fault diagnosis device is in communication connection with a temperature sensor and a pressure sensor of the machine tool and the surrounding environment through an RS-485 interface, and is used for collecting data such as temperature, pressure and the like.
In the embodiment, the fault diagnosis device is connected with the machine tool through an ethernet interface LAN1, so as to realize the acquisition and transmission of voltage, current data, x, y, z axes of a processing edge, rotation speed and other data inside the machine tool.
The data are transmitted into a linux system and are given to a C + + AI software library to process the data, and necessary data cleaning, feature extraction, normalization, feature correlation and other data are prepared in an early stage.
And then, carrying out neural network model training: acquiring a large amount of sensor data of a machine tool and the surrounding environment and operation data of the machine tool on line, preprocessing the sensor data and the operation data through a C + + AI software database, inputting the data preprocessed through the C + + AI software database into input nodes of a BP (back propagation) neural network, and assuming that the machine tool has 8-dimensional data, determining the number of the input nodes to be 8; the output nodes are set according to the number of the judged faults, for example, if the normal state plus other four fault states exist, the output nodes are set to be 5 outputs; the number of the intermediate nodes can be set to 30 according to the precision adjustment, and the connection weight between the neural networks is automatically adjusted by adopting a BP algorithm.
The trained model meeting the precision condition is locally deployed; after deployment is completed, real-time data acquisition and data diagnosis can be performed.
The upper-layer operation controller (upper computer) is interacted through the Ethernet interface LAN2, diagnosis results and collected data can be transmitted through the Ethernet interface to be checked and stored, the upper computer can also control and debug the fault diagnostors on a plurality of lower layers through instructions, for example, the fault diagnostors which are not trained sufficiently can be retrained through the instructions by the upper computer, and the stored data and the results of the upper computer are transmitted to the fault diagnostors on the lower layers again to carry out model training and deployment.
The AI edge calculation fault diagnotor provided by the invention uses a mode of combining edge calculation and artificial intelligence, analyzes and infers acquired real-time data by applying a bottom layer numerical model of a neural network, takes the neural network as a representation and processing mode of a knowledge source and is mutually fused with other inference mechanisms, realizes multi-mode inference, and can effectively diagnose an unexpected mode of equipment; and the privacy, the rapidity and the safety of data transmission are ensured by adopting localized deployment.
While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (9)

1. An AI-edge calculation fault diagnoser, comprising: the method comprises the following steps: the system comprises a platform architecture, a software system, a multi-protocol support module, an AI support module, a supervision learning module and a differentiation adaptation module;
the platform architecture and the software system are used for constructing an edge computing platform;
the multi-protocol support module comprises at least one bus interface or Ethernet interface, at least one protocol packet and at least one control module, wherein the at least one bus interface or Ethernet interface is used for downwards butting the field equipment and acquiring the operating data of the field equipment;
the AI supporting module is used for preprocessing the acquired data;
the supervised learning module is used for carrying out neural network model training through the preprocessed data acquired on line to acquire a fault diagnosis model; deploying a trained fault diagnosis model, and performing fault diagnosis according to the preprocessed data acquired on line;
the differentiation adapting module is used for saving and configuring parameters of the supervision learning module so as to adapt to field equipment.
2. The AI-edge computation fault diagnoser of claim 1, wherein: the platform architecture is an ARM platform architecture or an X86 platform architecture.
3. The AI-edge computation fault diagnoser of claim 1, wherein: the software system is a Linux operating system.
4. The AI-edge computation fault diagnoser of claim 1, wherein: the protocol packet comprises a plurality of protocols in an equipment networking protocol, an industrial field networking protocol and an application layer networking protocol.
5. The AI-edge computation fault diagnoser of claim 1, wherein: the bus interface at least comprises one of an RS-485 interface and a CAN interface.
6. The AI-edge computation fault diagnoser of claim 5, wherein: the multi-protocol support module at least comprises two Ethernet interfaces, an RS-485 interface and a CAN interface.
7. The AI-edge computation fault diagnoser of claim 1, wherein: the multi-sensor support module is also included, and the interface types supported by the multi-sensor module comprise a digital quantity input and output interface, an analog quantity signal acquisition interface, an RS-232 serial interface and an RS-422 serial interface.
8. The AI-edge computation fault diagnoser of claim 1, wherein: the AI support module includes an AI hardware acceleration module and an algorithm library, and/or includes an AI software library.
9. The AI-edge computation fault diagnoser of claim 1, wherein: the neural network model deployed by the supervised learning module includes but is not limited to: BP neural network, quantum neural network of multilayer excitation function and RBF neural network.
CN202011002985.3A 2020-09-22 2020-09-22 AI edge calculation fault diagnosis device Pending CN112164153A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011002985.3A CN112164153A (en) 2020-09-22 2020-09-22 AI edge calculation fault diagnosis device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011002985.3A CN112164153A (en) 2020-09-22 2020-09-22 AI edge calculation fault diagnosis device

Publications (1)

Publication Number Publication Date
CN112164153A true CN112164153A (en) 2021-01-01

Family

ID=73862685

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011002985.3A Pending CN112164153A (en) 2020-09-22 2020-09-22 AI edge calculation fault diagnosis device

Country Status (1)

Country Link
CN (1) CN112164153A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222170A (en) * 2021-03-30 2021-08-06 新睿信智能物联研究院(南京)有限公司 Intelligent algorithm and model for IOT (Internet of things) AI (Artificial Intelligence) collaborative service platform
CN113640027A (en) * 2021-08-12 2021-11-12 常州英集动力科技有限公司 Heat exchange unit fault diagnosis method and system based on edge calculation and neural network
CN114006793A (en) * 2021-11-29 2022-02-01 南京工业大学 Edge computing gateway for monitoring spontaneous combustion disasters of sulfide corrosion products
CN113222170B (en) * 2021-03-30 2024-04-23 新睿信智能物联研究院(南京)有限公司 Intelligent algorithm and model for AI collaborative service platform of Internet of things

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106291339A (en) * 2015-05-20 2017-01-04 国网上海市电力公司 A kind of circuit breaker failure diagnostic expert system based on artificial neural network
US20170206464A1 (en) * 2016-01-14 2017-07-20 Preferred Networks, Inc. Time series data adaptation and sensor fusion systems, methods, and apparatus
CN106981063A (en) * 2017-03-14 2017-07-25 东北大学 A kind of grid equipment state monitoring apparatus based on deep learning
CN107193271A (en) * 2017-06-13 2017-09-22 青岛科技大学 Preposition service adapter and trouble-shooter for industrial information physical system
CN108491580A (en) * 2018-02-26 2018-09-04 上海理工大学 A kind of equipment fault diagnosis apparatus and system
CN109063308A (en) * 2018-07-26 2018-12-21 北京航空航天大学 A kind of health evaluating method based on depth quantum learning
CN110119333A (en) * 2019-02-21 2019-08-13 北京天泽智云科技有限公司 A kind of abnormality detection edge calculations system
CN110221558A (en) * 2019-06-05 2019-09-10 镇江四联机电科技有限公司 A kind of electrohydraulic servo valve on-line fault diagnosis gateway based on edge calculations technology
CN111523660A (en) * 2020-04-15 2020-08-11 南京清然能源科技有限公司 Audio-visual-thermal integrated anomaly detection and alarm method based on artificial intelligence

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106291339A (en) * 2015-05-20 2017-01-04 国网上海市电力公司 A kind of circuit breaker failure diagnostic expert system based on artificial neural network
US20170206464A1 (en) * 2016-01-14 2017-07-20 Preferred Networks, Inc. Time series data adaptation and sensor fusion systems, methods, and apparatus
CN106981063A (en) * 2017-03-14 2017-07-25 东北大学 A kind of grid equipment state monitoring apparatus based on deep learning
CN107193271A (en) * 2017-06-13 2017-09-22 青岛科技大学 Preposition service adapter and trouble-shooter for industrial information physical system
CN108491580A (en) * 2018-02-26 2018-09-04 上海理工大学 A kind of equipment fault diagnosis apparatus and system
CN109063308A (en) * 2018-07-26 2018-12-21 北京航空航天大学 A kind of health evaluating method based on depth quantum learning
CN110119333A (en) * 2019-02-21 2019-08-13 北京天泽智云科技有限公司 A kind of abnormality detection edge calculations system
CN110221558A (en) * 2019-06-05 2019-09-10 镇江四联机电科技有限公司 A kind of electrohydraulic servo valve on-line fault diagnosis gateway based on edge calculations technology
CN111523660A (en) * 2020-04-15 2020-08-11 南京清然能源科技有限公司 Audio-visual-thermal integrated anomaly detection and alarm method based on artificial intelligence

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222170A (en) * 2021-03-30 2021-08-06 新睿信智能物联研究院(南京)有限公司 Intelligent algorithm and model for IOT (Internet of things) AI (Artificial Intelligence) collaborative service platform
CN113222170B (en) * 2021-03-30 2024-04-23 新睿信智能物联研究院(南京)有限公司 Intelligent algorithm and model for AI collaborative service platform of Internet of things
CN113640027A (en) * 2021-08-12 2021-11-12 常州英集动力科技有限公司 Heat exchange unit fault diagnosis method and system based on edge calculation and neural network
CN114006793A (en) * 2021-11-29 2022-02-01 南京工业大学 Edge computing gateway for monitoring spontaneous combustion disasters of sulfide corrosion products
CN114006793B (en) * 2021-11-29 2022-07-22 南京工业大学 Edge computing gateway for monitoring spontaneous combustion disasters of sulfide corrosion products

Similar Documents

Publication Publication Date Title
CN110554657B (en) Health diagnosis system and diagnosis method for operation state of numerical control machine tool
KR100532804B1 (en) Method for automatic operation of industrial plants
US8126679B2 (en) Automatic remote monitoring and diagnostics system
Mohebbi et al. Multi-criteria fuzzy decision support for conceptual evaluation in design of mechatronic systems: a quadrotor design case study
EP1672535A1 (en) Distributed intelligent diagnostic scheme
CN112164153A (en) AI edge calculation fault diagnosis device
CN110216680B (en) Cloud-ground cooperative fault diagnosis system and method for service robot
Mönks et al. Towards distributed intelligent sensor and information fusion
EP2287786A1 (en) Fuzzy inference apparatus and methods, systems and apparatuses using such inference apparatus
CN104615121B (en) A kind of train fault diagnostic method and system
CN109711062A (en) A kind of equipment fault diagnosis method and device based on cloud service
Villalonga et al. Visual analytics framework for condition monitoring in cyber-physical systems
CN113319462B (en) Welding robot control method and device based on edge cloud cooperation
Barig et al. Assistance systems for industry 4.0 environments
CN108494626B (en) Intelligent diagnosis method for communication faults of Profibus DP industrial field bus with improper physical installation
WO2018224649A1 (en) Method and distributed control system for carrying out an automated industrial process
US11494195B2 (en) Interface device and method for configuring the interface device
CN111650898B (en) Distributed control system and method with high fault tolerance performance
CN117372630B (en) Data visualization system and method based on digital twin technology
Shprekher et al. System of a Remotely Control of Technical Condition of Complex Electrical Objects
Strelec et al. IIoT device prototype design using state machine according to OPC UA
Velásquez et al. A Novel Architecture Definition for AI-Driven Industry 4.0 Applications
Viharos et al. AI techniques in modelling, assignment, problem solving and optimization
Westbrink et al. Integrated IPC for data-driven fault detection
CN115906990A (en) Efficient second order pruning of computer-implemented neural networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210101