A method for establishing a fault model based on alliance chain

Abstract

The present invention proposes a method for establishing a fault model based on an alliance chain, comprising the following steps: using historical production data to train and establish a production fault identification model; constructing an alliance chain for fault identification, and encrypting and transmitting data between nodes through the alliance chain; obtaining real-time production data of the sewing machine operation and transmitting it to the production fault identification model to determine whether the production fault identification model successfully identifies the fault type, and if so, feeding back the fault type to the enterprise node; if not, extracting the production data corresponding to the unknown fault and feeding it back to the enterprise node, and the enterprise node feeds back the production data of the unknown fault to the operator node through the alliance chain. The present invention encrypts the fault data by constructing an alliance chain to prevent data loss or tampering, so that subsequent fault analysis can be carried out safely, and at the same time, the subsequent fault solution is also transmitted through the alliance chain to prevent the solution from being tampered with.

Description

Technical Field

[0001] The present invention relates to the technical field of fault judgment, and in particular to a method for establishing a fault model based on an alliance chain. Background Art

[0002] With the development of the textile industry and the improvement of automatic control technology, the sewing industry has an increasing demand for automated equipment, which puts higher requirements on the level of automated control of sewing equipment. Industrial sewing machines are typical automated equipment in the sewing industry. They are mainly suitable for sewing factories or other industrial departments that mass-produce sewing products, such as various quilts, sleeping bags, curtains, etc. To achieve automatic control of industrial sewing machines, there must be a control system. The control system includes a main control unit MCU. The main control unit MCU transmits signals to the motor controller through the electronic control connection interface. The motor controller controls the motor to complete the sewing work. The automatic control of industrial sewing machines in the industrial field is to increase product output and improve work efficiency. Therefore, in the existing working environment, many industrial sewing machines often work at the same time. Industrial sewing machines must be managed by staff to deal with some unexpected situations in a timely manner. If each industrial sewing machine is equipped with a staff member, it is undoubtedly a waste of human resources; if the working status of the industrial sewing machine is managed by the mobile monitoring of the staff, the workload of the staff is very large, and accidents may occur due to inadequate monitoring; at the same time, once a problem occurs, the staff responsible for monitoring and management may not know where the problem lies, and still need to communicate with the manufacturer of the industrial sewing machine and have professional maintenance personnel to repair the industrial sewing machine. For users of industrial sewing machines, it not only wastes human resources, but also fails to solve the problem in time, which brings great inconvenience to users of industrial sewing machines and delays production. Referring to a remote monitoring and fault diagnosis system for an industrial sewing machine with a Chinese patent authorization announcement number of CN103451866B, the system comprises a motor, a motor controller, an electric control connection interface and a main control unit installed on the industrial sewing machine and connected in sequence, and also comprises a data collector connected to the main control unit, the data collector is connected to the repeater through a data transmission device, a GPRS / GSM module is provided in the repeater, the repeater is connected to the network server through a communication connection, the network server is respectively connected to the display terminal and the fault diagnosis system, the machine setting parameters of the motor and other information are transmitted to the repeater through the data transmission device through the data collector, the repeater transmits the received information to the customer service website server through the built-in GPRS / GSM module, and the customer service staff remotely monitors the working status of the industrial sewing machine through the display terminal and the fault diagnosis system connected to the website server, so as to help solve the problem in time. However, there is a lack of effective confidentiality measures in the process of network transmission, and data may be lost or tampered with during the network transmission process, resulting in the inability to achieve correct fault diagnosis later. Summary of the invention

[0003] The present invention solves the problem of insecure network transmission for remote monitoring and fault diagnosis, and proposes a fault model establishment method based on alliance chain. By building an alliance chain, fault data is encrypting to prevent data loss or tampering, so that subsequent fault analysis can be carried out safely. At the same time, subsequent fault solutions are also transmitted through the alliance chain to prevent the solutions from being tampered with.

[0004] In order to achieve the above objectives, the following technical solutions are proposed:

[0005] A method for establishing a fault model based on a consortium chain includes the following steps:

[0006] S1, obtaining historical production data of the sewing machine operation, using the historical production data to train and establish a production fault identification model, wherein the production fault identification model is provided with a fault type library;

[0007] S2, building a consortium chain for fault identification, the consortium chain includes several enterprise nodes and operator nodes, and data encryption is transmitted between the nodes through the consortium chain;

[0008] S3, obtain the real-time production data of the sewing machine operation and transmit it to the production fault identification model to determine whether the production fault identification model successfully identifies the fault type. If so, feedback the fault type to the enterprise node; if not, proceed to S4;

[0009] S4: Extract the production data corresponding to the unknown fault and feed it back to the enterprise node. The enterprise node feeds back the production data of the unknown fault to the operator node through the alliance chain.

[0010] S5, the operator node identifies the production data of unknown faults, obtains new fault types, and sends the data of new fault types to each enterprise node through the alliance chain. The fault identification model updates the fault type library.

[0011] In the present invention, one enterprise node corresponds to one production fault identification model, and one production fault identification model can simultaneously identify several sewing machines corresponding to the enterprise node. When an unknown fault occurs and needs to be resolved by the operator node, the fault data can be encrypted and uploaded to the operator node through the alliance chain. Due to the tamper-proof nature of the alliance chain transmission, the fault data is the original collected and real data, which facilitates the operator to find a solution. At the same time, due to data sharing within the alliance chain, when a sewing machine has a new fault, the operator node can synchronize the solution to the new fault to each enterprise node after resolving the fault. The production fault identification model of each enterprise node is updated regularly, so that timely response can be made when the same fault occurs.

[0012] Preferably, the production fault identification model is provided with a fault processing library, which records the corresponding fault processing methods. When the operator node obtains a new fault type, it provides a new fault processing method and transmits it to each enterprise node. Then the fault identification model updates the fault processing library.

[0013] Preferably, the S1 specifically comprises the following steps:

[0014] S101, simultaneously collecting and denoising several types of vibration data when the sewing machine fails to operate, sampling the several types of vibration data in time sequence, and obtaining historical production data;

[0015] S102, generating a training set and a test set based on the historical production data according to a set time step;

[0016] S103, setting a label representing a type of mechanical equipment failure for the training set;

[0017] S104. Train a double-layer LSTM neural network using the training set and the test set; the double-layer LSTM neural network is used to determine the fault type of mechanical equipment failure; the test set is used to verify the trained LSTM neural network; the trained LSTM neural network after verification is the production fault identification model.

[0018] Preferably, the two-layer LSTM neural network comprises: a first hidden layer, a second hidden layer, and a flattening module;

[0019] When training the two-layer LSTM network for the i-th time, i∈[1,L-w+1], where L is a constant, including:

[0020] The time series data x in the test set i ,…,x i+w The first hidden layer is input in sequence, and after the LSTM standard module of the first hidden layer performs w time steps of operation, the first output result of the corresponding time step is obtained; where x i ,…,x i+w They are respectively used as the input of the LSTM standard module of the first hidden layer at the w time steps;

[0021] The first output results are sequentially input into the second hidden layer, and after the LSTM standard module of the second hidden layer performs operations for w time steps, the second output results of the corresponding time step are obtained, wherein the first output results are respectively used as the input of the LSTM standard module of the second hidden layer in the w time steps;

[0022] The second output result is input into the flattening module to obtain a classification result of the mechanical equipment fault type, and the flattening module includes a classifier.

[0023] Preferably, the process of encrypting and transmitting data between the nodes through the alliance chain is as follows:

[0024] S201, each node stores a separate key. When a node uploads shared data, the node converts the data to be shared into a data block, annotates the data block with plain text, and uploads it to the alliance chain intranet platform;

[0025] S202, the alliance chain intranet platform divides the data block into several data blocks, and each data block is sent to several nodes separately according to a certain order; the several nodes encrypt the data blocks with keys, and the data blocks encrypted by the several nodes are combined into encrypted shared data blocks according to a certain order, and the encrypted shared data blocks are annotated with the plaintext, and the number of nodes participating in the encryption is more than half of the number of all nodes;

[0026] S203, the alliance chain intranet platform displays the encrypted shared data block and allows the encrypted shared data block to be copied to any node.

[0027] Preferably, the present invention further comprises a decryption step of the shared data block:

[0028] When a computer application in the intranet environment of the alliance chain requests to open shared data, a request is sent to the alliance chain intranet platform. The alliance chain intranet platform verifies whether the application is in the whitelist. If the application is in the whitelist, the alliance chain intranet platform sends a decryption request to the nodes involved in the encryption. When each node has completed decryption, the decrypted shared data is placed in a specified buffer, and the data reading address of the application is replaced with the buffer address. When the application is closed, the encryption and decryption listening node clears the buffer. If the application is not in the whitelist, no operation is performed.

[0029] The beneficial effects of the present invention are as follows: the fault data is encrypted and uploaded to the operator node through the alliance chain. Due to the tamper-proof nature of the alliance chain transmission, the fault data is the original collected and real data, which facilitates the operator to find a solution. At the same time, due to data sharing within the alliance chain, when a sewing machine has a new fault, the operator node can synchronize the solution to the new fault to each enterprise node after solving the fault. The production fault identification model of each enterprise node is regularly updated so that timely response can be made when the same fault occurs. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a method flow chart of an embodiment. DETAILED DESCRIPTION

[0031] Example:

[0032] This embodiment proposes a method for establishing a fault model based on a consortium chain. Figure 1 , including the following steps:

[0033] S1, obtain the historical production data of the sewing machine operation, use the historical production data to train and establish a production fault identification model, the production fault identification model is equipped with a fault type library; the production fault identification model is equipped with a fault processing library, the fault processing library records the corresponding fault processing methods, the operator node obtains a new fault type and provides a new fault processing method, which is transmitted to each enterprise node, and then the fault identification model updates the fault processing library.

[0034] S101, simultaneously collect and denoise several types of vibration data when the sewing machine fails, sample several types of vibration data in time sequence, and obtain historical production data; the sampled data set is recorded as E, E = {e r}r∈[1,num]; num is the total number of sampling times; e r ={e′ r1 ,…,e′ rp};e r ∈R m , e′ rp is the p-th type of vibration data sampled for the r-th time, p∈[1,m], and m is the total number of categories of vibration data.

[0035] S102, according to the set time step, generate a training set and a test set based on historical production data; let x i =[e (i-1)×s+1 ,e (i-1)×s+2 ,…,e i×s ]′,x i ∈R sm Where i∈[1,num / s], s is the set time step, [·]′ represents the transpose of the matrix; x1~x L As the training set, x L+1 ~x num / s Will be used as the test set, and L is the set constant.

[0036] S103, setting a label representing a type of mechanical equipment failure for the training set;

[0037] S104. Train a two-layer LSTM neural network through a training set and a test set; the two-layer LSTM neural network is used to determine the fault type of mechanical equipment failure; the test set is used to verify the trained LSTM neural network; the trained LSTM neural network after verification is the production fault identification model.

[0038] The two-layer LSTM neural network includes: the first hidden layer, the second hidden layer, and the flattening module;

[0039] When training the two-layer LSTM network for the i-th time, i∈[1,L-w+1], where L is a constant, including: i ,…,x i+w The first hidden layer is input sequentially, and after the LSTM standard module of the first hidden layer performs w time steps of operation, the first output result of the corresponding time step is obtained. where x i ,…,x i+w As the input of the LSTM standard module of the first hidden layer in the w time steps, the first output result The second hidden layer is input in sequence, and after the LSTM standard module of the second hidden layer performs w time steps of operation, the second output result of the corresponding time step is obtained. The first output result They are respectively used as the input of the LSTM standard module of the second hidden layer at the w time steps;

[0040] The second output result The flattening module is input to obtain the classification result of the mechanical equipment failure type, and the flattening module includes a classifier.

[0041] The LSTM standard module performs one time step of operations including:

[0042] S141, the number of hidden units of the LSTM standard module is n, s is the set time step; the current time step is t, and the time step t-1 is the previous time step of time step t; x t ∈R sm represents the input at time step t, h t-1 represents the hidden state at time step t-1, h t-1 ∈R n , the forget gate f of the LSTM standard module at time step t t and Memory Gate i t They are:

[0043] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0044] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0045] Among them, f t and i t ∈R sn; · represents matrix operation, σ is the sigmoid activation function, W f and W t ∈R (n+m)s is the weight parameter, b f and b i ∈R n is the bias parameter;

[0046] S142. Calculate the new state candidate at time step t:

[0047]

[0048] Among them, tanh is the activation function, W c is the weight parameter, b c is the bias parameter;

[0049] S143, calculate the update gate C of time step t t AND output gate o t :

[0050]

[0051] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0052] Among them, C t With o t ∈R sn , C t-1 is the update gate at time step t-1, C t-1 ∈R sn , W o is the weight parameter, b o is the bias parameter;

[0053] S144, h t is the hidden state of the LSTM standard module at time step t, and is also the output result of the LSTM standard module at time step t:

[0054] h t =o t *tanh(C t )

[0055] Among them, h t ∈R sn .

[0056] S2, build a consortium chain for fault identification. The consortium chain includes several enterprise nodes and operator nodes. Data is encrypted and transmitted between nodes through the consortium chain.

[0057] S3, obtain the real-time production data of the sewing machine operation and transmit it to the production fault identification model to determine whether the production fault identification model successfully identifies the fault type. If so, feedback the fault type to the enterprise node; if not, proceed to S4;

[0058] S4: Extract the production data corresponding to the unknown fault and feed it back to the enterprise node. The enterprise node feeds back the production data of the unknown fault to the operator node through the alliance chain.

[0059] S5, the operator node identifies the production data of unknown faults, obtains new fault types, and sends the data of new fault types to each enterprise node through the alliance chain. The fault identification model updates the fault type library.

[0060] The process of data encryption transmission between nodes through the alliance chain is as follows:

[0061] S201, each node stores a separate key. When a node uploads shared data, the node converts the data to be shared into a data block, annotates the data block with plain text, and uploads it to the alliance chain intranet platform;

[0062] S202, the alliance chain intranet platform divides the data block into several data blocks, and each data block is sent to several nodes separately according to a certain order; several nodes use keys to encrypt the data blocks, and the data blocks encrypted by several nodes are combined into encrypted shared data blocks according to a certain order. The encrypted shared data blocks are annotated with plain text, and the number of nodes participating in the encryption is more than half of the number of all nodes;

[0063] S203, the alliance chain intranet platform displays the encrypted shared data blocks and allows the encrypted shared data blocks to be copied to any node.

[0064] The present invention also includes a decryption step of the shared data block:

[0065] When a computer application in the intranet environment of the alliance chain requests to open shared data, a request is sent to the alliance chain intranet platform. The alliance chain intranet platform verifies whether the application is in the whitelist. If the application is in the whitelist, the alliance chain intranet platform sends a decryption request to the nodes involved in the encryption. When each node has completed decryption, the decrypted shared data is placed in the specified buffer, and the application's data reading address is replaced with the buffer address. When the application is closed, the encryption and decryption listening node clears the buffer. If the application is not in the whitelist, no action is taken.

[0066] In order to clearly express the encryption and decryption process of the alliance chain, the data sharing process of a specific node is explained below: the shared data includes fault data and fault solution data.

[0067] In the first step, node A converts the data to be shared into data block D k Node A annotates the data block in plain text and uploads it to the alliance chain intranet platform;

[0068] In the second step, the alliance chain intranet platform divides the data block into several data blocks, and each data block is sent to several nodes separately according to a certain order; several nodes use keys to encrypt the data blocks, and the data blocks encrypted by several nodes are combined into encrypted shared data blocks according to a certain order. The encrypted shared data blocks are annotated with plain text, and the number of nodes participating in the encryption is more than half of the number of all nodes; specifically, the following steps are included:

[0069] a. The alliance chain intranet platform will transfer data block D k Divided into 6 / 2 = 3 data blocks d ki，i∈[1，3] , k is the data block number, in this case k is A; when n is an odd number, it is divided into (n+1) / 2 data blocks;

[0070] b. Establish the encryption node sequence group corresponding to node A.

[0071] Select nodes other than node A, and perform permutations and combinations according to the time they entered the alliance chain intranet. Select the sequence combination whose number in the permutations and combinations reaches the set threshold as the encrypted node sequence group;

[0072] Each sequence in the encryption node sequence group divides the data into blocks d ki，i∈[1，3] Encryption is performed;

[0073] c. The alliance chain intranet platform displays the encrypted shared data block D′ A , and allows the encrypted shared data block D′ A Copy to any node.

[0074] Decryption steps:

[0075] When a computer application in the intranet environment of the alliance chain requests to open shared data, a request is sent to the alliance chain intranet platform. The alliance chain intranet platform verifies whether the application is in the whitelist. If the application is in the whitelist, the alliance chain intranet platform sends a decryption request to the nodes involved in the encryption. When each node has completed decryption, the decrypted shared data is placed in the specified buffer, and the application's data reading address is replaced with the buffer address. When the application is closed, the encryption and decryption listening node clears the buffer. If the application is not in the whitelist, no action is taken.

[0076] Node B obtains the shared data block D' uploaded by node A A Take this as an example, the specific decryption process is as follows:

[0077] d. Node B, in the intranet environment, uses a computer application to request to open shared data and sends a request to the alliance chain intranet platform;

[0078] e. The alliance chain intranet platform verifies whether the application is in the whitelist. If so, proceed to S6. If not, do nothing.

[0079] f. The alliance chain intranet platform sends a decryption request to the nodes involved in the encryption. When each node has completed the decryption, it puts the decrypted shared data into the specified buffer. The specific steps include:

[0080] The alliance chain intranet platform extracts the shared data block D′ A The encrypted data block combination; traverse the encrypted data block, divide the encrypted data block into encrypted data blocks, send them to each encryption node for decryption, until the encrypted data block can be completely decrypted, and restore the decrypted block combination to data block D k , and the data block D k Put into the specified buffer.

[0081] In the present invention, one enterprise node corresponds to one production fault identification model, and one production fault identification model can simultaneously identify several sewing machines corresponding to the enterprise node. When an unknown fault occurs and needs to be resolved by the operator node, the fault data can be encrypted and uploaded to the operator node through the alliance chain. Due to the tamper-proof nature of the alliance chain transmission, the fault data is the original collected and real data, which facilitates the operator to find a solution. At the same time, due to data sharing within the alliance chain, when a sewing machine has a new fault, the operator node can synchronize the solution to the new fault to each enterprise node after resolving the fault. The production fault identification model of each enterprise node is updated regularly, so that timely response can be made when the same fault occurs.

Claims

1. A method for establishing a fault model based on a consortium chain, characterized in that: The following steps are involved: S1, obtaining historical production data of the sewing machine operation, using the historical production data to train and establish a production fault identification model, wherein the production fault identification model is provided with a fault type library; S2, building a consortium chain for fault identification, the consortium chain includes several enterprise nodes and operator nodes, and data encryption is transmitted between the nodes through the consortium chain; S3, obtain the real-time production data of the sewing machine operation and transmit it to the production fault identification model to determine whether the production fault identification model successfully identifies the fault type. If so, feedback the fault type to the enterprise node; if not, proceed to S4; S4: Extract the production data corresponding to the unknown fault and feed it back to the enterprise node. The enterprise node feeds back the production data of the unknown fault to the operator node through the alliance chain. S5, the operator node identifies the production data of unknown faults, obtains new fault types, and sends the data of new fault types to each enterprise node through the alliance chain. The fault identification model updates the fault type library.

2. According to the method for establishing a fault model based on alliance chain according to claim 1, it is characterized in that: The production fault identification model is provided with a fault processing library, which records the corresponding fault processing methods. When the operator node obtains a new fault type, it provides a new fault processing method and transmits it to each enterprise node. Then the fault identification model updates the fault processing library.

3. The method for establishing a fault model based on alliance chain according to claim 1 is characterized in that: The S1 specifically includes the following steps: S101, simultaneously collecting and denoising several types of vibration data when the sewing machine fails to operate, sampling the several types of vibration data in time sequence, and obtaining historical production data; S102, generating a training set and a test set based on the historical production data according to a set time step; S103, setting a label representing a type of mechanical equipment failure for the training set; S104. Train a double-layer LSTM neural network using the training set and the test set; the double-layer LSTM neural network is used to determine the fault type of mechanical equipment failure; the test set is used to verify the trained LSTM neural network; the trained LSTM neural network after verification is the production fault identification model.

4. The method for establishing a fault model based on alliance chain according to claim 3 is characterized in that: The double-layer LSTM neural network includes: a first hidden layer, a second hidden layer, and a flattening module; When training the two-layer LSTM network for the i-th time, i∈[1,L-w+1], where L is a constant, including: The time series data x in the test set i ,…,x i+w The first hidden layer is input in sequence, and after the LSTM standard module of the first hidden layer performs w time steps of operation, the first output result of the corresponding time step is obtained; where x i ,…,x i+w They are respectively used as the input of the LSTM standard module of the first hidden layer at the w time steps; The first output results are sequentially input into the second hidden layer, and after the LSTM standard module of the second hidden layer performs operations for w time steps, the second output results of the corresponding time step are obtained, wherein the first output results are respectively used as the input of the LSTM standard module of the second hidden layer in the w time steps; The second output result is input into the flattening module to obtain a classification result of the mechanical equipment fault type, and the flattening module includes a classifier.

5. The method for establishing a fault model based on alliance chain according to claim 1 is characterized in that: The process of data encryption transmission between nodes through the alliance chain is as follows: S201, each node stores a separate key. When a node uploads shared data, the node converts the data to be shared into a data block, annotates the data block with plain text, and uploads it to the alliance chain intranet platform; S202, the alliance chain intranet platform divides the data block into several data blocks, and each data block is sent to several nodes separately according to a certain order; the several nodes encrypt the data blocks with keys, and the data blocks encrypted by the several nodes are combined into encrypted shared data blocks according to a certain order, and the encrypted shared data blocks are annotated with the plaintext, and the number of nodes participating in the encryption is more than half of the number of all nodes; S203, the alliance chain intranet platform displays the encrypted shared data block and allows the encrypted shared data block to be copied to any node.

6. The method for establishing a fault model based on alliance chain according to claim 1 is characterized in that: It also includes the decryption step of the shared data block: When a computer application in the intranet environment of the alliance chain requests to open shared data, a request is sent to the alliance chain intranet platform. The alliance chain intranet platform verifies whether the application is in the whitelist. If the application is in the whitelist, the alliance chain intranet platform sends a decryption request to the nodes involved in the encryption. When each node has completed decryption, the decrypted shared data is placed in a specified buffer, and the data reading address of the application is replaced with the buffer address. When the application is closed, the encryption and decryption listening node clears the buffer. If the application is not in the whitelist, no operation is performed.

Images