A distillation monitoring method and system based on neural network
By building a neural network-based distillation monitoring system, using historical data to train models and monitor the electronic waste liquid distillation process in real time, the problem of abnormal prediction of electronic waste liquid distillation was solved, and the distillation quality and data security were guaranteed.
Patent Information
- Application Number
- CN202311725187.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-15
- Publication Date
- 2025-10-03
- Estimated Expiration
- 2043-12-15
AI Technical Summary
Existing technologies make it difficult to predict anomalies during the distillation process of electronic waste liquid, resulting in reduced distillation quality and waste of resources.
By obtaining historical monitoring data of the distillation system, a data set is constructed and divided into training set, validation set and test set. A distillation monitoring model is created using a neural network. The data is monitored in real time and reports are pushed to predict anomalies.
It achieves timely prediction of electronic waste liquid distillation anomalies, ensures distillation quality, and improves data security and reliability.
Smart Images
Figure CN117906667B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of electronic waste liquid recycling, and in particular to a distillation monitoring method and system based on a neural network. Background Art
[0002] The electronics and optoelectronics industries generate a significant amount of waste electronic liquid during the production of chips and circuit boards. For example, a mixture of propylene glycol methyl ether (PM) and propylene glycol monomethyl ether acetate (PMA) is often used as a primary component in photoresist thinners, photoresist removers, photoresist buffers, and cleaning agents to remove esters such as photoresist from panels. Etching solutions and other reagents are also used during the actual production process, and these become waste electronic liquid after use.
[0003] Since electronic waste liquid contains a large amount of impurities such as propylene glycol methyl ether, propylene glycol monomethyl ether acetate, photoresist, etc., it poses a great threat to the environment. In addition, the recycling and reuse of propylene glycol methyl ether and propylene glycol monomethyl ether acetate has high commercial value. Therefore, there is a need to distill electronic waste liquid to recycle and reuse valuable raw materials.
[0004] The distillation of electronic waste liquid requires the use of a distillation system consisting of an evaporator, a first distillation tower, a second distillation tower, a finished product tank and other equipment. Since the distillation system is inevitably prone to failure and aging during long-term operation, or changes in distillation parameters due to changes in environmental factors, these will directly affect the distillation quality of the electronic waste liquid. If the raw materials separated by distillation still carry a large amount of impurities, this part of the raw materials will be discarded due to unusability, or need to be distilled again, which will undoubtedly cause great waste. If the distillation process can be predicted in advance and the problem can be handled accordingly, the waste of resources will be greatly reduced.
[0005] Therefore, how to provide a distillation monitoring method and system based on a neural network to predict the distillation anomaly of electronic waste liquid and ensure the distillation quality of electronic waste liquid has become a technical problem that needs to be solved urgently. Summary of the Invention
[0006] The technical problem to be solved by the present invention is to provide a distillation monitoring method and system based on a neural network, so as to predict the distillation anomaly of electronic waste liquid and ensure the distillation quality of electronic waste liquid.
[0007] In a first aspect, the present invention provides a distillation monitoring method based on a neural network, comprising the following steps:
[0008] Step S10: Acquire historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH, and distillation time, and construct a data set after preprocessing each of the historical monitoring data;
[0009] Step S20: dividing the data set into a training set, a validation set, and a test set based on a preset ratio;
[0010] Step S30: creating a distillation monitoring model based on a neural network;
[0011] Step S40: training the distillation monitoring model using the training set, verifying the trained distillation monitoring model using the validation set, and testing the verified distillation monitoring model using the test set;
[0012] Step S50: Acquire real-time monitoring data of the distillation system, input the real-time monitoring data into the tested distillation monitoring model, and obtain a distillation monitoring report;
[0013] Step S60: Display the distillation monitoring report in real time, and push the distillation monitoring report to the associated management terminal in real time.
[0014] Furthermore, the step S10 is specifically as follows:
[0015] Log into the database using the database address, database port, user name, and password, read historical monitoring data of the distillation system from the database, including at least temperature, flow, pressure, liquid level, pH, and distillation duration, and decrypt the read historical monitoring data using a national secret algorithm;
[0016] The historical monitoring data are pre-processed to standardize the format, and the pre-processed historical monitoring data are annotated with at least anomaly annotations and processing suggestions to construct a data set.
[0017] Furthermore, the step S20 is specifically as follows:
[0018] Grouping the historical monitoring data based on the distillation time, and dividing the grouped data sets into a training set, a validation set, and a test set based on a preset ratio of 7:2:1;
[0019] The step S30 is specifically as follows:
[0020] A distillation monitoring model for predicting distillation anomalies is created based on neural networks.
[0021] Furthermore, the step S40 specifically includes:
[0022] Step S41: training the distillation monitoring model using the training set until a preset convergence condition is met; during the training process, continuously optimizing the loss function of the distillation monitoring model and hyperparameters including at least a random dropout rate, a weight decay rate, and a learning rate;
[0023] Step S42: Validate the trained distillation monitoring model using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If so, proceed to step S43; if not, expand the training set and proceed to step S41.
[0024] Step S43: testing the verified distillation monitoring model using the test set.
[0025] Furthermore, in step S50, the real-time monitoring data includes at least temperature, flow, pressure, liquid level, pH and distillation time;
[0026] The step S60 is specifically as follows:
[0027] Displaying the distillation monitoring report in real time on a display screen, encrypting the distillation monitoring report using an AES algorithm and then pushing it to a pre-associated management terminal in real time;
[0028] A hash value is obtained by hashing the distillation monitoring report, and a symmetric key is randomly generated. The distillation monitoring report is encrypted using the symmetric key and stored in the IPFS system. The hash value and the index address returned by the IPFS system are bound and synchronized to the blockchain in real time.
[0029] In a second aspect, the present invention provides a distillation monitoring system based on a neural network, comprising the following modules:
[0030] A data set construction module is used to obtain historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH value and distillation time, and to construct a data set after preprocessing each of the historical monitoring data;
[0031] A data set partitioning module is used to partition the data set into a training set, a validation set, and a test set based on a preset ratio;
[0032] A distillation monitoring model creation module is used to create a distillation monitoring model based on a neural network;
[0033] a distillation monitoring model training module, configured to train the distillation monitoring model using the training set, verify the trained distillation monitoring model using the validation set, and test the verified distillation monitoring model using the test set;
[0034] A distillation monitoring module is used to obtain real-time monitoring data of the distillation system, input the real-time monitoring data into the tested distillation monitoring model, and obtain a distillation monitoring report;
[0035] The report push module is used to display the distillation monitoring report in real time and push the distillation monitoring report to the associated management terminal in real time.
[0036] Furthermore, the dataset construction module is specifically used to:
[0037] Log into the database using the database address, database port, user name, and password, read historical monitoring data of the distillation system from the database, including at least temperature, flow, pressure, liquid level, pH, and distillation duration, and decrypt the read historical monitoring data using a national secret algorithm;
[0038] The historical monitoring data are pre-processed to standardize the format, and the pre-processed historical monitoring data are annotated with at least anomaly annotations and processing suggestions to construct a data set.
[0039] Furthermore, the data set partitioning module is specifically used to:
[0040] Grouping the historical monitoring data based on the distillation time, and dividing the grouped data sets into a training set, a validation set, and a test set based on a preset ratio of 7:2:1;
[0041] The distillation monitoring model creation module is specifically used to:
[0042] A distillation monitoring model for predicting distillation anomalies is created based on neural networks.
[0043] Furthermore, the distillation monitoring model training module specifically includes:
[0044] A model training unit is configured to train the distillation monitoring model using the training set until a preset convergence condition is met; during the training process, the loss function of the distillation monitoring model and hyperparameters including at least a random dropout rate, a weight decay rate, and a learning rate are continuously optimized;
[0045] A model verification unit is used to verify the trained distillation monitoring model using the verification set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If so, the model test unit is entered; if not, the training set is expanded and the model training unit is entered;
[0046] The model testing unit is used to test the verified distillation monitoring model using the test set.
[0047] Furthermore, in the distillation monitoring module, the real-time monitoring data includes at least temperature value, flow value, pressure value, liquid level value, pH value and distillation time;
[0048] The report push module is specifically used for:
[0049] Displaying the distillation monitoring report in real time on a display screen, encrypting the distillation monitoring report using an AES algorithm and then pushing it to a pre-associated management terminal in real time;
[0050] A hash value is obtained by hashing the distillation monitoring report, and a symmetric key is randomly generated. The distillation monitoring report is encrypted using the symmetric key and stored in the IPFS system. The hash value and the index address returned by the IPFS system are bound and synchronized to the blockchain in real time.
[0051] The advantages of the present invention are:
[0052] 1. Obtain historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH and distillation time, perform pre-processing on each historical monitoring data in a standardized format, annotate each pre-processed historical monitoring data with at least anomaly annotation and processing suggestions, and then construct a data set. The data set is divided into a training set, a validation set and a test set based on a preset ratio; then, the training set is used to train a distillation monitoring model created based on a neural network, the validation set is used to validate the trained distillation monitoring model, and the test set is used to test the validated distillation monitoring model; finally, Then, the real-time monitoring data of the distillation system is obtained and input into the distillation monitoring model after the test to obtain the distillation monitoring report, which is displayed in real time and pushed to the associated management terminal; that is, the distillation monitoring model created based on the neural network is trained through the data set constructed by historical monitoring data, and the real-time monitoring data can be monitored by the distillation monitoring model subsequently. When the data changes, the impending anomaly can be predicted and processing suggestions can be given, so that the distillation system can be repaired or the process adjusted in time, and finally the distillation anomaly of the electronic waste liquid can be predicted, which greatly guarantees the distillation quality of the electronic waste liquid.
[0053] 2. The distillation monitoring report is encrypted and stored using a randomly generated symmetric key to prevent the distillation monitoring report from being stolen in plain text. The hash value obtained by hashing the distillation monitoring report is synchronized to the blockchain to prevent the hash value from being tampered with. The hash value stored in the blockchain can be used to quickly verify whether the distillation monitoring report is complete and whether it has been tampered with, which greatly improves the security and reliability of the distillation monitoring report management. The historical monitoring data is encrypted in combination with the national secret algorithm, and the distillation monitoring report pushed to the management terminal is encrypted using the AES algorithm, which greatly ensures data security.
[0054] 3. By continuously optimizing the loss function of the distillation monitoring model and its hyperparameters, including at least the random deactivation rate, weight decay rate, and learning rate, during the distillation monitoring model training process, the prediction accuracy of the distillation monitoring model is greatly improved, so that anomalies can be handled in a timely manner, thereby greatly ensuring the distillation quality of electronic waste liquid. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0056] Figure 1 The present invention is a flowchart of a distillation monitoring method based on a neural network.
[0057] Figure 2 It is a structural schematic diagram of a distillation monitoring system based on neural network of the present invention. DETAILED DESCRIPTION
[0058] The technical solution in the embodiments of the present application has the following overall idea: a distillation monitoring model created based on a neural network is trained through a data set constructed through historical monitoring data, and then the real-time monitoring data is monitored through the trained distillation monitoring model. When the data changes, the impending anomaly can be predicted and processing suggestions can be given, so that the distillation system can be repaired or the process adjusted in a timely manner, that is, the distillation anomaly of the electronic waste liquid can be predicted to ensure the distillation quality of the electronic waste liquid.
[0059] Please refer to Figures 1 to 2 As shown, a preferred embodiment of a distillation monitoring method based on a neural network of the present invention includes the following steps:
[0060] Step S10: Acquire historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH, and distillation time, and construct a data set after preprocessing each of the historical monitoring data;
[0061] Step S20: dividing the data set into a training set, a validation set, and a test set based on a preset ratio;
[0062] Step S30: creating a distillation monitoring model based on a neural network;
[0063] Step S40: training the distillation monitoring model using the training set, verifying the trained distillation monitoring model using the validation set, and testing the verified distillation monitoring model using the test set;
[0064] Step S50: Acquire real-time monitoring data of the distillation system, input the real-time monitoring data into the tested distillation monitoring model, and obtain a distillation monitoring report;
[0065] Step S60: Display the distillation monitoring report in real time, and push the distillation monitoring report to the associated management terminal in real time.
[0066] The step S10 is specifically as follows:
[0067] Log into the database using the database address, database port, user name, and password, read historical monitoring data of the distillation system from the database, including at least temperature, flow, pressure, liquid level, pH, and distillation duration, and decrypt the read historical monitoring data using a national secret algorithm;
[0068] The temperature value, flow value, pressure value, liquid level value and pH value are collected from the distillation system by a temperature sensor, a flow sensor, a pressure sensor, a liquid level sensor and a pH sensor respectively;
[0069] The historical monitoring data are pre-processed to standardize the format, and the pre-processed historical monitoring data are annotated with at least anomaly annotations and processing suggestions to construct a data set.
[0070] The step S20 is specifically as follows:
[0071] Grouping the historical monitoring data based on the distillation time, and dividing the grouped data sets into a training set, a validation set, and a test set based on a preset ratio of 7:2:1;
[0072] The step S30 is specifically as follows:
[0073] A distillation monitoring model for predicting distillation anomalies is created based on neural networks.
[0074] The distillation monitoring model includes a vector conversion module (Embedding), a self-attention module, a sampling module (Mean pooling & Max Pooling), and a linear projection module (Linera); the vector conversion module, the self-attention module, the sampling module, and the linear projection module are connected in sequence;
[0075] The vector conversion module is used to convert the data in the data set into a feature vector;
[0076] The self-attention module includes three Transformer layers. By adopting the three Transformer layers, the expressive power of the distillation monitoring model is greatly improved.
[0077] The sampling module is used to perform average sampling and maximum sampling on the output of the self-attention module;
[0078] The linear projection module is used to perform linear calculation on the output of the sampling module, thereby obtaining the predicted value of the distillation monitoring model.
[0079] The mathematical representation of the Transformer layer is:
[0080]
[0081] Z2=FFN(Z)=max(0,ZW1+b1)W2+b2;
[0082] Among them, Q, K, V are the same matrix with shape [3,128]; d k is the arithmetic square root of 128; FFN() is a linear projection layer, which eventually converts the [500,128]-shaped matrix into a [500,256]-shaped matrix; Z represents the output value of the Attention function, which is a [500,128]-shaped matrix; Z2 represents the output value of the FFN function, which is a [500,256]-shaped matrix; W1 represents parameter matrix 1, with a shape of [128,256]; W2 represents parameter matrix 2, with a shape of [128,256]; b1 represents bias vector 1 with a length of 500; b2 represents bias vector 2 with a length of 500;
[0083] The mathematical representation of the adoption module is:
[0084]
[0085]
[0086] in, represents the convolution window; x kpq Represents the original value in the convolution window; the matrix obtained by this formula is concatenated and finally converted into a matrix shape of [500,128];
[0087] The mathematical representation of the linear projection module is:
[0088] Y=LN(Z3)=Z3W+b;
[0089] The data shape sampled by the upper layer is [500,128], and then the data is stretched into a 1-dimensional vector by straightening, and finally converted into a matrix shape of [1].
[0090] The step S40 specifically includes:
[0091] Step S41: training the distillation monitoring model using the training set until a preset convergence condition is met; during the training process, continuously optimizing the loss function of the distillation monitoring model and hyperparameters including at least a random dropout rate, a weight decay rate, and a learning rate;
[0092] By continuously optimizing the loss function of the distillation monitoring model and the hyperparameters including at least the random deactivation rate, the weight decay rate, and the learning rate during the training process of the distillation monitoring model, the prediction accuracy of the distillation monitoring model is greatly improved, so that anomalies can be handled in a timely manner, thereby greatly ensuring the distillation quality of electronic waste liquid.
[0093] Step S42: Validate the trained distillation monitoring model using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If so, proceed to step S43; if not, expand the training set and proceed to step S41.
[0094] Step S43: testing the verified distillation monitoring model using the test set.
[0095] In step S50, the real-time monitoring data includes at least temperature, flow, pressure, liquid level, pH, and distillation time; that is, the real-time monitoring data and the historical monitoring data contain the same data content;
[0096] The step S60 is specifically as follows:
[0097] The distillation monitoring report is displayed in real time on a display screen, and the distillation monitoring report is encrypted using an AES algorithm and pushed to a pre-associated management terminal in real time, that is, an alarm is pushed in time for warning;
[0098] A hash value is obtained by hashing the distillation monitoring report, and a symmetric key is randomly generated. The distillation monitoring report is encrypted using the symmetric key and stored in the IPFS system. The hash value and the index address returned by the IPFS system are bound and synchronized to the blockchain in real time.
[0099] Since hash calculation is irreversible, the distillation monitoring report is hashed again subsequently. By comparing the calculated hash value with the hash value stored in the blockchain, it is possible to quickly determine whether the distillation monitoring report has been tampered with. The hash value and index address are notarized through the blockchain to prevent them from being tampered with. The distillation monitoring report is hash-checked using a trusted hash value to further ensure security.
[0100] A preferred embodiment of a distillation monitoring system based on a neural network of the present invention includes the following modules:
[0101] A data set construction module is used to obtain historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH value and distillation time, and to construct a data set after preprocessing each of the historical monitoring data;
[0102] A data set partitioning module is used to partition the data set into a training set, a validation set, and a test set based on a preset ratio;
[0103] A distillation monitoring model creation module is used to create a distillation monitoring model based on a neural network;
[0104] a distillation monitoring model training module, configured to train the distillation monitoring model using the training set, verify the trained distillation monitoring model using the validation set, and test the verified distillation monitoring model using the test set;
[0105] A distillation monitoring module is used to obtain real-time monitoring data of the distillation system, input the real-time monitoring data into the tested distillation monitoring model, and obtain a distillation monitoring report;
[0106] The report push module is used to display the distillation monitoring report in real time and push the distillation monitoring report to the associated management terminal in real time.
[0107] The dataset construction module is specifically used for:
[0108] Log into the database using the database address, database port, user name, and password, read historical monitoring data of the distillation system from the database, including at least temperature, flow, pressure, liquid level, pH, and distillation duration, and decrypt the read historical monitoring data using a national secret algorithm;
[0109] The temperature value, flow value, pressure value, liquid level value and pH value are collected from the distillation system by a temperature sensor, a flow sensor, a pressure sensor, a liquid level sensor and a pH sensor respectively;
[0110] The historical monitoring data are pre-processed to standardize the format, and the pre-processed historical monitoring data are annotated with at least anomaly annotations and processing suggestions to construct a data set.
[0111] The data set partitioning module is specifically used for:
[0112] Grouping the historical monitoring data based on the distillation time, and dividing the grouped data sets into a training set, a validation set, and a test set based on a preset ratio of 7:2:1;
[0113] The distillation monitoring model creation module is specifically used to:
[0114] A distillation monitoring model for predicting distillation anomalies is created based on neural networks.
[0115] The distillation monitoring model includes a vector conversion module (Embedding), a self-attention module, a sampling module (Mean pooling & Max Pooling), and a linear projection module (Linera); the vector conversion module, the self-attention module, the sampling module, and the linear projection module are connected in sequence;
[0116] The vector conversion module is used to convert the data in the data set into a feature vector;
[0117] The self-attention module includes three Transformer layers. By adopting the three Transformer layers, the expressive power of the distillation monitoring model is greatly improved.
[0118] The sampling module is used to perform average sampling and maximum sampling on the output of the self-attention module;
[0119] The linear projection module is used to perform linear calculation on the output of the sampling module, thereby obtaining the predicted value of the distillation monitoring model.
[0120] The mathematical representation of the Transformer layer is:
[0121]
[0122] Z2=FFN(Z)=max(0,ZW1+b1)W2+b2;
[0123] Among them, Q, K, V are the same matrix with shape [3,128]; d k is the arithmetic square root of 128; FFN() is a linear projection layer, which eventually converts the [500,128]-shaped matrix into a [500,256]-shaped matrix; Z represents the output value of the Attention function, which is a [500,128]-shaped matrix; Z2 represents the output value of the FFN function, which is a [500,256]-shaped matrix; W1 represents parameter matrix 1, with a shape of [128,256]; W2 represents parameter matrix 2, with a shape of [128,256]; b1 represents bias vector 1 with a length of 500; b2 represents bias vector 2 with a length of 500;
[0124] The mathematical representation of the adoption module is:
[0125]
[0126]
[0127] in, represents the convolution window; x kpq Represents the original value in the convolution window; the matrix obtained by this formula is concatenated and finally converted into a matrix shape of [500,128];
[0128] The mathematical representation of the linear projection module is:
[0129] Y=LN(Z3)=Z3W+b;
[0130] The data shape sampled by the upper layer is [500,128], and then the data is stretched into a 1-dimensional vector by straightening, and finally converted into a matrix shape of [1].
[0131] The distillation monitoring model training module specifically includes:
[0132] A model training unit is configured to train the distillation monitoring model using the training set until a preset convergence condition is met; during the training process, the loss function of the distillation monitoring model and hyperparameters including at least a random dropout rate, a weight decay rate, and a learning rate are continuously optimized;
[0133] By continuously optimizing the loss function of the distillation monitoring model and the hyperparameters including at least the random deactivation rate, the weight decay rate, and the learning rate during the training process of the distillation monitoring model, the prediction accuracy of the distillation monitoring model is greatly improved, so that anomalies can be handled in a timely manner, thereby greatly ensuring the distillation quality of electronic waste liquid.
[0134] A model verification unit is used to verify the trained distillation monitoring model using the verification set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If so, the model test unit is entered; if not, the training set is expanded and the model training unit is entered;
[0135] The model testing unit is used to test the verified distillation monitoring model using the test set.
[0136] In the distillation monitoring module, the real-time monitoring data includes at least temperature, flow, pressure, liquid level, pH and distillation time; that is, the real-time monitoring data and the historical monitoring data contain the same data content;
[0137] The report push module is specifically used for:
[0138] The distillation monitoring report is displayed in real time on a display screen, and the distillation monitoring report is encrypted using an AES algorithm and pushed to a pre-associated management terminal in real time, that is, an alarm is pushed in time for warning;
[0139] A hash value is obtained by hashing the distillation monitoring report, and a symmetric key is randomly generated. The distillation monitoring report is encrypted using the symmetric key and stored in the IPFS system. The hash value and the index address returned by the IPFS system are bound and synchronized to the blockchain in real time.
[0140] Since hash calculation is irreversible, the distillation monitoring report is hashed again subsequently. By comparing the calculated hash value with the hash value stored in the blockchain, it is possible to quickly determine whether the distillation monitoring report has been tampered with. The hash value and index address are notarized through the blockchain to prevent them from being tampered with. The distillation monitoring report is hash-checked using a trusted hash value to further ensure security.
[0141] In summary, the advantages of the present invention are:
[0142] 1. Obtain historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH and distillation time, perform pre-processing on each historical monitoring data in a standardized format, annotate each pre-processed historical monitoring data with at least anomaly annotation and processing suggestions, and then construct a data set. The data set is divided into a training set, a validation set and a test set based on a preset ratio; then, the training set is used to train a distillation monitoring model created based on a neural network, the validation set is used to validate the trained distillation monitoring model, and the test set is used to test the validated distillation monitoring model; finally, Then, the real-time monitoring data of the distillation system is obtained and input into the distillation monitoring model after the test to obtain the distillation monitoring report, which is displayed in real time and pushed to the associated management terminal; that is, the distillation monitoring model created based on the neural network is trained through the data set constructed by historical monitoring data, and the real-time monitoring data can be monitored by the distillation monitoring model subsequently. When the data changes, the impending anomaly can be predicted and processing suggestions can be given, so that the distillation system can be repaired or the process adjusted in time, and finally the distillation anomaly of the electronic waste liquid can be predicted, which greatly guarantees the distillation quality of the electronic waste liquid.
[0143] 2. The distillation monitoring report is encrypted and stored using a randomly generated symmetric key to prevent the distillation monitoring report from being stolen in plain text. The hash value obtained by hashing the distillation monitoring report is synchronized to the blockchain to prevent the hash value from being tampered with. The hash value stored in the blockchain can be used to quickly verify whether the distillation monitoring report is complete and whether it has been tampered with, which greatly improves the security and reliability of the distillation monitoring report management. The historical monitoring data is encrypted in combination with the national secret algorithm, and the distillation monitoring report pushed to the management terminal is encrypted using the AES algorithm, which greatly ensures data security.
[0144] 3. By continuously optimizing the loss function of the distillation monitoring model and its hyperparameters, including at least the random deactivation rate, weight decay rate, and learning rate, during the distillation monitoring model training process, the prediction accuracy of the distillation monitoring model is greatly improved, so that anomalies can be handled in a timely manner, thereby greatly ensuring the distillation quality of electronic waste liquid.
[0145] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.
Claims
1. A distillation monitoring method based on a neural network, characterized in that: The steps include: Step S10: Acquire historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH, and distillation time, and construct a data set after preprocessing each of the historical monitoring data; Step S20: dividing the data set into a training set, a validation set, and a test set based on a preset ratio; Step S30: creating a distillation monitoring model based on a neural network; Step S40: training the distillation monitoring model using the training set, verifying the trained distillation monitoring model using the validation set, and testing the verified distillation monitoring model using the test set; Step S50: Acquire real-time monitoring data of the distillation system, input the real-time monitoring data into the tested distillation monitoring model, and obtain a distillation monitoring report; Step S60: Display the distillation monitoring report in real time, and push the distillation monitoring report to the associated management terminal in real time.
2. A distillation monitoring method based on a neural network according to claim 1, characterized in that: The step S10 is specifically as follows: Log into the database using the database address, database port, user name, and password, read historical monitoring data of the distillation system from the database, including at least temperature, flow, pressure, liquid level, pH, and distillation duration, and decrypt the read historical monitoring data using a national secret algorithm; The historical monitoring data are pre-processed to standardize the format, and the pre-processed historical monitoring data are annotated with at least anomaly annotations and processing suggestions to construct a data set.
3. The neural network-based distillation monitoring method according to claim 1, wherein: The step S20 is specifically as follows: Grouping the historical monitoring data based on the distillation time, and dividing the grouped data sets into a training set, a validation set, and a test set based on a preset ratio of 7:2:1; The step S30 is specifically as follows: A distillation monitoring model for predicting distillation anomalies is created based on neural networks.
4. The neural network-based distillation monitoring method according to claim 1, wherein: The step S40 specifically includes: Step S41: training the distillation monitoring model using the training set until a preset convergence condition is met; during the training process, continuously optimizing the loss function of the distillation monitoring model and hyperparameters including at least a random dropout rate, a weight decay rate, and a learning rate; Step S42: Validate the trained distillation monitoring model using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If so, proceed to step S43; if not, expand the training set and proceed to step S41. Step S43: testing the verified distillation monitoring model using the test set.
5. The neural network-based distillation monitoring method according to claim 1, wherein: In step S50, the real-time monitoring data includes at least temperature, flow, pressure, liquid level, pH and distillation time; The step S60 is specifically as follows: Displaying the distillation monitoring report in real time on a display screen, encrypting the distillation monitoring report using an AES algorithm and then pushing it to a pre-associated management terminal in real time; A hash value is obtained by hashing the distillation monitoring report, and a symmetric key is randomly generated. The distillation monitoring report is encrypted using the symmetric key and stored in the IPFS system. The hash value and the index address returned by the IPFS system are bound and synchronized to the blockchain in real time.
6. A distillation monitoring system based on a neural network, characterized in that: Includes the following modules: A data set construction module is used to obtain historical monitoring data of the distillation system, including at least temperature, flow, pressure, liquid level, pH value and distillation time, and to construct a data set after preprocessing each of the historical monitoring data; A data set partitioning module is used to partition the data set into a training set, a validation set, and a test set based on a preset ratio; A distillation monitoring model creation module is used to create a distillation monitoring model based on a neural network; a distillation monitoring model training module, configured to train the distillation monitoring model using the training set, verify the trained distillation monitoring model using the validation set, and test the verified distillation monitoring model using the test set; A distillation monitoring module is used to obtain real-time monitoring data of the distillation system, input the real-time monitoring data into the tested distillation monitoring model, and obtain a distillation monitoring report; The report push module is used to display the distillation monitoring report in real time and push the distillation monitoring report to the associated management terminal in real time.
7. A neural network-based distillation monitoring system according to claim 6, characterized in that: The dataset construction module is specifically used for: Log into the database using the database address, database port, user name, and password, read historical monitoring data of the distillation system from the database, including at least temperature, flow, pressure, liquid level, pH, and distillation duration, and decrypt the read historical monitoring data using a national secret algorithm; The historical monitoring data are pre-processed to standardize the format, and the pre-processed historical monitoring data are annotated with at least anomaly annotations and processing suggestions to construct a data set.
8. The neural network-based distillation monitoring system according to claim 6, wherein: The data set partitioning module is specifically used for: Grouping the historical monitoring data based on the distillation time, and dividing the grouped data sets into a training set, a validation set, and a test set based on a preset ratio of 7:2:1; The distillation monitoring model creation module is specifically used to: A distillation monitoring model for predicting distillation anomalies is created based on neural networks.
9. The neural network-based distillation monitoring system according to claim 6, wherein: The distillation monitoring model training module specifically includes: A model training unit is configured to train the distillation monitoring model using the training set until a preset convergence condition is met; during the training process, the loss function of the distillation monitoring model and hyperparameters including at least a random dropout rate, a weight decay rate, and a learning rate are continuously optimized; A model verification unit is used to verify the trained distillation monitoring model using the verification set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If so, the model test unit is entered; if not, the training set is expanded and the model training unit is entered; The model testing unit is used to test the verified distillation monitoring model using the test set.
10. The neural network-based distillation monitoring system according to claim 6, characterized in that: In the distillation monitoring module, the real-time monitoring data includes at least temperature value, flow value, pressure value, liquid level value, pH value and distillation time; The report push module is specifically used for: Displaying the distillation monitoring report in real time on a display screen, encrypting the distillation monitoring report using an AES algorithm and then pushing it to a pre-associated management terminal in real time; A hash value is obtained by hashing the distillation monitoring report, and a symmetric key is randomly generated. The distillation monitoring report is encrypted using the symmetric key and stored in the IPFS system. The hash value and the index address returned by the IPFS system are bound and synchronized to the blockchain in real time.
Citation Information
Patent Citations
Rectification process quality monitoring service adaptive evolution method based on SVDD and Agent
CN114493235A
Method for operating a technical system
EP3736518A1