A distributed computing engine optimization control method based on artificial intelligence
By optimizing the control method using an AI-based distributed computing engine and optimizing the Long Short-Term Memory (LSTM) network using stochastic gradient descent, the problems of lag and manual intervention in traditional control systems are solved, achieving efficient and stable automated production control that adapts to changes in the production environment and market demands.
Patent Information
- Application Number
- CN202411044571.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-08-01
AI Technical Summary
Traditional control systems in factories suffer from lag, reliance on manual intervention, low production efficiency, high costs, inflexibility, and data silos, making them unable to adapt to changes in the production environment and market demands in a timely manner.
An AI-based distributed computing engine is used to optimize the control method. By collecting production data, the stochastic gradient descent method is used to optimize the Long Short-Term Memory (LSTM) network, and production parameters are adjusted in real time to achieve automated control and optimization.
It improves the precision and stability of the production process, reduces human intervention, lowers costs, and can flexibly adapt to changes in the production environment and market demand, supporting rapid conversion of production tasks and introduction of new products.
Smart Images

Figure FDA0005046226400000012 
Figure FDA0005046226400000014 
Figure FDA0005046226400000021
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of computer application technology, and specifically relates to an optimization control method for a distributed computing power engine based on artificial intelligence. Background Art
[0002] Traditional control systems have a certain lag in factories and cannot meet the production conditions of high quality, high efficiency and low power consumption under today's production methods. These shortcomings may limit the performance of enterprises in a highly competitive market environment. Traditional control systems usually rely on human intervention and manual adjustment, lacking real-time data and response speed. This makes them unable to adapt to changes in the production environment in a timely manner, resulting in low production efficiency and rising production costs. At the same time, the production line is inflexible, and traditional control systems are often rigidly designed and difficult to flexibly adjust to changes in market demand. This means that the production line may not be able to quickly switch production tasks or adapt to the requirements of new products, thus missing market opportunities. Traditional control systems are subject to human intervention and operating skills, so production quality is often difficult to guarantee. Due to the influence of human factors, products may have a high defect rate, resulting in reduced customer satisfaction and increased after-sales service costs. Information islands and data management difficulties, traditional control systems are usually unable to effectively integrate and analyze production data, leading to the problem of information islands. Lack of in-depth understanding and analysis of production data may cause companies to miss opportunities to optimize production processes.
[0003] Compared with traditional control systems, distributed computing engine systems have higher accuracy and stability, real-time optimization capabilities, higher production efficiency and quality, lower costs and energy consumption, and better adaptability and flexibility, thus bringing greater economic benefits to enterprises. Distributed computing engine systems use advanced algorithms and real-time data to monitor and adjust parameters in the production process, thereby achieving higher accuracy and stability. Compared with traditional manual or simple automatic control, distributed computing engine systems can control the production process more accurately and reduce fluctuations and changes. Distributed computing engine systems can optimize and adjust according to production data and goals in real time. It can quickly respond to changes in the production environment and take appropriate measures to maintain high efficiency and stability of production. Traditional control systems usually require manual intervention and adjustment, while distributed computing engine systems can automate these processes. Summary of the invention
[0004] In view of the technical problems existing in the above-mentioned traditional control system in the factory, the present invention proposes a distributed computing power engine optimization control method with a simple method and convenient operation. In order to achieve the above-mentioned purpose, the present invention proposes a distributed computing power engine optimization control method based on artificial intelligence, comprising the following steps:
[0005] S1. Collect historical production data from different production lines to provide a basis for model training and verification;
[0006] S2, using stochastic gradient descent method SGD to optimize the long short-term memory network LSTM to establish a multivariate model, based on the current production data and the established multivariate model, predict the output trajectory at the future moment;
[0007] S3. According to the prediction results of the optimized model, adjust the production parameters to adapt to the current production needs;
[0008] S4. Repeat the above steps to conduct feedback correction in real time to ensure that the production process is in the optimal state;
[0009] The specific steps of step S2 using stochastic gradient descent method SGD to optimize long short-term memory network LSTM are as follows:
[0010] S21, first is the forward propagation, the total time step number T of the input sequence {x1,x2,...,x T} Input into the LSTM network; calculate the hidden state and cell state at each time step: f t =σ(W f x t +U f h t-1 +b f ),i t =σ(W i x t +U i h t-1 +b i ), o t =σ(W o x t +U o h t-1 +b o ), g t =tanh(W g x t +U g h t -1+b g ), c t =f t ⊙c t-1 +i t ⊙g t ,h t =o t ⊙tanh(c t ), where f is the forget gate, which controls the cell state c at the previous moment t-1 The degree of information retention in; i is the input gate, which controls the new information g t Add to cell state c to is the output gate, which controls the cell state c t The output h after passing through the tanh function t g is the candidate cell state, the candidate update value, which is to be added to the cell state c t The information in; c is the cell state, the main memory part of LSTM, responsible for transmitting long-term dependency information; h is the hidden state, the output of LSTM, containing the information at the current moment; σ is the activation function, ⊙ is the element product, W, U, b are the weight matrix and bias vector of the network; calculate the final output Where W y , b y is the weight matrix and bias vector of the output layer;
[0011] S22. Then define the loss function and use the mean square error as the loss function: Where N is the number of samples, y l is the actual value, is the predicted value;
[0012] S23, then perform back propagation and calculate the gradient of the loss function to the output layer parameters: Calculate the error term δ at each time step t , backpropagating from the output layer to the hidden state and cell state at each time step, the gradient calculation for the current time step t is:
[0013] Calculate the gradient of the loss function with respect to the weights and biases:
[0014] S24, last updated parameters: Where θ is the parameter described above including W f , U f , b f , W i , U i , b i , W o , U o , b o , W g , U g , b g , η is the learning rate, which controls the step size of parameter update.
[0015] Preferably, the production data includes the deployment of workers, allocation of resources, and speed and temperature parameters of production equipment.
[0016] Preferably, the specific operation of adjusting the production parameters in step S3 is as follows:
[0017] S31, firstly calculate the new production parameters, calculate the production parameters that need to be adjusted according to the prediction results output by the model, and pass the calculated production parameters to the production line control system for actual adjustment;
[0018] S32, then adjust the production line, adjust the machine's operating speed, temperature setting, pressure production parameters, monitor the real-time status of the production line, and ensure that the adjusted production parameters can improve production efficiency and product quality;
[0019] S33, then calculate the prediction accuracy and error rate indicators of the model, record and analyze the performance of the model under different production conditions, and identify the weaknesses and improvement points of the model;
[0020] S34. Finally, continuously expand the training data set to include the latest production data, use the new training data and optimization algorithm to retrain the model, update the model parameters, and deploy the updated model to the production system to ensure that it can better adapt to changes in the production environment.
[0021] Preferably, the specific implementation of step S4 is that in each production cycle, the system first collects current production data and inputs it into a trained long short-term memory network LSTM model; the model predicts based on the input data and outputs future production parameter adjustment suggestions; based on these suggestions, the system adjusts the operating parameters of the production line, the speed, temperature and pressure of the equipment in real time; after the adjustment is completed, the system continues to monitor the operating status of the production line and collect new production data; these data are not only used for immediate feedback correction, but also continuously enrich and update the training data set; the system further improves the prediction accuracy and response speed of the model through continuous learning and optimization of these real-time data.
[0022] Compared with the prior art, the advantages and positive effects of the present invention are:
[0023] 1. The present invention adopts the stochastic gradient descent method SGD to optimize the long short-term memory network LSTM, which can provide higher prediction accuracy and production control stability compared with traditional control systems.
[0024] 2. Traditional control systems rely on human intervention and manual adjustment, while the present invention realizes automated production control through artificial intelligence technology, reduces the need for human intervention, improves production efficiency and reduces production costs.
[0025] 3. The present invention can flexibly adapt to changes in the production environment and market demand. By continuously expanding and updating the training data set, the model can continuously learn and optimize, adapt to different production conditions, and support the rapid conversion of production tasks and the rapid introduction of new products. DETAILED DESCRIPTION
[0026] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0027] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.
[0028] Embodiment, traditional control systems have a certain lag in factories, usually relying on human intervention and manual adjustment, lacking real-time data and response speed. This makes them unable to adapt to changes in the production environment in a timely manner, resulting in low production efficiency and rising production costs. The present invention proposes a distributed computing power engine optimization control method based on artificial intelligence, which uses a distributed computing power engine method to optimize and adjust according to production data and goals to improve the efficiency and stability of production. First of all, data collection is required to collect actual production data and historical production data on different production lines to provide a basis for model training and verification. The collected data described in the present invention includes various configurations of the production line, the deployment of workers, the allocation of resources, and the speed and temperature parameter data of production equipment. Collecting data helps model training and improves prediction accuracy and production optimization effects.
[0029] Considering the need to complete the control of process parameters in real time, the traditional long short-term memory network LSTM has low efficiency in processing large-scale data and does not respond quickly enough in real-time applications. The present invention uses the stochastic gradient descent method SGD to optimize the long short-term memory network LSTM. The stochastic gradient descent method SGD is a gradient-based optimization algorithm used to find the parameter configuration that minimizes the loss function. The stochastic gradient descent method SGD updates the parameters by calculating the gradient of each sample, and calculates the gradient of the loss function to the network parameters through the gradient descent algorithm, and gradually updates the parameters for optimization. This innovative method can optimize production parameters in real time, reduce fluctuations and changes, and is more stable than the traditional long short-term memory network LSTM. The specific operation is to first forward propagate and convert the input sequence {x1, x2,..., x} with a total time step of T. T} Input into the long short-term memory network LSTM; calculate the hidden state and cell state at each time step: f t =σ(W f x t +U f h t-1 +b f ),i t =σ(W i x t +U i ht-1 +b i ), o t =σ(W o x t +U o h t-1 +b o ), g t =tanh(W g x t +U g h t -1+b g ), c t =f t ⊙c t-1 +i t ⊙g t ,h t =o t ⊙tanh(c t ), where f is the forget gate, which controls the cell state c at the previous moment t-1 The degree of information retention in; i is the input gate, which controls the new information g t Add to cell state c t o is the output gate, which controls the cell state c t The output h after passing through the tanh function t g is the candidate cell state, the candidate update value, which is to be added to the cell state c t The information in; c is the cell state, the main memory part of LSTM, responsible for transmitting long-term dependency information; h is the hidden state, the output of LSTM, containing the information at the current moment; σ is the activation function, ⊙ is the element product, W, U, b are the weight matrix and bias vector of the network; calculate the final output Where W y , b y The weight matrix and bias vector of the output layer; define the loss function and use the mean square error as the loss function: Where N is the number of samples, y l is the actual value, is the predicted value; secondly, back propagation is performed to calculate the gradient of the loss function to the output layer parameters Calculate the error term δ at each time step t , backpropagating from the output layer to the hidden state and cell state at each time step, the gradient calculation for the current time step t is:
[0030] Calculate the gradient of the loss function with respect to the weights and biases:
[0031] Last updated parameters: Where θ is the parameter described above including W f , U f , b f , W i , U i , b i , W o , U o , b o , W g , U g , b g , η is the learning rate, which controls the step size of parameter update.
[0032] Taking into account the real-time update of production process parameters, after optimizing the long short-term memory network LSTM by stochastic gradient descent method SGD, the calculated parameters are applied to the model to perform the model work. First, the new production parameters are calculated, and the production parameters that need to be adjusted are calculated according to the prediction results output by the model. The calculated production parameters are passed to the production line control system for actual adjustment; then the production line is adjusted to adjust the machine's operating speed, temperature setting, and pressure production parameters, and the real-time status of the production line is monitored to ensure that the adjusted production parameters can improve production efficiency and product quality; then the prediction accuracy and error rate indicators of the model are calculated, and the performance of the model under different production conditions is recorded and analyzed to identify the model's weaknesses and improvement points; finally, the training data set is continuously expanded to include the latest production data, and the model is retrained using new training data and optimization algorithms, and the model parameters are updated. The updated model is deployed to the production system to ensure that it can better adapt to changes in the production environment. Adjusting production parameters according to the prediction results of the optimized model can respond to production needs in real time and improve production efficiency and product quality.
[0033] Finally, it is a core advantage of the present invention to repeat the above steps to perform feedback correction in real time and ensure that the production process is in the optimal state. Specifically, in each production cycle, the system first collects the current production data and inputs it into the trained long short-term memory network LSTM model; the model predicts based on the input data and outputs future production parameter adjustment suggestions; based on these suggestions, the system adjusts the operating parameters of the production line in real time, such as the speed, temperature and pressure of the equipment; after the adjustment is completed, the system continues to monitor the operating status of the production line and collects new production data; these data are not only used for immediate feedback correction, but also continuously enrich and update the training data set; the system further improves the prediction accuracy and response speed of the model through the continuous learning and optimization of these real-time data. Each feedback correction will make the production parameters more accurate and the production process more stable, thereby significantly improving production efficiency and product quality. This real-time feedback correction mechanism can quickly respond to changes in the production environment and avoid waste of resources and production delays caused by improper production parameters. Through continuous cycle optimization, the system ensures that the production process is always in the optimal state and realizes efficient, stable and intelligent production management.
[0034] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.
Claims
1. A distributed computing engine optimization control method based on artificial intelligence, characterized in that: The following steps are involved: S1. Collect historical production data from different production lines to provide a basis for model training and verification; S2, using stochastic gradient descent method SGD to optimize the long short-term memory network LSTM to establish a multivariate model, based on the current production data and the established multivariate model, predict the output trajectory at the future moment; S3. According to the prediction results of the optimized model, adjust the production parameters to adapt to the current production needs; S4. Repeat the above steps to conduct feedback correction in real time to ensure that the production process is in the optimal state; The specific steps of step S2 using stochastic gradient descent method SGD to optimize long short-term memory network LSTM are as follows: S21, first is the forward propagation, the total time step number T of the input sequence {x1,x2,...,x T } Input to the LSTM network; calculate the hidden state and cell state at each time step: f t =σ(W f x t +U f h t-1 +b f ),i t =σ(W i x t +U i h t-1 +b i ), o t =σ(W o x t +U o h t-1 +b o ), g t =tanh(W g x t +U g h t -1+b g ), c t =f t ⊙c t-1 +i t ⊙g t ,h t =o t ⊙tanh(c t ), where f is the forget gate, which controls the cell state c at the previous moment t-1 The degree of information retention in; i is the input gate, which controls the new information g t Add to cell state c t o is the output gate, which controls the cell state c t The output h after passing through the tanh function t g is the candidate cell state, the candidate update value, which is to be added to the cell state c t The information in; c is the cell state, the main memory part of LSTM, responsible for transmitting long-term dependency information; h is the hidden state, the output of LSTM, which contains the information of the current moment; σ is the activation function, ⊙ is the element product, W, U, b are the weight matrix and bias vector of the network; calculate the final output Where W y , b y is the weight matrix and bias vector of the output layer; S22. Then define the loss function and use the mean square error as the loss function: Where N is the number of samples, y l is the actual value, is the predicted value; S23, then perform back propagation and calculate the gradient of the loss function to the output layer parameters: Calculate the error term δ at each time step t , backpropagating from the output layer to the hidden state and cell state at each time step, the gradient calculation for the current time step t is: Calculate the gradient of the loss function with respect to the weights and biases: S24, last updated parameters: Where θ is the parameter described above including W f , U f , b f , W i , U i , b i , W o , U o , b o , W g , U g , b g , η is the learning rate, which controls the step size of parameter update.
2. The distributed computing engine optimization control method based on artificial intelligence according to claim 1 is characterized in that: The production data includes the deployment of workers, allocation of resources, and speed and temperature parameters of production equipment.
3. The distributed computing engine optimization control method based on artificial intelligence according to claim 1 is characterized in that: The specific operation of adjusting the production parameters in step S3 is as follows: S31, firstly calculate the new production parameters, calculate the production parameters that need to be adjusted according to the prediction results output by the model, and pass the calculated production parameters to the production line control system for actual adjustment; S32, then adjust the production line, adjust the machine's operating speed, temperature setting, pressure production parameters, monitor the real-time status of the production line, and ensure that the adjusted production parameters can improve production efficiency and product quality; S33, then calculate the prediction accuracy and error rate indicators of the model, record and analyze the performance of the model under different production conditions, and identify the weaknesses and improvement points of the model; S34. Finally, continuously expand the training data set to include the latest production data, use the new training data and optimization algorithm to retrain the model, update the model parameters, and deploy the updated model to the production system to ensure that it can better adapt to changes in the production environment.
4. The distributed computing engine optimization control method based on artificial intelligence according to claim 1 is characterized in that: The specific implementation of step S4 is that in each production cycle, the system first collects the current production data and inputs it into the trained long short-term memory network LSTM model; the model predicts based on the input data and outputs future production parameter adjustment suggestions; Based on these suggestions, the system adjusts the operating parameters of the production line, the speed, temperature and pressure of the equipment in real time; after the adjustment is completed, the system continues to monitor the operating status of the production line and collect new production data; this data is not only used for immediate feedback correction, but also continuously enriches and updates the training data set; the system further improves the prediction accuracy and response speed of the model through continuous learning and optimization of these real-time data.
