Cutting fluid cooling effect optimization method and system based on intelligent algorithm

Abstract

The invention discloses a cutting fluid cooling effect optimization method and system based on an intelligent algorithm, and the method comprises the steps: constructing an intelligent prediction model through fusing a convolutional neural network and a gating circulation unit, dynamically adjusting hyper-parameters in combination with a Bayesian optimization algorithm, and achieving the high-precision prediction and optimization of the cutting fluid cooling effect. The prediction efficiency is remarkably improved, the calculation cost is reduced, the method can adapt to complex nonlinear machining environments and different working condition requirements, the problems that a traditional method depends on empirical formulas, data are difficult to obtain and the model universality is poor are effectively solved, meanwhile, cutting fluid formula optimization is guided through an accurate prediction result, and the prediction efficiency is improved. The machining efficiency, the workpiece quality and the resource utilization rate are remarkably improved, and reliable technical support can be provided for intelligent upgrading of an industrial cutting fluid system.

Description

Technical Field

[0001] This application relates to the technical field of cutting fluid optimization, and particularly to a method and system for optimizing the cooling effect of cutting fluid based on intelligent algorithms. Background Art

[0002] Cutting fluid plays a key role in machining. Its main functions are to improve machining efficiency and workpiece quality through cooling, lubrication, and cleaning. The optimization of the cooling effect of traditional cutting fluid usually relies on empirical formulas and experimental verification. Although these methods are simple and direct, they are often difficult to adapt to complex non-linear machining environments. In recent years, numerical simulation technology has been widely used in the research of the cooling effect of cutting fluid. By simulating the flow and heat transfer mechanisms of cutting fluid, its cooling mechanism can be deeply revealed. However, with the improvement of machining accuracy and complexity, the computational cost and time consumption of numerical simulation have gradually become the bottleneck restricting its large-scale application. At the same time, the development of artificial intelligence technology provides new solutions for processing multi-dimensional complex data in the cooling effect of cutting fluid. These methods have demonstrated powerful capabilities in modeling and prediction, and have gradually become a research hotspot in the field of cutting fluid optimization.

[0003] Despite the significant progress made in the above technologies, many challenges still exist in practical applications. First, it is difficult and costly to obtain measurement data on the cooling effect of cutting fluid. Especially in real industrial scenarios, precise acquisition of data such as the temperature field, flow rate, and heat flux density of cutting fluid requires complex experimental conditions. Second, the computational cost of deep learning models is high, and the training process has high requirements for hardware resources and time. In addition, the universality of the model is also a key issue. There are significant differences in the requirements for cutting fluid in different machining scenarios. For example, existing models often have difficulty meeting the need for stronger cooling effects. Summary of the Invention

[0004] Embodiments of this application provide a method and system for optimizing the cooling effect of cutting fluid based on intelligent algorithms, which can achieve high-precision prediction of the cooling effect of cutting fluid under different working conditions and optimize its formulation to meet specific machining requirements.

[0005] To achieve the above object, the technical solution of the embodiments of the present invention is as follows:

[0006] In a first aspect, embodiments of the present invention provide a method for optimizing the cooling effect of cutting fluid based on intelligent algorithms, including: collecting physical property data of cutting fluid and machining environment parameters, and constructing a multi-dimensional input feature matrix, where the physical property data at least includes thermal conductivity, convective heat transfer coefficient, density, and evaporation mass flow rate;

[0007] Construct an intelligent prediction model based on a Convolutional Neural Network (CNN) and a Gated Recurrent Unit (GRU). Among them, the CNN is used to extract multi-dimensional data from the feature matrix, reduce the dimension of the multi-dimensional data, and extract non-linear features. The CNN includes at least one 1D convolutional layer, a pooling layer, and a fully connected layer; the GRU is used to process time series data and capture the long-term dependence relationship of time series data in the prediction of the cutting fluid cooling effect;

[0008] Use the Bayesian optimization method to optimize the hyperparameters of the prediction model, including constructing a surrogate model through Gaussian process regression, and iteratively optimizing the hyperparameter combination by combining the expected improvement or probability improvement acquisition function to minimize the prediction error;

[0009] Predict the cutting fluid cooling effect according to the optimized model, and adjust the cutting fluid formula based on the prediction result to achieve dynamic optimization of the cooling effect.

[0010] In some possible implementation manners, the convolutional kernel size of the 1D convolutional layer of the CNN is 3, the number of convolutional kernels is 32, the activation function is ReLU, the output is downsampled by the max pooling layer and flattened into a one-dimensional vector, and then input into a fully connected layer containing 64 neurons.

[0011] In some possible implementation manners, the reset gate and update gate of the GRU are controlled by the Sigmoid activation function. The reset gate is used to control the degree of forgetting of the historical state, the update gate is used to fuse the current input and the historical hidden state, the candidate activation value is generated by the hyperbolic tangent function, and the hidden state update formula is expressed as:

[0012]

[0013] where h t is the hidden state at the current moment, z t is the output of the update gate, is the candidate activation value, and t is the current moment.

[0014] In some possible implementation manners, use the Bayesian optimization method to optimize the hyperparameters of the prediction model, including:

[0015] Initialize the hyperparameter space, and set the learning rate, the number of iterations, and the kernel function parameters;

[0016] Calculate the mean and variance of the objective function based on Gaussian process regression, and select the next set of hyperparameters to be evaluated through the acquisition function;

[0017] Use the mean square error as the loss function, and iteratively update the hyperparameters until convergence; the loss function is expressed as:

[0018]

[0019] Among them, f(θ) represents the loss function, θ is the initial hyperparameter, and Y i is the true cooling effect, is the cooling effect predicted by the model under the given hyperparameter θ, and m is the number of samples.

[0020] In some possible implementation manners, the intelligent prediction model is represented by the following formula:

[0021]

[0022] Among them, Q cooling represents the intelligent prediction model, k is the thermal conductivity of the cutting fluid, is the temperature gradient, h is the convective heat transfer coefficient between the cutting fluid and the workpiece surface, T is the temperature field, and T ambient is the temperature of the surrounding environment, m evap is the evaporation mass flow rate of the cutting fluid, ρ is the density of the cutting fluid, L is the latent heat of evaporation of the cutting fluid, and Ω is the volume of the cutting area.

[0023] In a second aspect, an embodiment of the present invention provides an optimization system for the cooling effect of a cutting fluid based on an intelligent algorithm, which is used to implement the method described in the first aspect. The system includes: a data acquisition module, which is used to acquire the physical property data and processing condition data of the cutting fluid;

[0024] a model construction module, which is used to build an intelligent prediction model based on a convolutional neural network (CNN) and a gated recurrent unit (GRU);

[0025] a hyperparameter optimization module, which is used to adjust the model hyperparameters by using a Bayesian optimization algorithm;

[0026] a prediction and optimization module, which is used to predict the cooling effect of the cutting fluid according to the optimized model and adjust the cutting fluid formula based on the prediction result.

[0027] In some possible implementation manners, in the model construction module, the output of the CNN is flattened and then input into a fully connected layer, and the hidden state of the GRU is updated iteratively through time steps.

[0028] In some possible implementation manners, the hyperparameter optimization module constructs a surrogate model through Gaussian process regression and dynamically selects the optimal hyperparameter combination based on expected improvement or probability improvement.

[0029] One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages:

[0030] In the embodiments of the present invention, an intelligent prediction model is constructed by integrating a convolutional neural network and a gated recurrent unit, and the hyperparameters are dynamically adjusted in combination with the Bayesian optimization algorithm, realizing high-precision prediction and optimization of the cooling effect of cutting fluid, significantly improving the prediction efficiency and reducing the calculation cost. It can adapt to complex non-linear machining environments and different working condition requirements, effectively solving the problems of traditional methods relying on empirical formulas, difficult data acquisition, and poor model universality. At the same time, through accurate prediction results to guide the optimization of cutting fluid formulations, the machining efficiency, workpiece quality, and resource utilization rate are significantly improved, providing reliable technical support for the intelligent upgrade of industrial cutting fluid systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] To more clearly illustrate the embodiments of the present invention, the drawings required for use in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 It is a schematic flowchart of an embodiment of a method for optimizing the cooling effect of cutting fluid based on an intelligent algorithm provided for the embodiments of the present invention;

[0033] Figure 2 It is a schematic structural diagram of a system for optimizing the cooling effect of cutting fluid based on an intelligent algorithm in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the scope of protection of the present application.

[0035] In the relevant descriptions of this embodiment, terms such as "including", "containing", "having", etc. are all open terms, generally preferably understood as including but not limited to; the term "at least one" is generally preferably understood as one or more, where "a plurality" means two or more; the term "at least one (item) of the following" or its similar expression refers to any combination of these items, including any combination of single item (item) or plural items (items). For example, "at least one (item) of a, b or c", or, "at least one (item) of a, b and c" can all represent: a, b, c, a - b (i.e., a and b), a - c, b - c, or a - b - c, where a, b, c can be single or multiple respectively; the symbol "A / B" is used to describe the selection relationship of associated objects, generally representing an "or" relationship before and after.

[0036] In the following description of this embodiment, the terms used in the embodiments of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a" and "the" used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise.

[0037] Those skilled in the art should understand that in the following description of the embodiments of this application, the sequence numbers do not imply the order of execution, and some or all of the steps can be executed in parallel or sequentially. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of this application.

[0038] Those skilled in the art should understand that the numerical ranges in the embodiments of this application should be understood to specifically disclose each intermediate value between the upper and lower limits of the range. Intermediate values within any stated value or range, as well as each smaller range between any other stated value or intermediate value within the range, are also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded from the range.

[0039] Unless otherwise specified, the technical / scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. Although this application only describes preferred methods and materials, any methods and materials similar or equivalent to those described herein can also be used in the implementation or testing of this application. All documents mentioned in this specification are incorporated by reference to disclose and describe the methods and / or materials related to the documents. In case of conflict with any incorporated document, the content of this specification shall prevail.

[0040] To illustrate the technical solutions of the present invention, specific embodiments are used for illustration below.

[0041] Cutting fluid plays a crucial role in machining. Its main function is to improve machining efficiency and workpiece quality through cooling, lubrication, and cleaning effects. The optimization of the cooling effect of traditional cutting fluid usually relies on empirical formulas and experimental verification. Although these methods are simple and direct, they are often difficult to adapt to complex non-linear machining environments. In recent years, numerical simulation technology has been widely used in the study of the cooling effect of cutting fluid. By simulating the flow and heat transfer mechanisms of cutting fluid, its cooling mechanism can be deeply revealed. However, with the improvement of machining accuracy and complexity, the computational cost and time consumption of numerical simulation have gradually become the bottleneck restricting its large-scale application. At the same time, the development of artificial intelligence technology provides new solutions for processing multi-dimensional complex data in the cooling effect of cutting fluid. These methods have demonstrated powerful capabilities in modeling and prediction and have gradually become a research hotspot in the field of cutting fluid optimization.

[0042] Despite the significant progress made in the above technologies, there are still many challenges in practical applications. First, it is difficult and costly to obtain measurement data on the cooling effect of cutting fluid. Especially in real industrial scenarios, precise acquisition of data such as the temperature field, flow rate, and heat flux density of cutting fluid requires complex experimental conditions. Second, the computational cost of deep learning models is high, and the training process has high requirements for hardware resources and time. In addition, the universality of the model is also a key issue. There are significant differences in the requirements for cutting fluid in different machining scenarios. For example, existing models often struggle to meet the need for stronger cooling effects.

[0043] Based on this, the embodiments of the present invention provide an optimization method and system for the cooling effect of cutting fluid based on intelligent algorithms, which can achieve high-precision prediction of the cooling effect of cutting fluid under different working conditions and optimize its formulation to meet specific machining requirements.

[0044] Figure 1 The flowchart of the embodiment of an optimization method for the cooling effect of cutting fluid based on intelligent algorithms provided for the implementation of the present invention is shown in Figure 1 As shown, the above method may include:

[0045] S101, collect the physical property data of the cutting fluid and the machining environment parameters, and construct a multi-dimensional input feature matrix;

[0046] Among them, the physical property data at least includes thermal conductivity, convective heat transfer coefficient, density, and evaporation mass flow rate;

[0047] Specifically, thermal conductivity is an index to measure the heat conduction ability of cutting fluid, which directly affects the heat transfer efficiency during the cutting process. Professional thermal conductivity measuring instruments, such as hot-wire thermal conductivity meters, can be used to measure cutting fluid under different formulations and working conditions. The convective heat transfer coefficient reflects the ability of cutting fluid to transfer heat to the tool and workpiece surfaces through convection. This coefficient is affected by various factors such as the flow rate, temperature of the cutting fluid, and the roughness of the tool and workpiece surfaces, and can be obtained through a combination of experimental measurement and theoretical calculation. Different densities of cutting fluid will result in differences in flow resistance and flow distribution in the circulation system. Usually, a densitometer can be used for measurement. The evaporation mass flow rate reflects the mass rate at which the cutting fluid is lost due to heat evaporation during the machining process. By setting up a special evaporation measurement device at the machining site, the mass change of the cutting fluid within a certain time can be monitored in real time, and thus the evaporation mass flow rate can be calculated.

[0048] In some embodiments, the machining environment parameters may include but are not limited to the temperature field, cutting speed, and shear stress and contact area of the cutting area.

[0049] The temperature field can include the temperature of the cutting area, the temperature of the surrounding environment, and the temperatures of the tool and workpiece surfaces. Temperature measuring devices such as infrared thermometers and thermocouples can be used to monitor the temperatures at different positions in real time. The cutting speed is a parameter that affects the heat generation and cooling requirements during the cutting process. The cutting speed data can be directly obtained through the control system of the machine tool or measured using a dedicated speed measuring instrument. The shear stress is related to the yield strength of the material, while the contact area depends on the geometry of the tool and the cutting parameters. The shear stress and contact area in the cutting area can be estimated through methods such as finite element simulation and mechanical analysis, combined with the measurement data in actual machining.

[0050] After the data collection is completed, a multi-dimensional input feature matrix can be constructed by organizing and integrating these data. For example, assume that m sample data under different cutting fluid formulations and working conditions are collected, and each sample contains the above-mentioned physical property data of the cutting fluid and the machining environment parameters. Then the dimension of the constructed multi-dimensional input feature matrix X is m×n (n is the total number of selected features). Each row represents a sample and contains all the feature data corresponding to that sample; each column represents a specific feature. For example, the first column represents the thermal conductivity, the second column represents the convective heat transfer coefficient, and so on. In this way, the multi-dimensional input feature matrix X can completely record the detailed information of each sample.

[0051] S102, construct an intelligent prediction model based on a convolutional neural network (CNN) and a gated recurrent unit (GRU). Among them, CNN is used to extract multi-dimensional data in the feature matrix, reduce the dimension of the multi-dimensional data and extract non-linear features. CNN includes at least one 1D convolutional layer, a pooling layer, and a fully connected layer; GRU is used to process time series data and capture the long-term dependencies of time series data in the prediction of the cutting fluid cooling effect;

[0052] It should be noted that in the embodiments of the present invention, the prediction problem of the cutting fluid cooling effect is transformed into an optimization problem, and a corresponding intelligent optimization strategy is designed. The designed model not only considers the basic physical mechanism of heat transfer but also needs to combine the complex interactions in the actual machining environment.

[0053] Based on this, assume that each factor can be described by a comprehensive intelligent prediction model. Specifically, the comprehensive intelligent prediction model of the cooling effect can be expressed as a mathematical expression that includes the coupling of multiple factors such as heat conduction, convective heat transfer, evaporation effect, and shear heat generation.

[0054] Exemplarily, in some embodiments, the intelligent prediction model can be expressed by the following formula:

[0055]

[0056] Where Q coolingTo represent the intelligent prediction model, k is the thermal conductivity of the cutting fluid (W·m -1 ·K -1 ), is the temperature gradient, h is the convective heat transfer coefficient between the cutting fluid and the workpiece surface (W·m -2 ·K -1 ), T is the temperature field (K), T ambient is the temperature of the surrounding environment (K), m evap is the mass flow rate of cutting fluid evaporation (kg / s), ρ is the density of the cutting fluid (kg / m 3 ), L is the latent heat of evaporation of the cutting fluid (J / kg), and Ω is the volume of the cutting area.

[0057] In some embodiments, in the above intelligent prediction model, the heat conduction term follows the following formula:

[0058]

[0059] where, is the infinitesimal change in temperature T, is the infinitesimal change in time t, α is the thermal diffusivity, defined as where c p is the specific heat capacity of the cutting fluid; is the Laplace operator of the temperature field.

[0060] The variation law of the convective heat transfer term is:

[0061] q conv = h*(T surface - T liquid );

[0062] where, T surface is the temperature of the tool or workpiece surface (K); T liquid is the temperature of the cutting fluid (K). Heat conduction and convective heat transfer both increase as the temperature difference increases.

[0063] The evaporation term follows the following formula:

[0064]

[0065] where, m evap is the mass flow rate of evaporation (kg / s). The evaporation effect is affected by heat conduction and convective heat transfer, and higher temperatures will accelerate the evaporation process.

[0066] where, the generation of shear heat is usually proportional to the yield strength of the material and the cutting speed. It can be expressed by the following formula:

[0067] Q shear = τ*A cut *vcut ;

[0068] where τ is the shear stress in the cutting area; A cut is the contact area in the cutting area; v cut is the cutting speed. The shear heat generation directly affects the local temperature field, and further affects heat conduction and convective heat transfer.

[0069] In some embodiments, the cutting fluid optimization problem is transformed into an optimization problem, and the introduced feature matrix is established. For example, the thermal conductivity k of the cutting fluid, the convective heat transfer coefficient h between the cutting fluid and the workpiece surface, the density ρ of the cutting fluid, and the evaporation mass flow rate m of the cutting fluid are selected evap as four features. The function between the input features and the cooling effect is obtained through mathematical fitting, expressed as:

[0070] Q cooling = f(k, h, p, m evap );

[0071] After that, the input feature matrix X of the cutting fluid cooling effect can be constructed.

[0072] Exemplarily, assuming there are m samples (different cutting fluid formulations and working conditions), the dimension of matrix X is m×4, each row represents a sample, and each column represents a feature. Matrix X can be expressed as:

[0073]

[0074] After that, the complex multi-dimensional data involved in the cutting process is extracted by CNN features to reduce the dimension and simplify the subsequent processing steps.

[0075] Exemplarily, in some embodiments, the convolution kernel size of the 1D convolution layer of CNN is 3 (window size is 3), the number of convolution kernels is 32, the activation function is ReLU, and the output is flattened into a one-dimensional vector after being downsampled by the max pooling layer and input into a fully connected layer containing 64 neurons.

[0076] The output calculation formula of the convolution layer is expressed as:

[0077]

[0078] where W1 is the weight of the first convolution kernel, with a dimension of 3×32; b1 is the bias term of the convolution layer, with a dimension of 32. The activation function ReLU(x) = max(0, x), which is used to increase the non-linear characteristics of the network. After the convolution operation, the dimension of the output C1 is m×2×32. Here, 2 is because the convolution kernel size is 3, and the data size becomes twice the original after the convolution operation.

[0079] Furthermore, the pooling layer is used for downsampling. The pooling window size is 2, and the pooling stride is 2, following the formula below:

[0080] P1 = Maxpool(C1, 2);

[0081] where P1 is the output feature map after the max pooling operation.

[0082] After pooling, the output dimension is m×1×32. After the pooling layer, the data is flattened into a one-dimensional vector, following the formula below:

[0083] P1 = Flatten(P1);

[0084] Furthermore, the output dimension becomes m×32. The flattened output is fed into the fully connected layer with 64 neurons, using the ReLU activation function, following the formula below:

[0085]

[0086] where W2 is the weight matrix of the fully connected layer with dimension 32×64; b2 is the bias term of the fully connected layer with dimension 64. The dimension of the output F1 after the fully connected layer is m×64.

[0087] Furthermore, a fully connected layer is used to generate the predicted value of the cooling effect, following the formula below:

[0088] Y = F1 * W3 + b3;

[0089] where W3 is the weight matrix of the output layer with dimension 64×1; b3 is the bias term of the output layer with dimension 1; Y is the predicted value of the cooling effect, and the dimension of Y is m×1.

[0090] The input feature matrix X (including the thermal conductivity k, convective heat transfer coefficient h, density ρ, and evaporation mass flow rate m of the above cutting fluid evap ) is input into the convolutional neural network (CNN) model. The cooling effect is predicted through a series of convolutional operations, pooling operations, fully connected layers, and output layers. The input-output relationship of the final model can be expressed as:

[0091] Y = f(X);

[0092] where X = [k, h, ρ, m evap , Y = Q cooling is the output predicted value.

[0093] Furthermore, GRU is introduced to process time series data. Each time step of GRU has a reset gate and an update gate mechanism.

[0094] In some embodiments, the reset gate and update gate of the GRU are controlled by the Sigmoid activation function. The reset gate is used to control the degree of forgetting of the historical state, and the update gate is used to fuse the current input with the historical hidden state. The candidate activation value is generated by the hyperbolic tangent function.

[0095] The reset gate follows the formula as follows:

[0096] r t = σ(W r X t + U r h t-1 + b r );

[0097] Wherein, r t represents the reset gate, σ is the Sigmoid activation function, X t is the input feature at the current moment, h t-1 is the hidden state at the previous moment, W r is the weight matrix of the reset gate, multiplied by the input feature X t , U r is the weight matrix of the reset gate, multiplied by the hidden state h t-1 at the previous moment, and b r is the bias term of the reset gate.

[0098] The update gate follows the formula as follows:

[0099] z t = σ(W z X t + U z h t-1 + b z );

[0100] Wherein, z t is the update gate, W z , U z are the weight matrices of the update gate, and b z is the bias term.

[0101] After that, combining the reset gate and the current input, the candidate activation value is obtained, which can be expressed as:

[0102]

[0103] Wherein, is the candidate activation value, ⊙ represents element-wise multiplication, W h , U h are the weights of the candidate activation value, and b h is the bias term.

[0104] Furthermore, the hidden state update formula is obtained. Through the update gate zt and the candidate activation value Derive the relationship of the hidden state at the current moment, expressed as:

[0105]

[0106] where h t is the hidden state at the current moment, representing the encoding of the cooling effect at the current moment. When z t is close to 1, the hidden state remains almost unchanged; when z t is close to 0, the state at the current moment has a greater impact on the hidden state.

[0107] Assume that the cooling effect in the next 1 hour is predicted based on the cutting fluid characteristics in the past m moments. The input X of the model = [X1, X2,..., X m is the input data in the past m moments, where each X t is composed of the physical characteristics of the cutting fluid and the historical load data.

[0108] The output of the GRU network is the predicted value Y t , that is, the predicted value of the cooling effect at the future moment t+1. The GRU model is established as follows:

[0109] h t = GRU(X t , h t-1 ; W, U, b);

[0110] where, X t is the input feature at the current moment (i.e., the thermal conductivity (k) of the cutting fluid, the convective heat transfer coefficient (h) between the cutting fluid and the workpiece surface, the density (ρ) of the cutting fluid, the evaporation mass flow rate (m evap )) of the cutting fluid); h t-1 is the hidden state at the previous moment; W, U, b are the weights and biases of the network.

[0111] By using GRU in the cutting fluid cooling effect prediction model, time series data can be effectively processed, and the long-term dependence relationship in the cooling effect prediction can be captured, which can be expressed as:

[0112] GRU layer: h t = GRU(X t , h t-1 ; W, U, b);

[0113] Input predicted value: Y = f(h m ).

[0114] S103. Optimize the hyperparameters of the prediction model using the Bayesian optimization method, including constructing a surrogate model through Gaussian process regression, and iteratively optimizing the hyperparameter combination by combining the expected improvement or probability improvement acquisition function to minimize the prediction error;

[0115] In some embodiments, the above step S103 may specifically include:

[0116] S1031. Initialize the hyperparameter space, and set the learning rate, the number of iterations, and the kernel function parameters;

[0117] S1032. Calculate the mean and variance of the objective function based on Gaussian process regression, and select the next set of hyperparameters to be evaluated through the acquisition function;

[0118] S1033. Use the mean squared error as the loss function, and iteratively update the hyperparameters until convergence.

[0119] Specifically, select the initial hyperparameters θ1, θ2,..., θ0, and assume that the goal is to find a set of optimal hyperparameters θ * such that the prediction performance is optimal. Select the next hyperparameter combination θ to be evaluated by maximizing the acquisition function * , and evaluate the objective function f(θ * ) at θ * to update the Gaussian process model.

[0120] Then, by evaluating the objective function, construct the model loss function f(θ). Use the mean squared error (MSE) as the loss function, and the loss function can be specifically expressed as:

[0121]

[0122] where is the cooling effect predicted by the model given the hyperparameters θ; Y i is the true cooling effect of sample i; and m is the number of samples. Minimize the loss function f(θ) by optimizing the hyperparameters θ.

[0123] After that, perform Bayesian optimization, and approximate the objective function f(θ) through the Gaussian process regression surrogate model. Assume that the objective function f(θ) is described by the Gaussian process as follows:

[0124]

[0125] where u(θ) is the mean function of the objective function. According to the Gaussian process model Predict the mean \(u(\theta)\) and standard deviation \(\sigma(\theta)\) of the objective function \(f(\theta)\). Assume that \(\mu(\theta)=0\); \(k(\theta,\theta')\) is the covariance function (or kernel function), which describes the similarity between different points in the hyperparameter space and follows the following relationship:

[0126]

[0127] where \(\sigma\) 2 is the variance parameter of the kernel function; \(l\) is the length scale, which controls the correlation between input points.

[0128] Furthermore, select the next hyperparameter combination \(\theta\) to be evaluated * . Use the acquisition functions expected improvement and probability of improvement to determine the selection of the next point.

[0129] In some embodiments, the next point can be selected by calculating the expected value of the possible improvement based on the current optimal solution through the expected improvement (EI). The formula calculation can be expressed as:

[0130]

[0131] where \(f\) min is the currently known minimum loss, represents the mathematical expectation of the random variable \(\max(0,f min -f(\theta))\); according to the properties of the Gaussian process, EI also follows the following formula:

[0132]

[0133] where \(\mu(\theta)\) is the predicted mean at \(\theta\); \(\sigma(\theta)\) is the predicted standard deviation at \(\theta\); \(\varphi\) and \(\varPhi\) are the probability density function and cumulative distribution function of the standard normal distribution respectively.

[0134] In other embodiments, the probability that the objective function is less than the current optimal value \(f\) min can be calculated through the probability of improvement (PI) and follows the following formula:

[0135]

[0136] When the predetermined number of iterations is reached or the objective function converges, stop the optimization.

[0137] In the embodiments of the present invention, the optimal hyperparameter combination is selected by continuously exploring and exploiting the objective function, and finally the high-precision prediction of the cooling effect of the cutting fluid is realized.

[0138] In the embodiments of the present invention, an intelligent prediction model is constructed by integrating a convolutional neural network and a gated recurrent unit, and the hyperparameters are dynamically adjusted in combination with the Bayesian optimization algorithm, realizing high-precision prediction and optimization of the cooling effect of the cutting fluid, significantly improving the prediction efficiency and reducing the calculation cost, being able to adapt to complex non-linear machining environments and different working condition requirements, effectively solving the problems of traditional methods relying on empirical formulas, difficult data acquisition, and poor model universality. At the same time, the optimization of the cutting fluid formula is guided by accurate prediction results, significantly improving the machining efficiency, workpiece quality, and resource utilization rate, and being able to provide reliable technical support for the intelligent upgrading of the industrial cutting fluid system.

[0139] The following uses a specific embodiment to illustrate the method provided by the present invention.

[0140] In this embodiment, the data is taken from the specific cutting fluid data of an oil field. The data of the cutting fluid is processed by an algorithm, and the process of the algorithm is as follows.

[0141] It is assumed that each factor can be described by a comprehensive model. Specifically, the comprehensive model of the cooling effect can be expressed as a mathematical expression that couples multiple factors such as heat conduction, convective heat transfer, evaporation effect, and shear heat generation. An intelligent comprehensive prediction model of the cooling effect is established, expressed as:

[0142]

[0143] Among them, the heat conduction term follows the following formula:

[0144]

[0145] The variation law of the convective heat transfer term is:

[0146] q conv =h*(T surface -T liquid );

[0147] The evaporation term follows the following formula:

[0148]

[0149] The generation of shear heat is usually proportional to the yield strength of the material and the cutting speed. It can be expressed by the following formula:

[0150] Q shear =τ*A cut *v cut ;

[0151] The cutting fluid optimization problem is transformed into an optimization problem, and the introduced feature matrix is established. The thermal conductivity k of the cutting fluid, the convective heat transfer coefficient h between the cutting fluid and the workpiece surface, the density ρ of the cutting fluid, and the evaporation mass flow rate m of the cutting fluid are selected. evapFour features. The function between the input features and the cooling effect is obtained by mathematical fitting and is expressed as:

[0152] Q cooling = f(k, h, ρ, m evap );

[0153] After that, an input feature matrix X of the cutting fluid cooling effect is constructed.

[0154] For example, taking 10 actual samples as an example, the dimension of matrix X is m×4. Each row represents a sample, and each column represents a feature. Matrix X can be expressed as:

[0155]

[0156] Then a CNN model is constructed, using a 1D convolutional layer, following the formula as follows:

[0157]

[0158] The pooling layer is used for downsampling. The pooling window size is 2, and the pooling stride is 2, following the following formula:

[0159] P1 = Maxpool(C1, 2);

[0160] After the pooling operation, the output dimension is 10×1×32; the data is flattened into a one-dimensional vector, following the following formula:

[0161] P1 = Flatten(P1);

[0162] The output dimension after flattening becomes 10×32; the flattened output is passed into a fully connected layer with 64 neurons, using the ReLU activation function, following the formula as follows:

[0163]

[0164] The predicted value of the cooling effect is generated through a fully connected layer, following the formula as follows:

[0165] Y = F1 * W3 + b3;

[0166] The input-output relationship of the final model can be expressed by the following formula:

[0167] Y = f(X);

[0168] where X = [k, h, ρ, m evap , Y = Q cooling is the output predicted value, and the result can be expressed as:

[0169]

[0170] Furthermore, the GRU is introduced to process time series data, and the reset gate follows the formula as follows:

[0171] r t = σ(W r X t + U r h t-1 + b r );

[0172] The update gate follows the formula as follows:

[0173] z t = σ(W z X t + U z h t-1 + b z );

[0174] Combining the reset gate and the current input, the candidate activation value is obtained, which can be expressed as:

[0175]

[0176] Furthermore, the hidden state update formula is obtained. Through the update gate z t and the candidate activation value the relationship of the hidden state at the current moment is derived, expressed as:

[0177]

[0178] The input of the model X = [X1, X2,..., X m .

[0179] The output of the GRU network is the predicted value Y t , that is, the predicted value of the cooling effect at the future time t + 1. The GRU model is established as follows:

[0180] h t = GRU(X t , h t-1 ; W, U, b);

[0181] By applying the GRU to the cutting fluid cooling effect prediction model, time series data can be effectively processed, and the long-term dependence relationship in the cooling effect prediction can be captured, which can be expressed as:

[0182] GRU layer: h t = GRU(X t , h t-1 ; W, U, b);

[0183] Input predicted value: Y = f(h m ).

[0184]

[0185] After that, hyperparameter optimization is used to improve the model accuracy. Combining Bayesian optimization, Gaussian processes and acquisition functions are introduced.

[0186] The initial hyperparameters are denoted as:

[0187]

[0188] Construct the model loss function f(θ). Using the mean squared error as the loss function, it is denoted as:

[0189]

[0190] Suppose the evaluated hyperparameter combinations Θ and loss values F are respectively denoted as:

[0191]

[0192] Furthermore, assume that the objective function f(θ) is represented by a Gaussian process described as follows:

[0193]

[0194] The covariance matrix K can be expressed as:

[0195]

[0196] Further calculate each element specifically:

[0197]

[0198] Suppose the specific values of θ1, θ2, θ3, θ4 are respectively denoted as:

[0199]

[0200] The calculated covariance matrix K can be expressed as:

[0201]

[0202] After that, given a new hyperparameter combination θ * the predicted mean μ(θ * ) is denoted as:

[0203] u(θ * ) = k(θ * , Θ)K -1 F;

[0204] The standard deviation σ(θ * ) is denoted as:

[0205]

[0206] Specifically, assume a new combination of hyperparameters θ * is:

[0207]

[0208] Calculate k(θ * , Θ) as follows:

[0209] k(θ * , Θ) = [k(θ * , θ1)k(θ * , θ2)k(θ * , θ3)k(θ * , θ4)];

[0210] The specific calculation result is expressed as:

[0211] k(θ * , Θ) = [0.007 0.008 0.009 0.010];

[0212] Furthermore, calculate K -1 , the mean u(θ * ) and the standard deviation, and the results are expressed as:

[0213]

[0214] u(θ * ) = 0.0005314;

[0215] The standard deviation σ(θ * ), is expressed as:

[0216]

[0217] Assume k(θ * , θ * ) = 0.011, and the calculation result is expressed as:

[0218]

[0219] σ(θ * ) ≈ 0.0899;

[0220] In some embodiments, the acquisition functions expected improvement and probability of improvement are used to determine the selection of the next point. The formula for expected improvement is calculated as follows:

[0221]

[0222] EI(θ *) = 0.0114686Φ(1.297) + 0.0899φ(1.297);

[0223] where μ max is the minimum value among the evaluated loss values; Φ(z) is the cumulative distribution function of the standard normal distribution; φ(z) is the probability density function of the standard normal distribution.

[0224] Assume μ max = 0.012;

[0225]

[0226] Using the standard normal distribution, the results are as follows:

[0227] Φ(1.297) ≈ 0.901;

[0228] φ(1.297) ≈ 0.171;

[0229] EI(θ * ) = 0.0114686 × 0.901 + 0.0899 × 0.171;

[0230] EI(θ * ) = 0.01033 + 0.01537;

[0231] EI(θ * ) ≈ 0.0257.

[0232] In some embodiments, the probability improvement formula is expressed as:

[0233]

[0234] μ max = 0.012, μ(θ * ) ≈ 0.0005314, sigma(θ * ) ≈ 0.0899;

[0235]

[0236] Add θ * and f(θ * ) to the existing dataset, reconstruct the Gaussian process model, and stop the optimization when the predetermined number of iterations is reached or the objective function converges.

[0237] In the embodiments of the present invention, by continuously exploring and exploiting the objective function to select the optimal hyperparameter combination, the high-precision prediction of the cooling effect of the cutting fluid is ultimately achieved.

[0238] Table 1 is the measurement result and error analysis data table in the embodiments of the present invention.

[0239] Table 1:

[0240]

[0241] As shown in Table 1, the relative error between the actual data and the calibration data is controlled within 1%, which proves that the model provided by the embodiments of the present invention has a high accuracy and can simulate the error of the cutting fluid cooling effect for calibration.

[0242] S104. Predict the cooling effect of the cutting fluid according to the optimized model, and adjust the cutting fluid formulation based on the prediction results to achieve dynamic optimization of the cooling effect.

[0243] Specifically, during the process of following up the prediction results of the cooling effect of the cutting fluid and adjusting the cutting fluid formulation, the desired target of the cooling effect of the cutting fluid in the machining process can be clarified, for example, controlling the temperature in the cutting area within a certain specific range, or reducing the tool wear rate to a certain extent.

[0244] By comparing the cooling effect predicted by the model with the desired target, analyze the gap between the two. If the predicted cooling effect does not reach the target, it indicates that the current cutting fluid formulation may need to be adjusted; if the predicted effect is much higher than the target, it may mean that there is a waste of resources in the current formulation and optimization is also required.

[0245] According to the gap between the prediction result and the target, combined with the understanding of the physical properties of the cutting fluid and the machining environment, determine the direction of formulation adjustment. For example, if the prediction result shows that the temperature in the cutting area is too high, it may be necessary to increase the thermal conductivity or convective heat transfer coefficient of the cutting fluid. This can be achieved by adding additives with high thermal conductivity or adjusting the composition ratio of the cutting fluid.

[0246] In some embodiments, the interaction and influence between different components can also be considered. When adjusting the formulation, the synergistic effect between the components can be considered. For example, some additives may increase the thermal conductivity, but at the same time, they will also affect the density and evaporation mass flow rate of the cutting fluid, and thus have a comprehensive impact on the cooling effect. Therefore, comprehensive analysis and trade-off are required.

[0247] According to the determined direction of formulation adjustment, formulate a specific adjustment plan. It can clearly indicate the components to be adjusted and the adjustment range. For example, it is decided to increase the content of a certain additive from 5% to 8%, or reduce the proportion of another component from 10% to 7%. Using the optimized prediction model, simulate and predict the adjusted formulation, and evaluate whether the adjusted cooling effect can be closer to the desired target. If the simulation result is not ideal, the plan needs to be further modified and optimized.

[0248] During the actual machining process, the cooling effect of the cutting fluid can be continuously monitored. By real-time monitoring indicators such as the temperature in the cutting area and the tool wear situation, verify whether the adjusted formula has achieved the expected effect. If it is found that the effect is still not ideal, it is necessary to analyze the reasons again and readjust the formula to form a closed-loop dynamic optimization process, continuously improve the cooling effect of the cutting fluid to adapt to different machining requirements and environmental changes.

[0249] Based on the same inventive concept, the embodiment of the present application also provides an optimization system for the cooling effect of cutting fluid based on an intelligent algorithm. Figure 2 It is a schematic structural diagram of an optimization system for the cooling effect of cutting fluid based on an intelligent algorithm in the embodiment of the present invention. Refer to Figure 2 As shown, the optimization system 200 for the cooling effect of cutting fluid based on an intelligent algorithm may include:

[0250] A data acquisition module 201, configured to obtain the physical property data and machining condition data of the cutting fluid;

[0251] A model construction module 202, configured to build an intelligent prediction model based on a convolutional neural network (CNN) and a gated recurrent unit (GRU);

[0252] A hyperparameter optimization module 203, configured to adjust the model hyperparameters by using a Bayesian optimization algorithm;

[0253] A prediction and optimization module 204, configured to predict the cooling effect of the cutting fluid according to the optimized model and adjust the cutting fluid formula based on the prediction result.

[0254] In some possible implementation manners, in the model construction module, the output of the CNN is flattened and then input into a fully connected layer, and the hidden state of the GRU is updated iteratively through time steps.

[0255] In some possible implementation manners, the hyperparameter optimization module constructs a surrogate model through Gaussian process regression and dynamically selects the optimal hyperparameter combination based on expected improvement or probability improvement.

[0256] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, reference can be made to each other. The key points of each embodiment are the differences from other embodiments.

[0257] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting the present application; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the present application.

Claims

1. A method for optimizing the cooling effect of cutting fluid based on an intelligent algorithm, characterized in that: include: Collecting physical property data of cutting fluid and processing environment parameters to construct a multi-dimensional input feature matrix, wherein the physical property data at least includes thermal conductivity, convection heat transfer coefficient, density and evaporation mass flow rate; Constructing an intelligent prediction model based on a convolutional neural network (CNN) and a gated recurrent unit (GRU), wherein the CNN is used to extract multidimensional data in the feature matrix, reduce the dimension of the multidimensional data and extract nonlinear features, and the CNN includes at least one 1D convolutional layer, a pooling layer and a fully connected layer; the GRU is used to process time series data and capture the long-term dependency of time series data in the prediction of cutting fluid cooling effect; Optimizing the hyperparameters of the prediction model using a Bayesian optimization method, including constructing a surrogate model through Gaussian process regression, and iteratively optimizing the hyperparameter combination in combination with an expected improvement or a probability improvement acquisition function to minimize the prediction error; The cooling effect of the cutting fluid is predicted according to the optimized model, and the cutting fluid formula is adjusted based on the prediction results to achieve dynamic optimization of the cooling effect.

2. The method according to claim 1, characterized in that The convolution kernel size of the 1D convolution layer of the CNN is 3, the number of convolution kernels is 32, the activation function is ReLU, and the output is flattened into a one-dimensional vector after being downsampled by the maximum pooling layer and input into a fully connected layer containing 64 neurons.

3. The method according to claim 2, characterized in that The reset gate and update gate of the GRU are controlled by the Sigmoid activation function. The reset gate is used to control the degree of forgetting of the historical state, and the update gate is used to fuse the current input with the historical hidden state. The hidden state update formula is expressed as: Among them, h t is the hidden state at the current moment, z t is the update gate output, is the candidate activation value, and t is the current time.

4. The method according to claim 3, characterized in that The Bayesian optimization method is used to optimize the hyperparameters of the prediction model, including: Initialize the hyperparameter space, set the learning rate, number of iterations and kernel function parameters; Calculate the mean and variance of the objective function based on Gaussian process regression, and select the next set of hyperparameters to be evaluated through the acquisition function; The mean square error is used as the loss function, and the hyperparameters are iteratively updated until convergence; the loss function is expressed as: Among them, f(θ) represents the loss function, θ is the initial hyperparameter, and Y i For a real cooling effect, is the cooling effect predicted by the model given the hyperparameter θ, and m is the number of samples.

5. The method according to claim 4, characterized in that The intelligent prediction model is expressed by the following formula: Among them, Q cooling To represent the intelligent prediction model, k is the thermal conductivity of the cutting fluid, is the temperature gradient, h is the convective heat transfer coefficient, T is the temperature field, T ambient is the ambient temperature, m evap is the evaporation mass flow rate of the cutting fluid, ρ is the density of the cutting fluid, and L is the latent heat of evaporation of the cutting fluid.

6. A cutting fluid cooling effect optimization system based on intelligent algorithm, used to implement the method described in any one of claims 1 to 5, characterized in that: include: Data acquisition module, used to obtain physical property data of cutting fluid and processing condition data; Model building module for intelligent prediction models based on convolutional neural networks (CNN) and gated recurrent units (GRU); Hyperparameter optimization module, used to adjust model hyperparameters using Bayesian optimization algorithm; The prediction and optimization module is used to predict the cooling effect of the cutting fluid according to the optimized model and adjust the cutting fluid formula based on the prediction results.

7. The cutting fluid cooling effect optimization system based on intelligent algorithm according to claim 6 is characterized in that: In the model building module, the output of the CNN is flattened and then input into the fully connected layer, and the hidden state of the GRU is updated through time step iterations.

8. The cutting fluid cooling effect optimization system based on intelligent algorithm according to claim 7 is characterized in that: The hyperparameter optimization module constructs a proxy model through Gaussian process regression and dynamically selects the optimal hyperparameter combination based on expected improvement or probability improvement.

Images