A Machine Learning-Based Method and System for Preparing Aged Oil-Based and Water-Based Metal Cutting Fluids

Abstract

This application discloses a method and system for preparing aged oil-based water-based metal cutting fluid based on machine learning, comprising: obtaining stability characterization parameters of aged oil for stability classification and determining pretreatment parameters; sequentially performing preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation and solid-liquid separation to obtain a base oil for the cutting fluid; using the obtained base oil as the oil phase, compounding it with emulsifier, oiliness agent, extreme pressure anti-wear agent, rust inhibitor, pH stabilizer, bactericide and defoamer to construct a formulation sample and collect performance and cost data; using the formulation sample to train a wavelet neural network to establish a formulation-performance mapping model; using simulated annealing-particle swarm optimization algorithm to obtain the target formulation; preparing aged oil-based water-based metal cutting fluid according to the target formulation and verifying it; this method realizes the high-value resource utilization of aged oil, effectively demulsifies and dehydrates through multi-stage synergistic treatment, and efficiently obtains the globally optimal formulation by combining machine learning algorithms. The resulting cutting fluid has cooling, rust prevention, lubrication and cleaning properties, reducing production costs and reducing environmental pollution.

Description

Technical Field

[0001] This invention relates to the fields of industrial waste oil resource utilization technology and metalworking fluid technology, specifically to a method and system for preparing aged oil-based water-based metal cutting fluid based on machine learning. Background Technology

[0002] Aged oil refers to oil-water intermediate transition layers, oil-water emulsion layers in emergency tanks, recovered crude oil from spills, and difficult-to-treat oils containing various chemical agents formed during oilfield development, gathering and transportation, wastewater treatment, and tank cleaning. This type of aged oil typically contains clay, sand particles, corrosion products, scale, bacteria, asphaltenes, and various surfactants, resulting in a thick oil-water interface film, high interfacial strength, high potential, and strong stability. Conventional thermochemical sedimentation and simple gravity separation methods are insufficient for effective demulsification and dehydration. Currently, oilfield sites often re-mix aged oil into the crude oil system or use a single thermochemical treatment method for disposal. However, these methods easily disturb the electrostatic dehydration system, worsen the oil-water interface, increase treatment costs, and increase the risk of water pollution. If aged oil cannot be efficiently treated, it will not only occupy the effective volume of sedimentation equipment and affect oil-water separation efficiency, but also cause crude oil resource loss and a heavy environmental burden.

[0003] On the other hand, metal cutting fluids are important auxiliary materials in machining processes, requiring multiple functions such as cooling, rust prevention, lubrication, and cleaning. With the development of high-speed machining, precision machining, and green manufacturing, traditional oil-based cutting fluids are gradually being limited due to their high levels of smoke, insufficient cooling, and poor safety, while water-based cutting fluids have become the mainstream. However, existing commercially available emulsified cutting fluids still generally suffer from insufficient adaptability to ferrous and non-ferrous metals, low environmental friendliness, limited service life, and high cost. Aged oils, after proper treatment, have the potential to be converted into cutting fluid base oils. If, under the premise of ensuring dehydration, impurity removal, and stability, they can be utilized as the oil phase resource of emulsified cutting fluids, it can not only significantly increase the added value of aged oils but also reduce cutting fluid production costs and decrease waste oil discharge.

[0004] However, there are strong nonlinear coupling effects among the base oil, emulsifier, oiliness agent, extreme pressure anti-wear agent, rust inhibitor, pH stabilizer and other additives in the formulation of aged oil-based cutting fluids. It is difficult to find the globally optimal combination in a large variable space by simply relying on orthogonal experiments or empirical methods.

[0005] Therefore, there is an urgent need to provide a preparation method that combines efficient pretreatment of aged oil with intelligent optimization through machine learning, so as to achieve stable preparation of aged oil-based water-based metal cutting fluid. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for preparing aged oil-based and water-based metal cutting fluids based on machine learning.

[0007] To achieve the above objectives, the present invention provides the following technical solution: This application provides a method for preparing aged oil-based and water-based metal cutting fluid based on machine learning, including: Obtain the stability characterization parameters of the aged oil to be treated, classify the stability of the aged oil based on the stability characterization parameters, and determine the pretreatment parameters based on the classification results. The aged oil is subjected to preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation and solid-liquid separation in sequence to obtain the cutting fluid base oil. Using the cutting fluid base oil as the oil phase, multiple formulation samples were constructed by compounding it with emulsifiers, oiliness agents, extreme pressure anti-wear agents, rust inhibitors, pH stabilizers, bactericides, and defoamers. The corresponding cooling properties, rust prevention properties, lubrication properties, cleaning properties, stability, salt tolerance, defoaming performance, and cost indicators of each formulation sample were obtained. A wavelet neural network was trained using the composition and performance data of the aforementioned formula samples to establish a formula-performance mapping model; The simulated annealing-particle swarm optimization algorithm is used to optimize the formula-performance mapping model to obtain the target formula; An aged oil-based water-based metal cutting fluid was prepared according to the target formulation, and its performance was verified.

[0008] Optionally, the stability characterization parameters include physical property parameters and microscopic stability parameters. The physical property parameters include at least density, viscosity, water content, and solid content. The microscopic stability parameters include at least water droplet size distribution, median particle size, zeta potential, interfacial film strength, and stratification behavior under temperature changes. The stability grading includes: constructing a stability evaluation index based on the absolute value of zeta potential, the characterization value of interfacial film strength, and the median particle size. When the stability evaluation index reaches a preset threshold, the aged oil is determined to be a highly stable emulsion system, and at least one of the following is increased accordingly: dilution ratio, ultrasonic intensity, and electrostatic coalescence time.

[0009] Optionally, the process conditions for the preheating and diluting step are: heating to 45–70°C, holding at that temperature for 10–30 minutes, and adding a diluent of 5%–25% of the mass of the aged oil; The ultrasonic treatment process conditions are: frequency 20-40kHz, power density 0.2-1.0W / mL, treatment time 5-30min; The electrostatic coalescence process conditions are: electric field strength 1-6 kV / cm, treatment time 3-20 min; The centrifugal separation process conditions are: rotation speed 2000-8000 r / min, separation time 5-30 min; The solid-liquid separation employs a two-stage filtration process, with the first stage having a filtration accuracy of 20–50 μm and the second stage having a filtration accuracy of 1–10 μm.

[0010] Optionally, the water content of the resulting cutting fluid base oil shall not exceed 0.3%, and the solid impurity content shall not exceed 0.05%.

[0011] Optionally, the target formulation is an emulsified cutting fluid concentrate formulation, comprising, by mass percentage: 15%–45% of the cutting fluid base oil, 8%–20% of the composite emulsifier, 3%–12% of the oiliness agent, 1%–10% of the extreme pressure anti-wear agent, 1%–8% of the rust inhibitor, 0.5%–5% of the pH stabilizer, 0.05%–1% of the bactericide, 0.01%–0.5% of the defoamer, with the balance being water; wherein the composite emulsifier is formed by compounding anionic and nonionic surfactants, and the extreme pressure anti-wear agent is a chlorine-free, low-sulfur, or boron-containing environmentally friendly extreme pressure anti-wear agent.

[0012] Optionally, the preparation of the formulation sample includes: heating the base oil of the cutting fluid to 35-50°C, adding oiliness agent, extreme pressure anti-wear agent, rust inhibitor and composite emulsifier in sequence, and stirring at 300-800 r / min for 10-20 min; then adding pH stabilizer, bactericide, defoamer and the remaining water, and continuing to stir for 15-30 min to obtain emulsified cutting fluid concentrate; the emulsified cutting fluid concentrate is diluted with water at a mass ratio of 1:(10-30) when used.

[0013] Optionally, the input features of the formulation-performance mapping model include at least the proportion of base oil, the proportion of compound emulsifier, the proportion of oiliness agent, the proportion of extreme pressure anti-wear agent, the proportion of rust inhibitor, the proportion of pH stabilizer, the proportion of bactericide, the proportion of defoamer, the type of processed material, the cutting speed, the feed rate, and the water hardness. The output feature is a comprehensive performance index, which is determined by weighting the cooling performance, rust prevention, lubrication, cleaning performance, stability, defoaming performance, and cost indicators. The wavelet neural network uses wavelet basis functions as hidden layer activation functions. The simulated annealing-particle swarm optimization algorithm uses dynamic inertia weights and annealing acceptance probability for global optimization, with the pH value of the diluent being between 7.5 and 9.5 as a constraint.

[0014] A machine learning-based system for preparing aged oil-based and water-based metal cutting fluids includes: The stability analysis module is used to obtain stability characterization parameters of aged oil, perform stability classification, and output preprocessing parameters; The pretreatment module is used to sequentially perform preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugation and solid-liquid separation to obtain the cutting fluid base oil; The formulation sample construction module is used to form formulation samples by compounding the cutting fluid base oil with various additives as the oil phase, and to collect the performance data of each sample; The machine learning modeling module is used to train a wavelet neural network using the data from the formula samples to establish a formula-performance mapping model. The optimization module is used to optimize the formula-performance mapping model using a simulated annealing-particle swarm optimization algorithm and output the target formula. The quality verification module is used to prepare aged oil-based and water-based metal cutting fluids according to the target formulation and to complete performance verification.

[0015] An electronic device includes a memory and a processor, the memory storing executable instructions, and the processor executing the executable instructions to implement the fabrication method described above.

[0016] A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, implement the preparation method described above.

[0017] Compared with the prior art, this application has the following beneficial effects: This invention proposes a machine learning-based method and system for preparing aged oil-based water-based metal cutting fluids. By acquiring stability characterization parameters of the aged oil and performing stability classification, pretreatment parameters are determined based on the classification results. The aged oil is then subjected to synergistic demulsification, dehydration, and purification treatment in the order of preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation, and solid-liquid separation, resulting in a cutting fluid base oil that meets compatibility requirements. This solves the technical problem of difficult demulsification and dehydration of highly stable aged oils using conventional methods, achieving efficient purification and resource utilization of aged oils. Simultaneously, a formulation sample is constructed by compounding various functional additives with the obtained base oil as the oil phase. A wavelet neural network is trained using the composition and performance data of the formulation sample to establish a formulation-performance mapping model. Simulated annealing-particle swarm optimization is used to globally optimize the model to obtain the target formulation. This solves the technical problem of difficulty in efficiently obtaining the globally optimal combination under strong nonlinear coupling between components in the cutting fluid formulation, significantly improving formulation screening efficiency and reducing blind experiments. The final aged oil-based water-based metal cutting fluid possesses good cooling, rust prevention, lubrication, and cleaning properties, reducing production costs and environmental pollution. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the aging oil pretreatment process in an embodiment of the present invention; Figure 2This is a schematic diagram of the optimization process of wavelet neural network coupled with simulated annealing-particle swarm optimization algorithm in an embodiment of the present invention; Figure 3 This is a flowchart of the multi-parameter coupled BP neural network model algorithm system in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the method for calculating the stability of emulsion in high-speed steel cutting fluid according to an embodiment of the present invention. Figure 5 This is a flowchart illustrating the method for optimizing cutting fluid cooling effect based on intelligent algorithms in an embodiment of the present invention. Figure 6 This is a flowchart of a machine learning-based method for preparing aged oil-based and water-based metal cutting fluids. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Furthermore, in this invention, an element referred to as fixed to or disposed on another element may be directly disposed on the other element, or there may be an intermediate element. When an element is considered to be connected to another element, it may be directly connected to the other element, or there may be an intermediate element present simultaneously. The terms vertical, horizontal, left, right, and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0021] Unless otherwise stated, all percentages mentioned in this embodiment are mass percentages, temperature is in °C, time is in min, rotation speed is in r / min, and water hardness is expressed as CaCO3. In the performance tests, rust prevention, stability, salt tolerance, and defoaming performance are conducted in accordance with SH / T0365-1992 "Emulsified Oils". The water content is determined by the Karl Fischer method, the solid impurity content is determined by the centrifugation-drying method, the particle size distribution is determined by a laser particle size analyzer, and the zeta potential is determined by electrophoretic light scattering.

[0022] Firstly, see [the following] Figures 1-6 This application provides a method for preparing aged oil-based and water-based metal cutting fluid based on machine learning, including steps S1-S6: Step S1. Obtain the stability characterization parameters of the aged oil to be treated, classify the stability of the aged oil based on the stability characterization parameters, and determine the pretreatment parameters according to the classification results.

[0023] Stability characterization parameters of aged oil are key indicators for determining its ease of demulsification and dehydration. In this step, the physical properties and microscopic stability parameters of the aged oil to be treated are first collected. The physical properties include at least density, viscosity, water content, and solid content, reflecting the basic physical properties and contamination level of the aged oil. Density can be measured using a hydrometer, viscosity using a rotational viscometer, water content using the Karl Fischer method, and solid content using a centrifugation-drying method. The microscopic stability parameters include at least water droplet size distribution, median particle size (D50), zeta potential, interfacial film strength, and stratification behavior under temperature changes. Water droplet size distribution and median particle size (D50) can be measured using a laser particle size analyzer, zeta potential using electrophoretic light scattering, interfacial film strength can be characterized by measuring the interfacial film elastic modulus or critical rupture stress using an interfacial rheometer, and stratification behavior under temperature changes can be evaluated by placing the aged oil sample under different temperature conditions and observing the oil-water separation.

[0024] Based on the above parameters, the stability of aged oils is classified. Specifically, a stability evaluation index SI can be constructed based on the absolute value of zeta potential, the characteristic value of interfacial film strength, and the median particle size D50. The stability evaluation index satisfies SI = α × |Z| + β × IF + γ × (1 / D50), where Z is the zeta potential, IF is the characteristic value of interfacial film strength, and α, β, and γ are weighting coefficients. Generally, the larger the absolute value of zeta potential, the stronger the repulsive force of the oil-water interface double layer, and the more stable the emulsion system; the higher the interfacial film strength, the less likely water droplets will coalesce. The greater the resistance, the smaller the median particle size, the smaller the dispersed phase size, and the higher the thermodynamic stability. Therefore, the larger the SI value, the stronger the emulsification stability of the aged oil, and the greater the difficulty of demulsification and dehydration. When the stability evaluation index is greater than or equal to the preset threshold, the aged oil is determined to be a highly stable emulsion system; otherwise, it is determined to be a medium-to-low stable emulsion system. Based on the classification results, the subsequent pretreatment parameters are adjusted in a targeted manner. For example, when it is determined to be a highly stable emulsion system, at least one of the following is increased: dilution ratio, ultrasonic intensity, and electrostatic aggregation time, in order to enhance the demulsification effect.

[0025] Step S2. The aged oil is subjected to preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation and solid-liquid separation in sequence to obtain the cutting fluid base oil.

[0026] This step employs a multi-stage synergistic demulsification, dehydration, and purification process. The sequence and synergistic effect of each unit operation are key to achieving efficient dehydration.

[0027] Preheating and diluting is the first step, the purpose of which is to reduce the viscosity of aged oil and weaken the strength of the oil-water interface film, creating conditions for subsequent treatment. The preferred temperature for preheating and diluting is 45-70℃, and the holding time is 10-30 minutes. If the temperature is too low, the viscosity reduction effect will be insufficient, and if the temperature is too high, it may cause the volatilization of light components or the deterioration of components with poor thermal stability in the aged oil. The diluent is a low-viscosity base oil or light oil, and the amount added is 5%-25% of the mass of the aged oil. The diluent molecules can penetrate into the oil-water interface film, reduce the concentration of natural surface-active substances on the interface film, and weaken the strength of the interface film.

[0028] The second step is ultrasonic treatment, which utilizes the cavitation, mechanical vibration, and thermal effects generated by ultrasound in a liquid medium to apply high-frequency oscillations to the emulsified water droplets in the aged oil, causing the water droplets to vibrate, collide, and coalesce. The frequency of ultrasonic treatment is 20–40 kHz, the power density is 0.2–1.0 W / mL, and the treatment time is 5–30 min. The collapse of cavitation bubbles generates local high temperature and pressure and microjets, which can effectively destroy the oil-water interface film and cause small water droplets to coalesce into large water droplets.

[0029] Electrostatic coalescence is the third step, which uses an electric field to induce the dispersion of water droplets to polarize and attract each other to coalesce. The electric field strength for electrostatic coalescence is 1–6 kV / cm, and the processing time is 3–20 min. Under the action of the electric field, the emulsion droplets are polarized along the direction of the electric field, and dipole-dipole interaction forces are generated between adjacent droplets, driving droplet coalescence. Electrostatic coalescence is particularly effective for micron-sized water droplets, which can significantly increase the droplet size and create favorable conditions for subsequent centrifugal separation.

[0030] Centrifugation is the fourth step, which utilizes the density difference between the oil and water phases to achieve rapid stratification in a centrifugal force field. The centrifugal speed is 2000-8000 r / min, and the separation time is 5-30 min. After ultrasonic and electrostatic treatment, the water droplets that have agglomerated and grown rapidly settle under the action of strong centrifugal force, thus achieving oil-water separation. The centrifuge type can be a horizontal screw centrifuge or a disc centrifuge.

[0031] Solid-liquid separation is the fifth step, used to further remove residual trace amounts of water and solid impurities from the oil phase, ensuring the cleanliness of the base oil. The solid-liquid separation employs a two-stage filtration structure: the first stage has a filtration precision of 20–50 μm to trap larger particles; the second stage has a filtration precision of 1–10 μm to finely filter out tiny solid impurities. After the above treatment, the resulting cutting fluid base oil has a water content of no more than 0.3% and a solid impurity content of no more than 0.05%, meeting the compatibility requirements with subsequent emulsified cutting fluid formulations. Through the synergistic combination of preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation, and solid-liquid separation, each unit operation forms a progressive and complementary relationship in terms of mechanism, effectively reducing the water content and impurity content of aged oil, improving the compatibility and stability of subsequent base oils, and solving the problem that conventional single-treatment methods are difficult to effectively demulsify and dehydrate.

[0032] Step S3. Using the cutting fluid base oil as the oil phase, compound it with emulsifier, oiliness agent, extreme pressure anti-wear agent, rust inhibitor, pH stabilizer, bactericide and defoamer to construct multiple formulation samples, and obtain the corresponding cooling properties, rust prevention properties, lubrication properties, cleaning properties, stability, salt tolerance, defoaming performance and cost indicators for each formulation sample.

[0033] In this step, the components of the formulation and their dosage range are first determined. The target formulation is an emulsified cutting fluid concentrate formulation, which, by mass percentage, comprises: 15%–45% of the cutting fluid base oil, 8%–20% of the composite emulsifier, 3%–12% of the oiliness agent, 1%–10% of the extreme pressure anti-wear agent, 1%–8% of the rust inhibitor, 0.5%–5% of the pH stabilizer, 0.05%–1% of the bactericide, 0.01%–0.5% of the defoamer, and the balance being water. The composite emulsifier is formed by combining anionic and nonionic surfactants. Sodium petroleum sulfonate can be selected as the anionic surfactant. The surfactants can be fatty alcohol polyoxyethylene ethers and sorbitan monooleate, which achieve good emulsification stability and a wide HLB value range through the synergistic effect of anionic and nonionic surfactants; the extreme pressure anti-wear agent is preferably a chlorine-free, low-sulfur or boron-containing environmentally friendly extreme pressure anti-wear agent, such as a compound system of boron-containing extreme pressure agent and low-sulfur ester, to meet the requirements of green manufacturing; the oiliness agent can be a synthetic ester; the rust inhibitor can be a compound system of organic carboxylates and alcohol amine borate esters; the pH stabilizer can be a triethanolamine and carbonate buffer system; the bactericide can be isothiazolinones; and the defoamer can be polyether-modified siloxanes.

[0034] The preparation process of the formulation sample is as follows: First, heat the base oil of the cutting fluid to 35-50℃ to reduce viscosity and improve the solubility and dispersibility of each component; then add the oiliness agent, extreme pressure anti-wear agent, rust inhibitor and composite emulsifier in sequence, and stir at 300-800 r / min for 10-20 min to ensure that each additive is fully dissolved and forms a uniform oil phase; after the system is homogeneous, add the pH stabilizer, bactericide, defoamer and the remaining water, and continue stirring for 15-30 min to form a stable water-in-oil emulsion concentrate through the phase inversion emulsification method; the resulting emulsion cutting fluid concentrate is diluted with water at a mass ratio of 1:(10-30) to form the working fluid.

[0035] After the preparation of each formulation sample, performance tests were conducted to obtain the corresponding cooling performance, rust prevention performance, lubrication performance, cleaning performance, stability, salt tolerance, defoaming performance, and cost indicators for each formulation sample. Among them, cooling performance can be characterized by the peak temperature drop in the grinding zone or the cooling curve; rust prevention performance is tested according to SH / T0365-1992 for single-piece and stacked cast iron pieces; lubrication performance is determined by the four-ball method to measure the maximum non-seize load PB value and the wear scar diameter; cleaning performance is characterized by the residual rate of standard dirt pieces; stability is evaluated by the stratification, floating oil, and sedimentation of the diluted solution after standing for 24 hours and 7 days; salt tolerance and defoaming performance are tested according to SH / T0365-1992; and cost indicators are calculated based on the production cost per unit of concentrated solution.

[0036] Step S4. Train a wavelet neural network using the composition and performance data of the formula sample to establish a formula-performance mapping model.

[0037] This step establishes a quantitative mapping relationship between formulation composition and overall performance using machine learning methods. First, a training dataset is constructed. The input features of this training dataset include at least the following: base oil ratio, compound emulsifier ratio, oiliness agent ratio, extreme pressure anti-wear agent ratio, rust inhibitor ratio, pH stabilizer ratio, bactericide ratio, defoamer ratio, processing material type, cutting speed, feed rate, and water hardness. Among these, the base oil and additive ratios are continuous variables, the processing material type is a categorical variable (which can be processed using unique thermal coding), the cutting speed and feed rate are process parameters, and the water hardness reflects the difference in dilution water quality. All input features are normalized to the [-1, 1] interval before modeling to eliminate the impact of dimensional differences on model training; the output label of the training dataset is the comprehensive performance index F, which satisfies F=w1·C+w2·R+w3·L+w4·Q+w5·S+w6·D-w7·K, where C, R, L, Q, S, D and K represent the normalized scores of cooling performance, rust prevention, lubrication, cleaning performance, stability, defoaming performance and cost indicators, respectively, and w1~w7 are the corresponding weight coefficients, which are determined according to the emphasis on each performance in the actual application scenario.

[0038] Wavelet neural networks are an organic combination of wavelet analysis and artificial neural networks, possessing both the excellent time-frequency localization characteristics of wavelet transform and the self-learning ability of neural networks, effectively approximating complex nonlinear function relationships. The wavelet neural network adopts an input layer, two hidden layers, and an output layer structure. The number of nodes in the input layer is consistent with the dimension of the input features. The number of nodes in the first hidden layer can be determined based on empirical formulas combined with cross-validation, for example, set to 14 nodes. The number of nodes in the second hidden layer can be set to 7 nodes. The number of nodes in the output layer is 1, i.e., the output comprehensive performance index F is used. The activation function of the hidden layers adopts a wavelet transform. The preferred wavelet basis function is the Morlet wavelet function, expressed as ψ(t) = cos(1.75t)·exp(-t² / 2). The wavelet basis function achieves multi-scale localization approximation through translation and scaling parameters, which can better capture the complex nonlinear mapping relationship between recipe and performance compared to traditional activation functions such as Sigmoid. The network training adopts the error backpropagation algorithm, which continuously adjusts the network weights, translation parameters, and scaling parameters of the wavelet basis function by minimizing the mean square error between the predicted output and the measured comprehensive performance index until the model converges.

[0039] Step S5. The simulated annealing-particle swarm optimization algorithm is used to optimize the formula-performance mapping model to obtain the target formula.

[0040] After establishing the recipe-performance mapping model, it is necessary to search for the recipe combination that achieves the best overall performance in the multidimensional recipe variable space. The traditional particle swarm optimization algorithm has strong global search capabilities, but it is prone to getting trapped in local optima. The simulated annealing algorithm, on the other hand, has the ability to escape local optima by probabilistically accepting inferior solutions. This invention combines the two to form a simulated annealing-particle swarm hybrid algorithm for optimization.

[0041] Specifically, the position vector of each particle in the particle swarm represents a set of formulation variables, namely, the proportions of base oil, compound emulsifier, oiliness agent, extreme pressure anti-wear agent, rust inhibitor, pH stabilizer, bactericide, and defoamer. The algorithm parameters are set as follows: the number of particles is 20–60, the maximum number of iterations is 50–200, the dynamic inertia weight is decreased from 0.9 to 0.4 to balance global search and local convergence capabilities, and the learning factors c1 and c2 are set to 1.5–2.5 respectively; particle velocity and position... The update of the position follows the rules of the standard particle swarm optimization algorithm. In each iteration, the fitness value of the current particle is calculated, i.e., the trained wavelet neural network model is used to predict its comprehensive performance index F. When the particle tends to converge to a certain position, a simulated annealing mechanism is introduced: a solution worse than the current global optimum is accepted with a certain probability. This acceptance probability is controlled by the annealing temperature, which gradually decreases with the increase of the number of iterations. The initial annealing temperature T0 is preferably 1 to 10, and the cooling coefficient λ is 0.85 to 0.98. Each cooling is based on T. k+1=λ·T k This hybrid strategy can maintain good diversity in the early stages of the search and gradually converge to the global optimum in the later stages of the search.

[0042] The constraints during the optimization process include: the pH value of the diluent is in the range of 7.5 to 9.5 to meet the corrosion protection requirements for metal workpieces and the skin safety of operators; there is no obvious floating oil and sedimentation after 24 hours to ensure the stability of the diluent; the rust prevention test is qualified; the defoaming time is no more than 10 seconds; and the unit cost of the concentrate does not exceed the preset upper limit. Through the above constraints, it is ensured that the optimized formula not only has excellent comprehensive performance, but also has engineering practicality and economic feasibility.

[0043] Step S6. Prepare aged oil-based water-based metal cutting fluid according to the target formulation and verify its performance.

[0044] According to the target formulation obtained in step S5, the preparation and verification were carried out on a large scale. The preparation process was the same as that of the formulation sample in step S3. That is, the base oil of the cutting fluid was first heated to 35-50°C, and the oiliness agent, extreme pressure anti-wear agent, rust inhibitor and composite emulsifier were added in sequence and stirred evenly. Then, the pH stabilizer, bactericide, defoamer and the balance water were added and stirred to obtain a concentrate. The concentrate was diluted with water at a mass ratio of 1:(10-30). The rust prevention, stability, salt tolerance and defoaming performance of the diluted solution were tested and verified according to the SH / T0365-1992 "Emulsified Oil" standard. At the same time, the cooling, lubrication and cleaning performance were verified under actual cutting conditions and compared with the model prediction results to ensure that the performance of the obtained product meets the usage requirements.

[0045] In a preferred embodiment, the stability characterization parameters include physical property parameters and microscopic stability parameters. The physical property parameters include at least density, viscosity, water content, and solid content. The microscopic stability parameters include at least water droplet size distribution, median particle size, zeta potential, interfacial film strength, and stratification behavior under temperature changes.

[0046] The stability grading includes: constructing a stability evaluation index based on the absolute value of zeta potential, the characterization value of interfacial film strength, and the median particle size. When the stability evaluation index reaches a preset threshold, the aged oil is determined to be a highly stable emulsion system, and at least one of the following is increased accordingly: dilution ratio, ultrasonic intensity, and electrostatic coalescence time.

[0047] In this embodiment, the specific composition of the stability characterization parameters and the specific method for stability grading are as follows: Physical properties of the aged oil were collected, including density, viscosity, water content, and solid content. Simultaneously, microscopic stability parameters were collected, including droplet size distribution, median particle size D50, zeta potential, interfacial film strength, and stratification behavior under temperature changes. A stability evaluation index SI was constructed based on the absolute value of zeta potential |Z|, the interfacial film strength characterization value IF, and the median particle size D50, satisfying SI = α × |Z| + β × IF + γ × (1 / D50). The weighting coefficients α, β, and γ were determined according to the relative importance of each parameter's contribution to emulsification stability, for example, they could be 0.45, 0.35, and 0.20, respectively. When the SI value is greater than or equal to a preset threshold (e.g., 45), the aged oil is determined to be a highly stable emulsion system; when the SI value is lower than the threshold, it is determined to be a medium- or low-stability emulsion system. When it is determined to be a highly stable emulsion system, at least one of the following should be increased: the dilution ratio (e.g., from 10% to 15% or more), the intensity of ultrasonic action (e.g., from 0.3 W / mL to 0.45 W / mL or more), and the electrostatic coalescence time (e.g., from 6 min to 10 min or more) to ensure the demulsification and dehydration effect. Conversely, when it is determined to be a medium- or low-stability emulsion system, the pretreatment intensity can be appropriately reduced to save energy and processing costs.

[0048] As a preferred embodiment, the process conditions for the preheating and diluting step are: heating to 45-70°C, holding for 10-30 minutes, and adding diluent at 5%-25% of the mass of the aged oil; The ultrasonic treatment process conditions are: frequency 20-40kHz, power density 0.2-1.0W / mL, treatment time 5-30min; The electrostatic coalescence process conditions are: electric field strength 1-6 kV / cm, treatment time 3-20 min; The centrifugal separation process conditions are: rotation speed 2000-8000 r / min, separation time 5-30 min; The solid-liquid separation employs a two-stage filtration process, with the first stage having a filtration accuracy of 20–50 μm and the second stage having a filtration accuracy of 1–10 μm.

[0049] In this embodiment, the specific process parameters for preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation, and solid-liquid separation, and the basis for their setting, are as follows: In the preheating and dilution step, the aged oil is heated to 45–70°C and held for 10–30 minutes. The reason for setting the temperature in this range is that the viscosity reduction effect is not obvious below 45°C and the improvement of interfacial film fluidity is limited; above 70°C, some light components in the aged oil may volatilize and energy consumption increases; the holding time of 10–30 minutes can ensure that the overall temperature of the oil sample is uniform and the diluent is fully diffused; the diluent is a low-viscosity base oil, and the amount added is 5%–25% of the mass of the aged oil; if the dilution ratio is too small, the effect of weakening the interfacial film will be insufficient; if the dilution ratio is too large, it will increase the volume and cost of subsequent processing.

[0050] In the ultrasonic treatment step, the frequency is 20–40 kHz, the power density is 0.2–1.0 W / mL, and the treatment time is 5–30 min. When the frequency is below 20 kHz, the noise is relatively large and the cavitation effect is weak. When the frequency is above 40 kHz, the cavitation threshold increases and the effective cavitation area decreases. A power density in the range of 0.2–1.0 W / mL can generate a cavitation effect of sufficient intensity to destroy the oil-water interface film, while avoiding excessive oil temperature rise or exacerbation of emulsification due to excessive power. A treatment time of 5–30 min can ensure sufficient cavitation while taking into account treatment efficiency.

[0051] In the electrostatic coalescence step, the electric field strength is 1–6 kV / cm, and the treatment time is 3–20 min. When the electric field strength is below 1 kV / cm, the dipole-dipole interaction force is insufficient, making it difficult to effectively induce droplet coalescence. When it is above 6 kV / cm, it may cause electrodispersion, leading to droplet breakage rather than coalescence. A treatment time of 3–20 min can allow the droplets to fully coalesce and grow, creating favorable conditions for subsequent centrifugal separation.

[0052] In the centrifugation separation step, the rotation speed is 2000-8000 r / min and the separation time is 5-30 min. If the rotation speed is too low, the separation efficiency will be insufficient; if the rotation speed is too high, the equipment requirements and energy consumption will increase. Within this rotation speed and time range, water droplets and solid particles that have agglomerated and grown after ultrasonic and electrostatic treatment can be effectively separated.

[0053] The solid-liquid separation step employs a two-stage filtration structure. The first stage has a filtration precision of 20–50 μm, which is used to trap larger particles and flocculent matter. The second stage has a filtration precision of 1–10 μm, which is used to finely filter out tiny solid impurities and ensure the cleanliness of the base oil.

[0054] As a preferred embodiment, the water content of the resulting cutting fluid base oil is not higher than 0.3%, and the solid impurity content is not higher than 0.05%.

[0055] In this embodiment, after the above-mentioned synergistic demulsification, dehydration, and purification treatment, the quality indicators of the obtained cutting fluid base oil are strictly controlled as follows: water content not exceeding 0.3% and solid impurity content not exceeding 0.05%. The water content is determined by the Karl Fischer method. When the water content is higher than 0.3%, it will affect the emulsion stability and the transparency of the concentrate in the subsequent emulsion formulation process, and may cause the risk of corrosion of metal workpieces. The solid impurity content is determined by the centrifugation-drying method. When the solid impurity content is higher than 0.05%, it will cause accelerated tool wear and workpiece surface scratches during the cutting process. Therefore, by strictly controlling these two indicators, the good compatibility between the base oil and the subsequent emulsified cutting fluid formulation system is ensured.

[0056] In a preferred embodiment, the target formulation is an emulsified cutting fluid concentrate formulation, comprising, by mass percentage: 15%–45% of the cutting fluid base oil, 8%–20% of the composite emulsifier, 3%–12% of the oiliness agent, 1%–10% of the extreme pressure anti-wear agent, 1%–8% of the rust inhibitor, 0.5%–5% of the pH stabilizer, 0.05%–1% of the bactericide, 0.01%–0.5% of the defoamer, with the balance being water; wherein the composite emulsifier is formed by compounding anionic and nonionic surfactants, and the extreme pressure anti-wear agent is a chlorine-free, low-sulfur, or boron-containing environmentally friendly extreme pressure anti-wear agent.

[0057] In this embodiment, the target formulation is an emulsified cutting fluid concentrate formulation; the preferred ranges of each component by mass percentage are as follows: 15%–45% cutting fluid base oil, 8%–20% composite emulsifier, 3%–12% oiliness agent, 1%–10% extreme pressure anti-wear agent, 1%–8% rust inhibitor, 0.5%–5% pH stabilizer, 0.05%–1% bactericide, 0.01%–0.5% defoamer, and the balance being water.

[0058] The base oil is the core oil phase component of the emulsified cutting fluid. When its content is below 15%, the lubricating performance is insufficient; when it is above 45%, the emulsification stability and cooling properties deteriorate. The composite emulsifier is formed by blending anionic surfactants (such as sodium petroleum sulfonate) and nonionic surfactants (such as fatty alcohol polyoxyethylene ether and sorbitan monooleate) in a certain proportion, with a total content of 8%–20%, to provide sufficient emulsification capacity and long-term storage stability. The oiliness agent is a synthetic ester, with a content of 3%–12%, used to form a physical adsorption film on the metal surface, providing boundary lubrication. The extreme pressure anti-wear agent is a chlorine-free, low-sulfur, or boron-containing environmentally friendly extreme pressure anti-wear agent, such as a boron-containing extreme pressure anti-wear agent. The compound system of pressure-reducing agent and low-sulfur ester, with a content of 1% to 10%, is used to form a chemical reaction film in the high-temperature and high-pressure cutting zone to prevent adhesive wear between the tool and the workpiece; the rust inhibitor is a compound system of organic carboxylate and alcohol amine borate, with a content of 1% to 8%, providing rust protection for ferrous and non-ferrous metals; the pH stabilizer is a triethanolamine and carbonate buffer system, with a content of 0.5% to 5%, maintaining the pH value of the diluted solution within a suitable range; the bactericide is isothiazolinone, with a content of 0.05% to 1%, inhibiting the growth of microorganisms; the defoamer is polyether-modified siloxane, with a content of 0.01% to 0.5%, controlling the generation of foam during processing.

[0059] As a preferred embodiment, the preparation of the formulation sample includes: heating the base oil of the cutting fluid to 35-50°C, adding oiliness agent, extreme pressure anti-wear agent, rust inhibitor and composite emulsifier in sequence, and stirring at 300-800 r / min for 10-20 min; then adding pH stabilizer, bactericide, defoamer and the remaining water, and continuing to stir for 15-30 min to obtain emulsified cutting fluid concentrate; the emulsified cutting fluid concentrate is diluted with water at a mass ratio of 1:(10-30) when used.

[0060] In this embodiment, the preparation process of the formulation sample is as follows: First, the base oil of the cutting fluid is heated to 35-50°C. This temperature range can effectively reduce the viscosity of the base oil, promote the dissolution and dispersion of subsequent additives, and at the same time, it will not cause the decomposition of heat-sensitive additives. Then, the oiliness agent, extreme pressure anti-wear agent, rust inhibitor and composite emulsifier are added in sequence, and stirred at a stirring speed of 300-800 r / min for 10-20 min. The mixing effect is not good when the stirring speed is lower than 300 r / min, and excessive air bubbles may be introduced when the stirring speed is higher than 800 r / min. By adding them in sequence, each additive can be fully dissolved in the oil phase and form a uniform oil phase mixture. After the system is uniform, the pH stabilizer, bactericide, defoamer and the balance water are added, and stirring is continued for 15-30 min. A stable water-in-oil emulsion concentrate is formed through the phase inversion emulsification process. When using the emulsion cutting fluid concentrate, it is diluted with water at a mass ratio of 1:(10-30). The choice of dilution ratio depends on the different requirements of cooling and lubrication under specific processing conditions.

[0061] As a preferred embodiment, the input features of the formulation-performance mapping model include at least the proportion of base oil, the proportion of compound emulsifier, the proportion of oiliness agent, the proportion of extreme pressure anti-wear agent, the proportion of rust inhibitor, the proportion of pH stabilizer, the proportion of bactericide, the proportion of defoamer, the type of processing material, the cutting speed, the feed rate, and the water hardness. The output feature is a comprehensive performance index, which is determined by weighting the cooling performance, rust prevention, lubrication, cleaning performance, stability, defoaming performance, and cost indicators. The wavelet neural network uses wavelet basis functions as hidden layer activation functions. The simulated annealing-particle swarm optimization algorithm uses dynamic inertia weights and annealing acceptance probability for global optimization, with the pH value of the diluent being between 7.5 and 9.5 as a constraint.

[0062] In this embodiment, the specific construction and optimization process of the recipe-performance mapping model is as follows: The input features of the training dataset include: base oil ratio, compound emulsifier ratio, oiliness agent ratio, extreme pressure anti-wear agent ratio, rust inhibitor ratio, pH stabilizer ratio, bactericide ratio, defoamer ratio, processing material type (e.g., 45 steel, HT250 cast iron, 6061 aluminum alloy, etc., represented by unique thermal encoding), cutting speed, feed rate, and water hardness (in CaCO3). The output feature is the comprehensive performance index F, satisfying F=w1·C+w2·R+w3·L+w4·Q+w5·S +w6·D-w7·K, where C, R, L, Q, S, D, and K represent the normalized scores of cooling, rust prevention, lubrication, cleaning, stability, defoaming, and cost indicators (each indicator is normalized to 0-100 points), and w1-w7 are the corresponding weighting coefficients. The weighting coefficients can be adjusted according to the emphasis of each performance in different application scenarios. For example, for scenarios mainly involving grinding, the weight of cooling can be appropriately increased; for scenarios mainly involving difficult machining processes such as tapping, the weight of lubrication can be appropriately increased.

[0063] The wavelet neural network adopts an input layer, two hidden layers, and an output layer structure. The number of nodes in the input layer is consistent with the dimension of the input features, which is 12 (corresponding to 12 input features). The first hidden layer has 14 nodes, and the second hidden layer has 7 nodes. The output layer has 1 node and outputs a comprehensive performance index F. The activation function of the hidden layer adopts the Morlet wavelet basis function, which has excellent time-frequency localization characteristics and can effectively approximate the complex high-dimensional nonlinear mapping relationship between recipe and performance. The model training adopts the backpropagation algorithm, the loss function is the mean squared error, and the parameters of the network weights and wavelet basis functions are iteratively updated by gradient descent until the model's performance on the validation set no longer improves or reaches the preset training rounds.

[0064] The simulated annealing-particle swarm optimization algorithm is implemented as follows: the position vector of each particle in the particle swarm represents a set of candidate recipes; the preferred number of particles is 40, and the maximum number of iterations is 120; the inertia weight ω adopts a linear decreasing strategy, starting from the initial value ω. max =0.90 linearly decreases to the termination value ω min =0.40, that is, ω(t)=ω max -(ω max -ω min )·t / T max Where t is the current iteration number, T max The maximum number of iterations is set to 2.0. This decreasing strategy gives the algorithm strong global exploration capabilities in the early stages of the search and strong local exploration capabilities in the later stages. Learning factors c1 and c2 are both set to 2.0, controlling the degree to which particles learn from their individual historical best position and the global best position, respectively. In each iteration, the particle velocity and position are updated according to the standard particle swarm optimization algorithm. Simulated annealing is also introduced: the energy difference ΔE (i.e., the difference in the comprehensive performance index F) between the new solution and the current global best solution is calculated. If the new solution is better than the global best solution, it is accepted unconditionally; if the new solution is worse than the global best solution... The solution is accepted with probability P=exp(-ΔE / T), where T is the current annealing temperature; the initial annealing temperature T0 is set to 5, the cooling coefficient λ is set to 0.93, and the temperature is reduced by T=λ·T after each iteration; the optimization constraints include: the pH value of the diluted solution is between 7.5 and 9.5, there is no obvious floating oil and sedimentation after 24 hours, the rust prevention test is qualified, the defoaming time is not greater than 10 seconds, and the unit cost of the concentrated solution is not higher than the preset value; by satisfying these constraints, it is ensured that the obtained optimal formula meets the actual application requirements in terms of comprehensive performance, stability in use and economy.

[0065] Secondly, this application also provides a machine learning-based system for preparing aged oil-based and water-based metal cutting fluids, comprising: The stability analysis module is used to obtain stability characterization parameters of aged oil, perform stability classification, and output preprocessing parameters; The pretreatment module is used to sequentially perform preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugation and solid-liquid separation to obtain the cutting fluid base oil; The formulation sample construction module is used to form formulation samples by compounding the cutting fluid base oil with various additives as the oil phase, and to collect the performance data of each sample; The machine learning modeling module is used to train a wavelet neural network using the data from the formula samples to establish a formula-performance mapping model. The optimization module is used to optimize the formula-performance mapping model using a simulated annealing-particle swarm optimization algorithm and output the target formula. The quality verification module is used to prepare aged oil-based and water-based metal cutting fluids according to the target formulation and to complete performance verification.

[0066] Corresponding to the above method embodiments, this invention also provides a machine learning-based system for preparing aged oil-based water-based metal cutting fluids; the system includes: The stability analysis module is used to obtain stability characterization parameters (including physical property parameters and microscopic stability parameters) of the aged oil to be treated. Based on the absolute value of zeta potential, interfacial film strength and median particle size, it constructs a stability evaluation index and performs stability classification. Based on the classification results, it outputs preprocessing parameters. This module can integrate the corresponding testing instrument data interface and analysis algorithm to automatically complete the stability determination.

[0067] The pretreatment module is used to perform synergistic demulsification, dehydration and purification treatment on aged oil in the order of preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugation and solid-liquid separation to obtain cutting fluid base oil; this module may include the control system and execution equipment of each process unit.

[0068] The formulation sample construction module is used to compound multiple formulation samples with cutting fluid base oil as the oil phase, and emulsifiers, oiliness agents, extreme pressure anti-wear agents, rust inhibitors, pH stabilizers, bactericides, and defoamers according to preset dosage ranges. It also collects data on the cooling properties, rust prevention properties, lubrication properties, cleaning properties, stability, salt tolerance, defoaming performance, and cost of each sample. This module may include formulation design tools and a performance testing database.

[0069] The machine learning modeling module is used to train a wavelet neural network using the composition and performance data of the recipe samples to establish a recipe-performance mapping model. This module may include a data preprocessing unit, a model training unit, and a model evaluation unit.

[0070] The optimization module is used to perform global optimization of the formula-performance mapping model using simulated annealing-particle swarm optimization algorithm. Under the premise of meeting the constraints of pH value, stability, rust prevention, defoaming time and cost of the diluent, it outputs the target formula. This module may include an optimization algorithm engine and a constraint management unit.

[0071] The quality verification module is used to prepare aged oil-based and water-based metal cutting fluids according to the target formula, and to verify the performance of the resulting products in accordance with standards such as SH / T0365-1992, and output a verification report.

[0072] The data flow and operation sequence between the modules correspond to the steps in the aforementioned method embodiments. Through the collaborative work of the modules, the system realizes the resource utilization of aged oil and the intelligent optimization of cutting fluid formulation.

[0073] Thirdly, this application provides an electronic device, including a memory and a processor, wherein the memory stores executable instructions, and the processor executes the executable instructions to implement the manufacturing method described above.

[0074] Fourthly, this application provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, implement the preparation method described above.

[0075] This invention also provides an electronic device, including a memory and a processor; the memory stores executable instructions, and the processor executes the executable instructions to implement the steps in the machine learning-based method for preparing aged oil-based and water-based metal cutting fluid described in any of the above embodiments; the electronic device may be an industrial computer, an embedded controller, or a cloud server; this invention also provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, implement the steps in the above method embodiments; the storage medium may be a non-volatile storage medium, such as ROM, flash memory, hard disk, etc.

[0076] Example 1: Pretreatment of Highly Stable Aging Oil and Preparation of Base Oil Oil from the settling tank of oilfield wastewater, the intermediate layer of oil and water in the emergency tank, and the bottom oil of the cleaning tank were selected as raw materials. The mass ratio of the three was 5:3:2. After mixing, 50L of aged oil to be treated was obtained. The aged oil was tested and found to have a density of 0.972g / cm³, a kinematic viscosity of 138mm² / s at 40℃, a water content of 35.6%, a solid content of 2.8%, an acid value of 3.9mgKOH / g, an absolute value of zeta potential of 41mV, a median droplet size D50 of 4.2μm, and an interfacial film strength characterization value IF of 0.86. Taking the weighting coefficients α=0.45, β=0.35, and γ=0.20, the stability evaluation index SI was calculated as SI=0.45×41+0.35×0.86×100+0.20×(1 / 4.2)×100≈53.6, which was judged to be a highly stable emulsion system.

[0077] For this highly stable system, the following parameters were used for synergistic demulsification and dehydration: First, the aged oil was heated to 62℃ and held for 18 min, then 15% low-viscosity base oil was added as a diluent; subsequently, it was ultrasonically treated at 30 kHz for 12 min with a power density of 0.45 W / mL; then it was introduced into an electrostatic coalescence unit and treated under an electric field strength of 3.5 kV / cm for 10 min; then it was centrifuged in a disc centrifuge at 4200 r / min for 15 min; finally, it was sequentially filtered through a 25 μm coarse filter and a 5 μm fine filter; after treatment, 26.8 kg of base oil was obtained, with a base oil recovery rate of 82.3%.

[0078] The treatment results showed that the water content of the aged oil decreased from 35.6% to 0.21%, the solid impurity content decreased from 2.8% to 0.028%, the acid value decreased from 3.9 mg KOH / g to 1.2 mg KOH / g, and the oil content in the wastewater was 37 ppm. When this base oil was used in subsequent emulsification tests, the concentrate prepared with a base oil content of 30% showed no obvious stratification after a 1:20 dilution, indicating good compatibility between the obtained base oil and the emulsion system. Compared with the control group that only used 60°C heating and static sedimentation, the water content of the base oil in this embodiment decreased by approximately 96.7%, and the solid content decreased by approximately 93.4%, demonstrating the significant advantages of multi-stage synergistic treatment for highly stable aged oil.

[0079] Example 2: Grading treatment of medium-stabilized aged oil and comparison of base oil quality The recovered oil from the land and the floating oil recovered from the sewage equalization tank were mixed at a mass ratio of 7:3 to obtain 40L of aged oil. Testing showed a density of 0.934 g / cm³, a kinematic viscosity of 82 mm² / s at 40℃, a water content of 24.8%, a solid content of 1.4%, an absolute zeta potential of 28 mV, a D50 of 7.6 μm, an interfacial film strength (IF) of 0.54, and an SI of approximately 32.4 calculated using the same formula as in Example 1, classifying it as a moderately stable emulsion system. Based on these classification results, the pretreatment intensity was appropriately reduced.

[0080] The specific process parameters are as follows: preheating temperature 55℃, holding time 12 min, diluent addition 10%; ultrasonic frequency 28kHz, treatment time 8 min, power density 0.30W / mL; electrostatic coalescence electric field strength 2.5kV / cm, treatment time 6 min; horizontal screw centrifuge speed 3500r / min, separation time 10 min; two-stage filtration at the end: 30μm and 8μm; after treatment, 24.5kg of base oil was obtained, with a base oil recovery rate of 80.6%; the water content of the base oil was 0.27%, the solid impurity content was 0.041%, and the oil content in the external drainage was 44ppm.

[0081] To verify the effectiveness of stability grading and parameter matching, Comparative Example A was set up: the high-intensity process of Example 1 was directly applied to this stable sample; Comparative Example B was set up: a lower-intensity process (50℃ preheating, no ultrasound, 2kV / cm electrostatic coalescence, 3000r / min centrifugation) was used. The results showed that although Comparative Example A could reduce the water content to 0.20%, the energy consumption per ton of oil processed increased by about 17.8%, resulting in poor economic efficiency; after treatment in Comparative Example B, the water content of the base oil was still 0.92%, and the fine filtration pressure difference increased significantly, affecting continuous operation; This example, while ensuring the quality of the base oil, reduced the unit processing cost by about 11.6% compared to Comparative Example A, proving that selecting process parameters according to stability level is more practical for engineering applications.

[0082] Example 3: Formulation Sample Construction, Model Training, and Determination of Optimal Formulation Using the base oils prepared in Examples 1 and 2, formulation samples were designed with the following proportions: base oil 15%–45%, composite emulsifier 8%–20%, oiliness agent 3%–12%, extreme pressure anti-wear agent 1%–10%, rust inhibitor 1%–8%, pH stabilizer 0.5%–5%, bactericide 0.05%–1%, and defoamer 0.01%–0.5%. A total of 96 emulsified cutting fluid concentrate samples were prepared, of which 72 were used as the training set, 12 as the validation set, and 12 as the test set. During sample preparation, the base oil was first heated to 42°C, and the oiliness agent, extreme pressure anti-wear agent, rust inhibitor, and composite emulsifier were added sequentially, and stirred at 500 r / min for 15 min. Then, the pH stabilizer, bactericide, defoamer, and the remaining deionized water were added, and stirring was continued for 20 min to obtain the concentrate. After the concentrate was allowed to stand for 24 h, samples were taken and diluted 1:20 with tap water with a hardness of 200 mg / L for performance evaluation.

[0083] The composite emulsifier is composed of sodium petroleum sulfonate, fatty alcohol polyoxyethylene ether, and sorbitan monooleate; the oiliness agent is a synthetic ester; the extreme pressure anti-wear agent is a compound system of boron-containing extreme pressure agent and low-sulfur ester; the rust inhibitor is a compound system of organic carboxylate and alcohol amine borate; the pH stabilizer is a triethanolamine and carbonate buffer system; the bactericide is isothiazolinone; and the defoamer is polyether-modified siloxane.

[0084] The test indicators include: cooling performance (characterized by peak temperature drop in the grinding zone), rust prevention performance (single and stacked cast iron tests), lubrication performance (maximum non-seizure load PB and wear scar diameter of four balls), cleaning performance (standard dirt residue rate), stability (24h / 7d stratification), allowable salt content, defoaming time, and unit concentrated solution cost; the weights of the comprehensive performance index F are set as follows: cooling performance w1=0.22, rust prevention performance w2=0.18, lubrication performance w3=0.20, cleaning performance w4=0.12, stability w5=0.12, defoaming performance w6=0.06, and cost index w7=0.10; all individual indicators are normalized to 0-100 points.

[0085] Before modeling, the input features were normalized to the interval [-1, 1], and one-hot encoding was used for the processing material category. The wavelet neural network adopted a 12-14-7-1 structure, that is, 12 input nodes, two hidden layers with 14 and 7 nodes respectively, and 1 output node. The Morlet wavelet function was selected as the activation function of the hidden layer. After training, the coefficient of determination R² of the training set was 0.973, the R² of the validation set was 0.948, the R² of the test set was 0.941, and the root mean square error (RMSE) of the test set was 2.87, indicating that the model has high prediction accuracy for the comprehensive performance index.

[0086] During the optimization phase, the particle count was set to 40, the maximum number of iterations to 120, the initial inertia weight ωmax to 0.90, the final inertia weight ωmin to 0.40, the learning factors c1=c2=2.0, the initial annealing temperature T0=5, the cooling coefficient λ=0.93, and the constraints were set as follows: the pH of the diluted solution was 7.8~9.2, there was no obvious floating oil and sedimentation after 24 hours, the rust prevention test was qualified, the defoaming time was no more than 10s, and the unit cost of the concentrated solution was no more than 12.5 yuan / kg; After iteration, a set of optimal concentrate formulations were obtained: 29.5% base oil, 12.8% composite emulsifier, 6.2% oiliness agent, 4.8% extreme pressure anti-wear agent, 3.4% rust inhibitor, 1.6% pH stabilizer, 0.18% bactericide, 0.08% defoamer, and the balance being water. The model predicted an F-value of 89.7 for this formulation, while the measured F-value was 88.9, with a prediction error of only 0.89%. This demonstrates that the formulation-performance mapping model and optimization method established in this invention have good accuracy and practicality.

[0087] Example 4: Pilot-scale amplification and processing application verification of the optimal formulation A 200kg pilot-scale test was conducted according to the preferred formulation obtained in Example 3. The specific steps were as follows: 59.0kg of base oil was added to a 300L jacketed stirred tank and heated to 40°C; 25.6kg of composite emulsifier, 12.4kg of oiliness agent, 9.6kg of extreme pressure anti-wear agent, and 6.8kg of rust inhibitor were added sequentially and stirred at 550r / min for 15min; then 3.2kg of pH stabilizer, 0.36kg of bactericide, 0.16kg of defoamer, and 83.0kg of deionized water were added, and stirring was continued for 22min. After filtration, the final concentrate was obtained. The concentrate was a uniform brownish-yellow semi-transparent liquid. After storage at 25°C for 30 days, there was no stratification or obvious sediment, indicating that the concentrate had good storage stability.

[0088] The concentrate was diluted 1:20 with process water with a hardness of 180 mg / L and used for continuous machining verification in turning of 45 steel, milling of HT250 cast iron, and drilling of 6061 aluminum alloy. Two comparative products were used: Comparative Sample 1 was an emulsified cutting fluid with the same formula but using commercial white oil instead of the base oil of this invention; Comparative Sample 2 was a commercially available general-purpose emulsified cutting fluid. The results after 10 days of continuous operation are as follows: In this embodiment, the pH of the diluted solution was maintained at 8.6–8.9, which is within the specified range of 7.5–9.5; the defoaming time was 4 seconds, which meets the constraint of no more than 10 seconds; the allowable salt content was 3.5%, indicating good hard water resistance; no rust spots were found in the 24-hour rust prevention test of cast iron, indicating that the rust prevention performance was qualified.

[0089] Under the turning conditions of 45 steel, the average tool life using the cutting fluid of this embodiment was 118 min, which was 16.8% higher than that of the control sample 2; under the milling conditions of HT250, the average temperature of the main cutting zone was 11.4℃ lower than that of the control sample 2; under the drilling conditions of 6061 aluminum alloy, the surface roughness Ra of the workpiece was stable at 0.82 μm, which was 12.8% lower than that of the control sample 2. The above results show that the aged oil-based water-based metal cutting fluid prepared in this embodiment is superior to commercially available general products in terms of cooling performance, lubrication and improvement of the surface quality of the machined parts.

[0090] Further analysis of overall economic efficiency: The production cost of the concentrated solution in this embodiment is 11.3 yuan / kg, compared to 12.1 yuan / kg for control sample 1 and 12.9 yuan / kg for control sample 2. Based on the same processing volume, the unit workpiece processing fluid cost in this embodiment is reduced by approximately 6.6% compared to control sample 1 and by approximately 12.4% compared to control sample 2. Furthermore, after 30 days of repeated use, the diluted solution did not exhibit any obvious putrid odor, and the total bacterial count was controlled at 1.2 × 10⁻⁶. 4 A concentration below CFU / mL indicates that the cutting fluid has good stability and field applicability.

[0091] As can be seen from the above embodiments, the present invention, through the technical path of quantitative analysis of the stability of aged oil—multi-stage synergistic demulsification and dehydration—machine learning modeling—simulated annealing—particle swarm optimization—pilot-scale application verification, can not only transform aged oil with high water content and high emulsification stability into base oil that meets the compatibility requirements of cutting fluid, but also obtain an optimal formula that takes into account cooling, rust prevention, lubrication, cleaning, stability, defoaming and cost within a large formula space, and has strong data support and engineering feasibility.

[0092] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0093] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for preparing aged oil-based and water-based metal cutting fluid based on machine learning, characterized in that, include: Obtain the stability characterization parameters of the aged oil to be treated, classify the stability of the aged oil based on the stability characterization parameters, and determine the pretreatment parameters based on the classification results. The aged oil is subjected to preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugal separation and solid-liquid separation in sequence to obtain the cutting fluid base oil. Using the cutting fluid base oil as the oil phase, multiple formulation samples were constructed by compounding it with emulsifiers, oiliness agents, extreme pressure anti-wear agents, rust inhibitors, pH stabilizers, bactericides, and defoamers. The corresponding cooling properties, rust prevention properties, lubrication properties, cleaning properties, stability, salt tolerance, defoaming performance, and cost indicators of each formulation sample were obtained. A wavelet neural network was trained using the composition and performance data of the aforementioned formula samples to establish a formula-performance mapping model; The simulated annealing-particle swarm optimization algorithm is used to optimize the formula-performance mapping model to obtain the target formula; An aged oil-based water-based metal cutting fluid was prepared according to the target formulation, and its performance was verified.

2. The preparation method according to claim 1, characterized in that, The stability characterization parameters include physical property parameters and microscopic stability parameters. The physical property parameters include at least density, viscosity, water content, and solid content. The microscopic stability parameters include at least water droplet size distribution, median particle size, zeta potential, interfacial film strength, and stratification behavior under temperature changes. The stability grading includes: constructing a stability evaluation index based on the absolute value of zeta potential, the characterization value of interfacial film strength, and the median particle size. When the stability evaluation index reaches a preset threshold, the aged oil is determined to be a highly stable emulsion system, and at least one of the following is increased accordingly: dilution ratio, ultrasonic intensity, and electrostatic coalescence time.

3. The preparation method according to claim 1, characterized in that, The process conditions for the preheating and diluting step are as follows: heating to 45-70°C, holding for 10-30 minutes, and adding diluent at a rate of 5%-25% of the mass of the aged oil. The ultrasonic treatment process conditions are: frequency 20-40kHz, power density 0.2-1.0W / mL, treatment time 5-30min; The electrostatic coalescence process conditions are: electric field strength 1-6 kV / cm, treatment time 3-20 min; The centrifugal separation process conditions are: rotation speed 2000-8000 r / min, separation time 5-30 min; The solid-liquid separation employs a two-stage filtration process, with the first stage having a filtration accuracy of 20–50 μm and the second stage having a filtration accuracy of 1–10 μm.

4. The preparation method according to claim 1, characterized in that, The resulting cutting fluid base oil has a water content of no more than 0.3% and a solid impurity content of no more than 0.05%.

5. The preparation method according to claim 1, characterized in that, The target formulation is an emulsified cutting fluid concentrate formulation, comprising, by mass percentage: 15%–45% of the cutting fluid base oil, 8%–20% of the composite emulsifier, 3%–12% of the oiliness agent, 1%–10% of the extreme pressure anti-wear agent, 1%–8% of the rust inhibitor, 0.5%–5% of the pH stabilizer, 0.05%–1% of the bactericide, 0.01%–0.5% of the defoamer, with the balance being water; wherein, the composite emulsifier is formed by compounding anionic and nonionic surfactants, and the extreme pressure anti-wear agent is a chlorine-free, low-sulfur, or boron-containing environmentally friendly extreme pressure anti-wear agent.

6. The preparation method according to claim 5, characterized in that, The preparation of the formulation sample includes: heating the base oil of the cutting fluid to 35-50°C, adding oiliness agent, extreme pressure anti-wear agent, rust inhibitor and composite emulsifier in sequence, and stirring at 300-800 r / min for 10-20 min; then adding pH stabilizer, bactericide, defoamer and the balance water, and continuing to stir for 15-30 min to obtain emulsified cutting fluid concentrate; the emulsified cutting fluid concentrate is diluted with water at a mass ratio of 1:(10-30) when used.

7. The preparation method according to claim 1, characterized in that, The input features of the formulation-performance mapping model include at least the proportion of base oil, the proportion of compound emulsifier, the proportion of oiliness agent, the proportion of extreme pressure anti-wear agent, the proportion of rust inhibitor, the proportion of pH stabilizer, the proportion of bactericide, the proportion of defoamer, the type of processed material, the cutting speed, the feed rate, and the water hardness. The output feature is a comprehensive performance index, which is determined by weighting the cooling performance, rust prevention, lubrication, cleaning performance, stability, defoaming performance, and cost indicators. The wavelet neural network uses wavelet basis functions as hidden layer activation functions. The simulated annealing-particle swarm optimization algorithm uses dynamic inertia weights and annealing acceptance probability for global optimization, with the pH value of the diluent being between 7.5 and 9.5 as a constraint.

8. A machine learning-based system for preparing aged oil-based and water-based metal cutting fluids, characterized in that, include: The stability analysis module is used to obtain stability characterization parameters of aged oil, perform stability classification, and output preprocessing parameters; The pretreatment module is used to sequentially perform preheating and dilution, ultrasonic treatment, electrostatic coalescence, centrifugation and solid-liquid separation to obtain the cutting fluid base oil; The formulation sample construction module is used to form formulation samples by compounding the cutting fluid base oil with various additives as the oil phase, and to collect the performance data of each sample; The machine learning modeling module is used to train a wavelet neural network using the data from the formula samples to establish a formula-performance mapping model. The optimization module is used to optimize the formula-performance mapping model using a simulated annealing-particle swarm optimization algorithm and output the target formula. The quality verification module is used to prepare aged oil-based and water-based metal cutting fluids according to the target formulation and to complete performance verification.

9. An electronic device comprising a memory and a processor, the memory storing executable instructions, characterized in that, When the processor executes the executable instructions, it implements the preparation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having executable instructions stored thereon, characterized in that, When the executable instructions are executed by the processor, they implement the preparation method as described in any one of claims 1 to 7.

Images