Machine tool cutting fluid circulating filtration intelligent control method and system

Abstract

The invention discloses an intelligent control method and system for machine tool cutting fluid circulating filtration, relates to the technical field of cutting fluid filtration control, and can pre-judge a pollution critical state and improve control perspectiveness by calculating a pollution evolution trend function. Unnecessary frequent filtering is avoided, and the service cycle of equipment parts is effectively prolonged; the cleanliness of the cutting fluid is guaranteed through real-time filtering control, and chain anomalies such as machining burn and aggravation of tool abrasion are prevented; iterative optimization of system parameters is realized through a feedback learning correction module, so that the system automatically adapts to an optimal filtering strategy in different processing scenes, and the system carries out proportional dynamic correction on a factor calibration coefficient and a periodic compression regulation factor lambda used in pollution trend modeling through a parameter adaptive correction unit. Lambda is moderately strengthened, so that the next round of filtering responds more positively. The mechanism endows the system with evolutionary ability and adaptive regulation and control ability, and can gradually optimize a self control strategy in a complex and uncertain processing scene.

Description

Technical Field

[0001] The present invention relates to the technical field of chip fluid filtration control, and in particular to an intelligent control method and system for circulating filtration of chip fluid in machine tools. Background Art

[0002] In the modern industrial landscape, where high-precision equipment such as CNC lathes, vertical machining centers, and multi-tasking machine tools are widely used, cutting fluid not only serves as a cooling and lubricating medium but also plays a key role in machine tool operational stability, surface finish quality, and tool life control. To ensure long-lasting performance and controllable fluid cleanliness, a cutting fluid filtration and circulation unit is typically required to remove contaminants such as metal particles, emulsified debris, and oily foam during machining, thereby enabling continuous, high-quality reuse of the cutting fluid.

[0003] Currently, most common machine tool chip fluid filtration systems utilize a periodic filtration logic with a constant time interval, triggering the pumping and filtration process at a fixed cycle. However, this strategy seriously overlooks the dynamic nature of actual contaminant accumulation. The rate and form of contamination vary significantly under different operating conditions and fluid compositions. This can lead to wasted energy and filter life if filtration is performed too early, while delayed filtration can lead to filtration lag and fluid degradation.

[0004] The root cause of these issues lies in the existing system's failure to establish an intelligent, adaptive connection between contamination status and filtration cycles—in other words, a cyclical, self-adjusting model based on the dynamic evolution of contamination. Because the chip fluid contamination process is characterized by significant nonlinearity, multi-stage, and multi-factor coupling, the continued use of fixed filtration cycle control logic can easily lead to system-wide, cascading problems during periods of rapid contamination accumulation, including a sharp increase in liquid viscosity, imbalanced interfacial tension, increased pump resistance, abnormal fluid return, burns on machined surfaces, and a dramatic reduction in tool life. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides an intelligent control method and system for circulating and filtering cutting fluid of machine tools, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent control system for circulating filtration of cutting fluid in machine tools, comprising a data acquisition and time series preprocessing module, a pollution state factor extraction module, a pollution trend evolution modeling module, a filtering cycle adaptive scheduling module, an intelligent control execution module, and a feedback and learning correction module;

[0007] The data acquisition and time series preprocessing module collects the cutting fluid data through the sensor installed on the machine tool, fits it into the original data set W, and performs preprocessing to obtain the filtered data set WG;

[0008] The pollution state factor extraction module extracts features from the obtained filtered data set WG, obtains the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, and fits them into the pollution feature set FRw;

[0009] The pollution trend evolution modeling module analyzes the acquired pollution feature set FRw and constructs a pollution trend function Ψ(t) that comprehensively reflects the change of pollution severity over time;

[0010] The filtering cycle adaptive scheduling module converts the period of the pollution trend index Ψ(t) into the actual filtering period ΔTfter, replacing the fixed period logic;

[0011] The intelligent control execution module triggers the actual pump control action according to the actual filtration cycle ΔTfter, realizes the cycle control closed loop, and obtains the post-filtration pollution trend function nΨ;

[0012] The feedback and learning correction module compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ, and compares it with the preset pollution deviation threshold TΨ to judge the filtering effect.

[0013] Preferably, the data acquisition and time series preprocessing module includes a pollution parameter acquisition unit and a data standardization processing unit;

[0014] The pollution parameter acquisition unit collects the cutting fluid data through sensors, including liquid flow rate Vy, liquid color Sy, viscosity Ny, conductivity Cy, foam interference coefficient FIO, particle disturbance density Rk, emulsification separation speed Ve and interfacial tension disturbance Dg, and fits them into the original data set W;

[0015] The liquid flow rate Vy is collected and obtained by a turbine flowmeter installed in the filtration loop pipe section;

[0016] The body color Sy is collected and acquired through an online color sensor installed in the transparent window of the return liquid pipe section;

[0017] The viscosity Ny is collected by a micro-thermal impedance viscometer installed in a low-flow bypass loop;

[0018] The conductivity Cy is collected and obtained through the conductivity electrode sensor installed at the outlet of the liquid tank circuit;

[0019] The foam interference coefficient FIO is acquired through the image recognition module and light reflection sensor installed in the irradiation area above the liquid surface;

[0020] The particle disturbance density Rk is acquired by collecting the fluid pressure micro-disturbance transducer;

[0021] The emulsification separation speed Ve is acquired through image processing algorithm;

[0022] The interfacial tension perturbation Dg is acquired by microbubble image analysis and high-frequency light scattering;

[0023] The data standardization processing unit performs denoising, sliding average and standardization on the original data set W to obtain the filtered data set WG;

[0024] Denoising is done by removing the noise from the original dataset W using the median filter method;

[0025] Sliding average removes local spike interference by using a sliding window for smoothing;

[0026] By using the standardization method, the data of different dimensions in the original data set W are converted to a unified standard scale to obtain the filtered data set WG;

[0027] The filtered dataset WG is obtained by the following formula:

[0028]

[0029] Where WGo represents the o-th data item in the filtered dataset WG, Wo represents the o-th data item in the original dataset W, μWo represents the mean of the o-th data item in the original dataset W, and σ represents the standard deviation of the o-th data item in the original dataset W.

[0030] Preferably, the pollution state factor extraction module includes a multi-parameter coupling feature construction unit and a pollution feature set acquisition unit;

[0031] The multi-parameter coupling feature construction unit performs nonlinear combination and differential operations on the data in the filtered data set WG to construct a dynamic coupling relationship and extract pollution sensitivity indicators, including the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3;

[0032] The particle disturbance index FR1 is obtained by the following formula:

[0033]

[0034] Where Rk(t) represents the particle disturbance density at time t, Dg(t) represents the interfacial tension disturbance at time t, c represents the perturbation stability constant to prevent the denominator from being zero, β and π represent the coupling term adjustment coefficients, β = 0.2, π = 3.14, FIO(t) represents the foam interference coefficient at time t, and d represents the differential sign;

[0035] The emulsion stability degradation function FR2 is obtained by the following formula:

[0036]

[0037] Wherein, Ve(t) represents the emulsification separation rate at time t, log represents the logarithmic function, Ny(t) represents the viscosity at time t, and Cy(t) represents the conductivity at time t;

[0038] The tension fluctuation intensity FR3 is obtained by the following formula:

[0039]

[0040] Where T represents the time window length, Dg(t) represents the interfacial tension disturbance at time t, and μDg represents the average tension disturbance value of the interfacial tension disturbance;

[0041] The pollution feature set acquisition unit fits the acquired particle disturbance index FR1, emulsion stability degradation function FR2 and tension fluctuation intensity FR3 to acquire the pollution feature set FRw.

[0042] Preferably, the pollution trend evolution modeling module includes a feature dynamic integration unit and a pollution trend function calculation unit;

[0043] The feature dynamic normalization unit performs standard normalization processing on the obtained pollution feature set FRw to obtain the pollution normalization set nFRw;

[0044] The pollution normalization set nFRw is obtained by the following formula:

[0045]

[0046] Where nFRwi represents the i-th feature in the pollution normalization set nFRw, FRwi represents the i-th feature in the pollution feature set FRw, μFRwi represents the mean of the i-th feature in the pollution feature set FRw, and σFRwi represents the standard deviation of the i-th feature in the pollution feature set FRw.

[0047] Preferably, the pollution trend function calculation unit constructs indicators for the features in the pollution normalization set nFRw and calculates and obtains the pollution trend function Ψ(t);

[0048] The pollution trend function Ψ(t) is obtained by the following formula:

[0049]

[0050] Where α1, α2 and α3 represent the factor calibration coefficients of the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, respectively; xt represents the integral variable; FR1(xt) represents the particle disturbance index at time xt; FR2(xt) represents the emulsion stability degradation function at time xt; and FR3(xt) represents the tension fluctuation intensity at time xt.

[0051] The current pollution status is judged by the obtained pollution trend function Ψ(t);

[0052] When 0<pollution trend function Ψ(t)<0.5, it means the current pollution state is normal;

[0053] When 0.5≤pollution trend function Ψ(t)<1.0, it means that the current pollution state is abnormal and needs to be filtered.

[0054] Preferably, the filtering cycle adaptive scheduling module performs mapping conversion processing on the pollution trend index Ψ(t) according to the numerical interval and change form thereof, corresponding to the input scale required by the periodic control function, and obtains the pollution normalization index nΨ;

[0055] The pollution normalization index GYΨ is obtained by the following formula:

[0056]

[0057] Where Ψmin represents the valley value of the pollution trend index, and Ψmax represents the peak value of the pollution trend index;

[0058] According to the obtained pollution normalization index GYΨ, a nonlinear periodic function is constructed to adjust the basic period Tbase to obtain the actual filtration period ΔTfter;

[0059] The actual filtration period ΔTfter is obtained by the following formula:

[0060]

[0061] Where Tbase represents the base period, and λ represents the period compression adjustment factor.

[0062] Preferably, the intelligent control execution module includes a filtration cycle control unit and a post-filtration pollution trend feedback unit;

[0063] The filtration cycle control unit compares the current running time Tnew with the last filtration time Tlast to determine whether the actual filtration cycle ΔTfter has been reached;

[0064] The judgment formula is as follows:

[0065]

[0066] Where Pctr(t) represents the pump status instruction, control output: 1 for start, 0 for hold or stop

[0067] After the filter pump completes one action, the post-filtration contamination trend feedback unit monitors the new contamination trend data in real time, including the new particle disturbance index nFR1, the new emulsification stability degradation function nFR2, and the new tension fluctuation intensity nFR3, and constructs the post-filtration contamination trend function nΨ;

[0068] The pollution trend function nΨ after filtering is obtained by the following formula:

[0069]

[0070] Where Ts represents the short-term feedback window.

[0071] Preferably, the feedback and learning correction module includes a pollution deviation determination unit and a parameter adaptive correction unit;

[0072] The pollution deviation determination unit compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ;

[0073] The contamination deviation ΔΨ is obtained by the following formula:

[0074] ΔΨ=Ψ(t)-nΨ;

[0075] Compare the obtained pollution deviation ΔΨ with the preset pollution deviation threshold TΨ to judge the filtering effect;

[0076] When the contamination deviation ΔΨ ≥ the contamination deviation threshold TΨ, it means that the filtering is effective;

[0077] When the contamination deviation ΔΨ is less than the contamination deviation threshold TΨ, it means that the filtering is invalid, a correction signal is output, and the parameter correction process is started.

[0078] Preferably, the parameter adaptive correction unit corrects the parameters according to the correction signal, including correcting the factor calibration coefficients α1, α2 and α3 and correcting the periodic compression adjustment factor λ;

[0079] Correcting the factor calibration coefficients α1, α2, and α3 of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3 to obtain new factor calibration coefficients nα1, nα2, and nα3 of the factor calibration coefficients of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3;

[0080] The correction formula is as follows:

[0081]

[0082] Where, α i represents the calibration coefficient of the i-th factor, and η represents the sensitivity adjustment factor;

[0083] Correct the periodic compression adjustment factor λ to obtain a new periodic compression adjustment factor nλ;

[0084] The new period compression adjustment factor nλ is obtained by the following formula:

[0085]

[0086] An intelligent control method for circulating and filtering cutting fluid of a machine tool comprises the following steps:

[0087] Step 1: The data acquisition and time series preprocessing module collects the cutting fluid data through the sensor installed on the machine tool, fits it into the original data set W, and performs preprocessing to obtain the filtered data set WG;

[0088] Step 2: The pollution state factor extraction module extracts features from the obtained filtered data set WG, obtains the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, and fits them into the pollution feature set FRw;

[0089] Step 3: The pollution trend evolution modeling module analyzes the obtained pollution feature set FRw and constructs a pollution trend function Ψ(t) that comprehensively reflects the change of pollution severity over time;

[0090] Step 4: The filtering cycle adaptive scheduling module converts the period of the pollution trend index Ψ(t) into the actual filtering period ΔTfter, replacing the fixed period logic;

[0091] Step 5: The intelligent control execution module triggers the actual pump control action according to the actual filtration cycle ΔTfter, realizes the cycle control closed loop, and obtains the post-filtration pollution trend function nΨ;

[0092] Step 6: The feedback and learning correction module compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ, and compares it with the preset pollution deviation threshold TΨ to determine the filtering effect.

[0093] The present invention provides an intelligent control method and system for circulating and filtering cutting fluid in machine tools, which has the following beneficial effects:

[0094] (1) When the system is running, it can dynamically adjust the filtration frequency according to the actual pollution changes, significantly reducing the energy waste caused by excessive filtration or the liquid degradation problem caused by delayed filtration; by calculating the pollution evolution trend function, it can predict the critical state of pollution in advance and improve the control foresight; avoid unnecessary frequent filtration and effectively extend the service life of equipment components; real-time filtration control ensures the cleanliness of the cutting fluid and prevents chain abnormalities such as processing burns and increased tool wear; through the feedback learning correction module, it realizes the iterative optimization of system parameters, so that it can automatically adapt to the best filtration strategy in different processing scenarios.

[0095] (2) The system acquisition method integrates embedded, low-cost sensing solutions such as image recognition modules, micro-electro-thermal impedance viscometers, fluid perturbation transducers, and high-frequency light scattering meters without relying on the addition of expensive hardware. Combined with image processing algorithms, it obtains indirect contamination parameters such as the emulsification separation velocity Ve and interfacial tension perturbation Dg, achieving accurate parameter modeling. This approach not only makes the system scalable and low-cost, but also adapts to a variety of cutting fluid formulations and complex processing environments, effectively solving the problem of poor adaptability of traditional systems.

[0096] (3) A highly sensitive and structured pollution feature recognition system is implemented to improve the response speed to early changes in pollution evolution; a multi-dimensional pollution path expression model is constructed to improve the comprehensiveness and accuracy of the pollution status description; the stability of the characteristic indicators in expressing time series trends is enhanced to lay a solid foundation for subsequent trend modeling and control scheduling; the probability of misjudgment and missed judgment is reduced to avoid the control error caused by the rough setting of thresholds in traditional systems; the interpretability and traceability of pollution judgment are improved to facilitate system anomaly analysis and fault backtracking.

[0097] (4) The problem of unbalanced response of pollution trends in different numerical intervals is solved, so that the pollution trend can be mapped to the standardized input range of the periodic function regardless of whether it is in a mild or severe fluctuation stage, realizing high-resolution control of period scaling. This avoids the hysteresis caused by the traditional "single numerical threshold control" and improves the real-time sensitivity of period regulation to the pollution status. In the scheduling strategy, linear scaling or fixed time interval logic is no longer used. Instead, the nonlinear combination adjustment mechanism of the pollution normalization index and the period compression factor is used to achieve intelligent adjustment of the basic period Tbase. This method can not only quickly compress the filtration cycle in the stage of rapid growth of the pollution trend, but also automatically slow down the filtration frequency in the stage of slow evolution or stability of pollution, effectively balancing the contradiction between filtration efficiency and system resource consumption, significantly extending the life of the filter element and reducing energy usage costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 This is a flow chart of the intelligent control system for circulating and filtering cutting fluid in machine tools according to the present invention;

[0099] Figure 2 Schematic diagram of the steps of the intelligent control method for circulating and filtering cutting fluid of machine tools according to the present invention;

[0100] Figure 3 A schematic diagram of a process for obtaining a pollution trend function according to the present invention;

[0101] Figure 4 A bar graph showing the pollution trend function of the present invention. DETAILED DESCRIPTION

[0102] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0103] The present invention provides an intelligent control system for circulating and filtering cutting fluid of machine tools, such as Figures 1 to 4 As shown, it includes data acquisition and time series preprocessing module, pollution state factor extraction module, pollution trend evolution modeling module, filtering cycle adaptive scheduling module, intelligent control execution module and feedback and learning correction module.

[0104] The data acquisition and time series preprocessing module collects cutting fluid data using sensors installed on the machine tool, fitting it into a raw data set W, and performing preprocessing to obtain a filtered data set WG. The pollution state factor extraction module extracts features from the filtered data set WG, obtaining the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3, and fitting them into the pollution feature set FRw. The pollution trend evolution modeling module analyzes the acquired pollution feature set FRw and constructs a pollution trend function Ψ(t) that comprehensively reflects the temporal evolution of pollution severity. The filtration cycle adaptive scheduling module converts the period of the pollution trend index Ψ(t) into an actual filtration cycle ΔTfter, replacing the fixed cycle logic. The intelligent control execution module triggers the actual pump control action based on the actual filtration cycle ΔTfter, implementing a closed-loop periodic control and obtaining the post-filtration pollution trend function nΨ. The feedback and learning correction module compares the acquired pollution trend function Ψ(t) with the post-filtration pollution trend function nΨ to obtain the pollution deviation ΔΨ. This is then compared with the preset pollution deviation threshold TΨ to determine the filtering effectiveness.

[0105] In this embodiment, based on the actual development trend of the cutting fluid contamination state, a contamination trend function Ψ(t) is constructed to achieve real-time adaptive updating of the filtration period ΔTfter. This improvement overcomes the drawback of the existing logic being disconnected from the contamination state, enabling the system to "adjust the filtration frequency as the contamination evolves," fundamentally improving control accuracy.

[0106] By collecting data from existing sensors and fusing and extracting deep pollution factors such as the particle disturbance index, emulsification stability degradation function, and tension fluctuation intensity, multi-dimensional modeling of pollution development can be achieved without increasing any hardware costs. This "soft perception + dynamic modeling" approach achieves higher cost-effectiveness, stronger adaptability, and a lower threshold for system transformation. By collecting a new pollution trend function nΨ(t) after filtration and comparing it with the original pollution trend function Ψ(t), and calculating the pollution deviation ΔΨ, the system can determine whether the filtration effect meets the standards and adaptively optimize the trend modeling parameters and cycle control parameters accordingly. This "self-evaluation + self-correction" feedback mechanism builds a true closed-loop intelligent control system, effectively overcoming the inherent defects of traditional systems of "lack of feedback and non-adjustability."

[0107] The system dynamically adjusts filtration frequency based on actual contamination changes, significantly reducing energy waste from over-filtration and fluid degradation caused by delayed filtration. By calculating the contamination evolution trend function, it can predict critical contamination states in advance, improving control foresight, avoiding unnecessary frequent filtration, and effectively extending the life cycle of equipment components. Real-time filtration control ensures the cleanliness of the cutting fluid and prevents chain reactions such as machining burns and increased tool wear. A feedback learning correction module enables iterative optimization of system parameters, automatically adapting to the optimal filtration strategy for different machining scenarios.

[0108] Furthermore, the data acquisition and time series preprocessing module includes a pollution parameter acquisition unit and a data standardization processing unit. The pollution parameter acquisition unit collects cutting fluid data (including liquid flow rate Vy, liquid color Sy, viscosity Ny, conductivity Cy, foam interference coefficient FIO, particle disturbance density Rk, emulsification separation velocity Ve, and interfacial tension disturbance Dg) through sensors and fits them into the original data set W.

[0109] The liquid flow rate Vy is acquired by a turbine flowmeter installed in the filtration loop. The liquid color Sy is acquired by an online color sensor installed in the transparent window of the return liquid pipe. The viscosity Ny is acquired by a micro-thermal impedance viscometer installed in the low-flow bypass loop. The conductivity Cy is acquired by a conductivity electrode sensor installed at the outlet of the tank loop. The foam interference coefficient Fio is acquired by an image recognition module and a light reflectance sensor installed in the irradiated area above the liquid surface. The particle disturbance density Rk is acquired by a fluid pressure micro-perturbation transducer. The emulsification separation velocity Ve is acquired using an image processing algorithm. The interfacial tension disturbance Dg is acquired by microbubble image analysis and a high-frequency light scatterometer. The data normalization unit performs denoising, sliding averaging, and normalization on the raw dataset W to obtain a filtered dataset WG. Denoising removes noise from the raw dataset W using a median filter. Sliding averaging removes local spikes using a sliding window smoothing method. Normalization converts the different-dimensional data in the raw dataset W to a unified standard scale using a normalization method to obtain the filtered dataset WG.

[0110] The filtered dataset WG is obtained by the following formula: Where WGo represents the o-th data item in the filtered dataset WG, Wo represents the o-th data item in the original dataset W, μWo represents the mean of the o-th data item in the original dataset W, and σ represents the standard deviation of the o-th data item in the original dataset W.

[0111] By constructing an original data set W covering eight contamination characteristics—liquid flow rate Vy, liquid color Sy, viscosity Ny, conductivity Cy, foam interference coefficient FIO, particle disturbance density Rk, emulsification separation velocity Ve, and interfacial tension disturbance Dg—a total of eight contamination characteristics, a multi-dimensional, three-dimensional perception of the contamination status of the cutting fluid is achieved. This collection system fully considers the synergistic effects of contaminants in terms of physical properties (such as particle deposition), chemical properties (such as emulsification dissolution), and interfacial behavior (such as tension changes), avoiding the blind spots caused by existing systems that rely solely on pressure or single-point turbidity for judgment, significantly improving the comprehensiveness and accuracy of contamination identification.

[0112] Without relying on expensive new hardware, the system integrates embedded, low-cost sensing solutions such as image recognition modules, micro-electro-thermal impedance viscometers, fluid perturbation transducers, and high-frequency light scatterers. Combined with image processing algorithms, it acquires indirect contamination parameters such as the emulsification separation velocity Ve and interfacial tension disturbance Dg, enabling precise parameter modeling. This approach not only ensures scalability and low deployment costs, but also adapts to a variety of cutting fluid formulations and complex machining environments, effectively addressing the poor adaptability of traditional systems.

[0113] The median filtering method is used to achieve anti-interference and denoising of the original data, and the sliding average processing is used to reduce the misleading effect of spike noise on system judgment. Combined with the standardization processing method, data of different physical dimensions are unified to the same scale, so that the subsequent pollution factor extraction and trend modeling are consistent in data expression and physical meaning.

[0114] By building a multi-dimensional pollution parameter collection and preprocessing mechanism, the comprehensiveness and refinement of pollution perception is achieved, effectively covering all types of key pollution evolution characteristics and avoiding data omissions. The system significantly improves the accuracy of pollution identification and the stability of characteristic parameters through multi-parameter fusion processing and noise suppression methods, thereby providing a high-quality, computable data foundation for subsequent pollution modeling. In addition, the system relies on the existing sensor configuration for intelligent upgrades and can be deployed without adding hardware, greatly reducing the cost of system transformation and having good engineering practicality. Relying on this high-quality perception data, the system's adaptive filtering mechanism can respond to changes in pollution trends more quickly and accurately in subsequent modules, achieving dynamic control optimization. The pollution state factor extraction module includes a multi-parameter coupling feature construction unit and a pollution feature set acquisition unit.

[0115] The multi-parameter coupling feature construction unit performs nonlinear combination and differential operations on the data in the filtered dataset WG to construct a dynamic coupling relationship and extract pollution sensitivity indicators, including the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3. The particle disturbance index FR1 is obtained using the following formula:

[0116]

[0117] Where Rk(t) represents the particle disturbance density at time t, Dg(t) represents the interfacial tension disturbance at time t, c represents the perturbation stability constant, β and π represent the coupling term adjustment coefficients, FIO(t) represents the foam interference coefficient at time t, and d represents the differential sign.

[0118] The emulsion stability degradation function FR2 is obtained by the following formula:

[0119]

[0120] Wherein, Ve(t) represents the emulsification separation rate at time t, log represents the logarithmic function, Ny(t) represents the viscosity at time t, and Cy(t) represents the conductivity at time t.

[0121] The tension fluctuation intensity FR3 is obtained by the following formula:

[0122]

[0123] Where T represents the length of the time window, Dg(t) represents the interfacial tension disturbance at time t, and μDg represents the average tension disturbance value of the interfacial tension disturbance.

[0124] The pollution feature set acquisition unit fits the acquired particle disturbance index FR1, emulsion stability degradation function FR2 and tension fluctuation intensity FR3 to acquire the pollution feature set FRw.

[0125] This embodiment establishes a "multi-parameter coupled feature construction unit" that nonlinearly combines, couples, and differentially processes key pollution parameters in the raw collected data, achieving a dynamic coupled extraction mechanism for pollution features. Compared to traditional systems that rely on single-variable threshold judgments, this system is more sensitive to pollution and can capture subtle changes in pollution at an early stage of evolution, significantly improving the ability to identify increasing pollution trends.

[0126] The system abstractly constructs three pollution-sensitive factors: the particle perturbation index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3. It then establishes a pollution model based on the three dimensions of particle microflow interference, emulsion interface stability, and molecular tension perturbation amplitude. This structured decomposition approach effectively addresses the existing system's reliance on fuzzy empirical indicators for pollution identification, improving the model's logical interpretability and providing independent observation channels for each key path in the pollution evolution process, facilitating subsequent dynamic modeling and control decisions.

[0127] In constructing the tension fluctuation intensity FR3, the system introduces a window integration approach to statistically analyze the fluctuation amplitude of tension disturbances within a time period. This enhances the model's average suppression of short-term, severe disturbances and its ability to stably represent medium- and long-term pollution trends. This design, unlike single-point burst judgment mechanisms, more comprehensively characterizes the continuity and fluctuation characteristics of pollution development, providing high-quality input for the construction of the pollution trend function Ψ(t).

[0128] By building a highly sensitive and structured pollution feature recognition system, the system's response speed to early changes in pollution evolution has been significantly improved, allowing pollution trends to be perceived and processed more promptly. Through dynamic integration of indicators and optimization of filtering algorithms, the stability of each feature's expression of pollution time series trends has been enhanced, laying a reliable data foundation for subsequent pollution trend modeling and periodic control strategies. Through the model-driven discrimination mechanism, this system effectively reduces the risk of misjudgment and missed judgment caused by rough threshold settings, and improves control accuracy and the scientific nature of filtering decisions. In addition, the construction process of pollution features has clear logic and clear sources, which significantly enhances the interpretability and traceability of pollution judgment results, and provides data support and technical basis for anomaly analysis and fault tracing during system operation. The pollution trend evolution modeling module includes a feature dynamic integration unit and a pollution trend function calculation unit. The feature dynamic integration unit performs standard normalization on the obtained pollution feature set FRw to obtain the pollution normalization set nFRw. The pollution normalization set nFRw is obtained using the following formula:

[0129]

[0130] Where nFRwi represents the i-th feature in the pollution normalization set nFRw, FRwi represents the i-th feature in the pollution feature set FRw, μFRwi represents the mean of the i-th feature in the pollution feature set FRw, and σFRwi represents the standard deviation of the i-th feature in the pollution feature set FRw.

[0131] The pollution trend function calculation unit constructs indicators based on the features in the pollution normalization set nFRw and calculates the pollution trend function Ψ(t). The pollution trend function Ψ(t) is obtained by the following formula:

[0132]

[0133] where α1, α2, and α3 represent the factor calibration coefficients of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3, respectively; xt represents the integral variable; FR1(xt) represents the particle disturbance index at time xt; FR2(xt) represents the emulsion stability degradation function at time xt; and FR3(xt) represents the tension fluctuation intensity at time xt.

[0134] Specific examples:

[0135] Table 1: Pollution trend function acquisition table;

[0136]

[0137]

[0138] The current pollution status is determined by the obtained pollution trend function Ψ(t). When 0 < pollution trend function Ψ(t) < 0.5, the current pollution status is normal. When 0.5 ≤ pollution trend function Ψ(t) < 1.0, the current pollution status is abnormal and filtering is required.

[0139] In this example, each pollution indicator in the pollution feature set FRw is normalized to a standard, unifying the differences in physical dimensions, numerical ranges, and frequency of change among different pollution factors. This results in the construction of the pollution normalization set nFRw. This mechanism effectively addresses the model calculation bias caused by varying indicator scales in traditional systems, significantly improving the mathematical stability and modeling accuracy of pollution trend assessments and laying a unified foundation for the subsequent construction of trend functions.

[0140] The system integrates multiple normalized pollution indicators through time series integration and factor calibration to generate a pollution trend function Ψ(t) that can reflect the development trend of pollution. This pollution trend function is no longer a static threshold judgment, but a continuous indicator that integrates multiple pollution evolution signals and has dynamic fluctuation capabilities, realizing the transformation of evolutionary prediction from "current pollution level" to "pollution development trend". This gives the system predictive capabilities, not just reactive capabilities, avoiding the problems of delayed response or false triggering. By constructing a hierarchical judgment logic based on Ψ(t), the system can automatically determine whether the current cutting fluid needs to enter the filtration cycle based on the real-time output of the pollution trend function. This mechanism realizes the transformation from the traditional "manual setting of the filtration cycle" to "data-driven pollution status triggering", making the filtration control truly on-demand response, hierarchical triggering, and dynamic adaptation, greatly improving the intelligence and autonomy of the system.

[0141] By unifying the expression standards for pollution data features, the consistency and computational accuracy of pollution trend modeling are effectively improved, providing a solid data foundation for pollution state analysis. The system constructs a trend-based pollution state indicator with temporal continuity, enabling dynamic identification and continuous tracking of pollution evolution, avoiding the lag problem of traditional methods that rely solely on burst signals for judgment. By introducing a pollution level judgment mechanism, the system achieves hierarchical identification of pollution states, thereby enhancing the sophistication and automated decision-making capabilities of filtering strategies during execution. Driven by dynamic trends, the system can rationally adjust filtering timing based on the rhythm of pollution changes, effectively avoiding over-frequency or lags in filtering operations, and achieving dual optimization of energy conservation and consumption reduction and system efficiency. Furthermore, this trend indicator, as a key input parameter for periodic regulation, provides a structured and interpretable basis for pollution evolution for subsequent adjustments to filtering cycles, significantly enhancing the engineering applicability and model transparency of the overall control logic.

[0142] The filter cycle adaptive scheduling module maps and transforms the pollution trend index Ψ(t) according to its numerical range and change form, and obtains the pollution normalization index nΨ according to the input scale required by the periodic control function. The pollution normalization index GYΨ is obtained by the following formula:

[0143]

[0144] Where Ψmin represents the valley value of the pollution trend index, and Ψmax represents the peak value of the pollution trend index.

[0145] Based on the obtained pollution normalization index GYΨ, a nonlinear periodic function is constructed to adjust the basic period Tbase to obtain the actual filtration period ΔTfter. The actual filtration period ΔTfter is obtained using the following formula:

[0146]

[0147] Where Tbase represents the base period, and λ represents the period compression adjustment factor.

[0148] The intelligent control execution module includes a filtration cycle control unit and a post-filtration pollution trend feedback unit. The filtration cycle control unit compares the current running time Tnew with the previous filtration time Tlast to determine whether the actual filtration cycle ΔTfter has been reached. The judgment formula is as follows:

[0149]

[0150] Where Pctr(t) represents the pump status instruction and the control output: 1 for start and 0 for hold or stop.

[0151] After the filter pump completes one action, the post-filtration contamination trend feedback unit monitors the new contamination trend data in real time, including the new particle disturbance index nFR1, the new emulsification stability degradation function nFR2, and the new tension fluctuation intensity nFR3, and constructs the post-filtration contamination trend function nΨ. The post-filtration contamination trend function nΨ is obtained by the following formula:

[0152]

[0153] Where Ts represents the short-term feedback window.

[0154] This embodiment solves the problem of unbalanced response of pollution trends in different numerical intervals, so that the pollution trend can be mapped to the standardized input range of the periodic function regardless of whether it is in a mild or severe fluctuation stage, and high-resolution control of period scaling is achieved. This avoids the hysteresis brought about by the traditional "single numerical threshold control" and improves the real-time sensitivity of periodic regulation to the pollution status. In the scheduling strategy, linear proportional scaling or fixed time interval logic is no longer used. Instead, a nonlinear combination adjustment mechanism of the pollution normalization index and the period compression factor is used to achieve intelligent adjustment of the basic period Tbase. This method can not only quickly compress the filtration cycle during the rapid growth stage of the pollution trend, but also automatically slow down the filtration frequency during the slow evolution or stable stage of pollution, effectively balancing the contradiction between filtration efficiency and system resource consumption, significantly extending the life of the filter element and reducing energy usage costs.

[0155] This embodiment utilizes a filtration cycle control unit that compares the current operating time with the last filtration trigger time in real time, ensuring that pump activation is strictly based on the cycle judgment logic, eliminating the uncertainty caused by manual judgment or procedural delays. More importantly, the system also incorporates a post-filtration contamination trend feedback unit that automatically collects new contamination data generated after filtration and reconstructs the contamination trend function nΨ. This mechanism enables the system to self-verify filtration effectiveness and dynamically adjust scheduling parameters, providing a precise basis for the subsequent feedback and learning correction modules.

[0156] This embodiment improves the universality and cross-platform scalability of the control algorithm under different pollution conditions by normalizing the expression of the pollution trend function, ensuring that the control strategy has good adaptability and generalizability. The system introduces a nonlinear periodic control function to achieve a precise mapping relationship between the severity of pollution and the filtering period, thereby effectively avoiding the problem of premature start or delayed response during the filtering process, and ensuring the scientific nature and timeliness of the scheduling decision. In terms of control execution, the system transforms the pump control logic from a traditional fixed timing mechanism to a triggering method based on real-time judgment of the pollution status, significantly improving the rationality of the execution behavior and the response efficiency of the filtering action. The supporting feedback mechanism can collect and update pollution trend data in real time after each filtration is completed, providing a closed-loop verification basis for the system, further enhancing the system's adaptive learning ability and long-term operational stability. Overall, this method makes the filtering behavior more flexible and adjustable. While ensuring the pollution control effect, the system operation significantly improves the energy saving level, response speed and integrity of the control closed loop.

[0157] The feedback and learning correction module includes a pollution deviation determination unit and a parameter adaptive correction unit. The pollution deviation determination unit compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ. The pollution deviation ΔΨ is obtained by the following formula:

[0158] ΔΨ=Ψ(t)-nΨ;

[0159] The obtained contamination deviation ΔΨ is compared with the preset deviation threshold TΨ to determine the filtering effect. When the contamination deviation ΔΨ ≥ the deviation threshold TΨ, the filtering is effective. When the contamination deviation ΔΨ < the deviation threshold TΨ, the filtering is ineffective, and a correction signal is output, initiating the parameter correction process.

[0160] The parameter adaptive correction unit corrects the parameters according to the correction signal, including correcting the factor calibration coefficients α1, α2 and α3 and correcting the period compression adjustment factor λ.

[0161] Correct the factor calibration coefficients α1, α2, and α3 of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3 to obtain new factor calibration coefficients nα1, nα2, and nα3 of the factor calibration coefficients of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3. The correction formula is as follows:

[0162]

[0163] Where, α i represents the calibration coefficient of the i-th factor, and η represents the sensitivity adjustment factor.

[0164] The periodic compression adjustment factor λ is modified to obtain a new periodic compression adjustment factor nλ. The new periodic compression adjustment factor nλ is obtained by the following formula:

[0165] In this embodiment, a contamination deviation determination unit performs a differential analysis between the pre-filter contamination trend function Ψ(t) and the post-filter trend function nΨ to construct a key indicator, ΔΨ, which is used to determine the actual improvement in the filtering operation. This improvement overcomes the drawback of traditional filtering control systems, which only control but do not evaluate, by enabling dynamic closed-loop verification between filtering behavior and results. This fundamentally enhances the system's ability to understand its own operational effectiveness and avoids the waste of resources caused by frequent execution of ineffective filtering.

[0166] The system sets a contamination deviation threshold, TΨ. When the contamination deviation, ΔΨ, falls below the preset threshold, the system no longer assumes the current filtering cycle is effective. Instead, it automatically outputs a correction signal, triggering the optimization process for modeling and control parameters. This design establishes a results-oriented, self-regulating mechanism with clear filtering quality standards. This not only provides a basis for determining whether filtering is being executed, but also provides a quantitative basis for determining whether filtering is effective, enhancing the reliability of system control and the logical consistency of the closed-loop.

[0167] The system uses a parameter adaptive correction unit to dynamically and proportionally correct the calibration coefficients and cycle compression adjustment factor λ used in pollution trend modeling. When filtering effectiveness is insufficient, the system automatically amplifies the relevant modeling factors based on the magnitude of the shortfall in ΔΨ, improving the model's responsiveness to changes in pollution characteristics. Simultaneously, λ is moderately strengthened to ensure a more responsive next round of filtering. This mechanism empowers the system with evolutionary and adaptive control capabilities, enabling it to gradually optimize its control strategies in complex and uncertain machining scenarios.

[0168] This embodiment achieves a quantitative effect evaluation of filtering behavior by introducing a pollution deviation analysis mechanism, enabling the system to have a clear judgment capability on "whether the filtering is effective" and no longer relying on manual experience or static judgment. The system constructs an accurate deviation judgment model based on the comparison of pollution trends before and after filtering, making it clear at a glance whether the filtering effect meets the standards, significantly improving the response accuracy and reliability of the feedback link. At the same time, the system supports a linkage adjustment mechanism between pollution modeling factors and filtration cycle control factors, and can automatically correct model parameters according to actual filtering deviations, thereby enhancing the system's self-learning ability and adaptability under complex working conditions. Through automatic identification of invalid filtering situations and parameter correction, this solution effectively avoids the rigidity problem of traditional control strategies after feedback failure.

[0169] The present invention also provides an intelligent control method for circulating and filtering cutting fluid in machine tools, please refer to Figure 2 , including the following steps:

[0170] Step 1: The data acquisition and time series preprocessing module collects the cutting fluid data through the sensor installed on the machine tool, fits it into the original data set W, and performs preprocessing to obtain the filtered data set WG;

[0171] Step 2: The pollution state factor extraction module extracts features from the obtained filtered data set WG, obtains the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, and fits them into the pollution feature set FRw;

[0172] Step 3: The pollution trend evolution modeling module analyzes the obtained pollution feature set FRw and constructs a pollution trend function Ψ(t) that comprehensively reflects the change of pollution severity over time;

[0173] Step 4: The filtering cycle adaptive scheduling module converts the period of the pollution trend index Ψ(t) into the actual filtering period ΔTfter, replacing the fixed period logic;

[0174] Step 5: The intelligent control execution module triggers the actual pump control action according to the actual filtration cycle ΔTfter, realizes the cycle control closed loop, and obtains the post-filtration pollution trend function nΨ;

[0175] Step 6: The feedback and learning correction module compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ, and compares it with the preset pollution deviation threshold TΨ to determine the filtering effect.

[0176] In this embodiment, the filtration cycle is adaptively adjusted to be driven by the degree of pollution through real-time modeling of the pollution evolution trend. This improvement significantly improves the accuracy of the system's filtration response and the flexibility of regulation, effectively avoids the phenomenon of early filtration (resulting in waste of resources) or delayed filtration (resulting in pollution diffusion), and truly realizes the "on-demand filtration" operation logic. Through steps two and three, this method extracts key pollution factors from multiple physical levels such as particle disturbance, emulsification separation, and interfacial tension, and constructs a high-dimensional, interpretable pollution trend function, which not only accurately reflects the current pollution status, but also predicts the development trend of pollution. Compared with traditional systems that rely only on a single parameter (such as turbidity or pressure difference) to trigger filtration, this method significantly enhances the system's perception and forward-looking judgment capabilities of pollution changes under complex working conditions.

[0177] This method uses a feedback and learning correction module to evaluate the actual contamination improvement effect after each filtration. If the effect does not meet the preset standard, the system automatically adjusts the contamination trend modeling parameters and the filtration cycle compression strategy. This adaptive and self-correcting mechanism effectively solves the problem of rigid control parameters caused by environmental changes and changing cutting conditions in traditional systems. It enables the entire control method to continuously optimize and self-evolve, and is applicable to different machine tool platforms and processing scenarios.

[0178] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent control system for circulating and filtering cutting fluid of machine tools, characterized by: It includes data acquisition and time series preprocessing module, pollution state factor extraction module, pollution trend evolution modeling module, filtration cycle adaptive scheduling module, intelligent control execution module and feedback and learning correction module; The data acquisition and time series preprocessing module collects the cutting fluid data through the sensor installed on the machine tool, fits it into the original data set W, and performs preprocessing to obtain the filtered data set WG; The pollution state factor extraction module extracts features from the obtained filtered data set WG, obtains the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, and fits them into the pollution feature set FRw; The pollution trend evolution modeling module analyzes the acquired pollution feature set FRw and constructs a pollution trend function Ψ(t) that comprehensively reflects the change of pollution severity over time; The filtering cycle adaptive scheduling module converts the period of the pollution trend index Ψ(t) into the actual filtering period ΔTfter, replacing the fixed period logic; The intelligent control execution module triggers the actual pump control action according to the actual filtration cycle ΔTfter, realizes the cycle control closed loop, and obtains the post-filtration pollution trend function nΨ; The feedback and learning correction module compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ, and compares it with the preset pollution deviation threshold TΨ to judge the filtering effect.

2. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 1 is characterized in that: The data acquisition and time series preprocessing module includes a pollution parameter acquisition unit and a data standardization processing unit; The pollution parameter acquisition unit collects the cutting fluid data through sensors, including liquid flow rate Vy, liquid color Sy, viscosity Ny, conductivity Cy, foam interference coefficient FIO, particle disturbance density Rk, emulsification separation speed Ve and interfacial tension disturbance Dg, and fits them into the original data set W; The liquid flow rate Vy is collected and obtained by a turbine flowmeter installed in the filtration loop pipe section; The body color Sy is collected and acquired through an online color sensor installed in the transparent window of the return liquid pipe section; The viscosity Ny is collected by a micro-thermal impedance viscometer installed in a low-flow bypass loop; The conductivity Cy is collected and obtained through the conductivity electrode sensor installed at the outlet of the liquid tank circuit; The foam interference coefficient FIO is acquired through the image recognition module and light reflection sensor installed in the irradiation area above the liquid surface; The particle disturbance density Rk is acquired by collecting the fluid pressure micro-disturbance transducer; The emulsification separation speed Ve is acquired through image processing algorithm; The interfacial tension perturbation Dg is acquired by microbubble image analysis and high-frequency light scattering; The data standardization processing unit performs denoising, sliding average and standardization on the original data set W to obtain the filtered data set WG; Denoising is done by removing the noise from the original dataset W using the median filter method; Sliding average removes local spike interference by using a sliding window for smoothing; By using the standardization method, the data of different dimensions in the original data set W are converted to a unified standard scale to obtain the filtered data set WG; The filtered dataset WG is obtained by the following formula: Where WGo represents the o-th data item in the filtered dataset WG, Wo represents the o-th data item in the original dataset W, μWo represents the mean of the o-th data item in the original dataset W, and σ represents the standard deviation of the o-th data item in the original dataset W.

3. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 2, characterized in that: The pollution state factor extraction module includes a multi-parameter coupling feature construction unit and a pollution feature set acquisition unit; The multi-parameter coupling feature construction unit performs nonlinear combination and differential operations on the data in the filtered data set WG to construct a dynamic coupling relationship and extract pollution sensitivity indicators, including the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3; The particle disturbance index FR1 is obtained by the following formula: Where Rk(t) represents the particle perturbation density at time t, Dg(t) represents the interfacial tension perturbation at time t, c represents the perturbation stability constant, β and π represent the coupling term adjustment coefficients, FIO(t) represents the foam interference coefficient at time t, d represents the differential sign, and dt represents the derivative with respect to time; The emulsion stability degradation function FR2 is obtained by the following formula: Where Ve(t) represents the emulsification separation rate at time t, log represents the logarithmic function, Ny(t) represents the viscosity at time t, and Cy(t) represents the conductivity at time t; The tension fluctuation intensity FR3 is obtained by the following formula: Where T represents the length of the time window, Dg(t) represents the interfacial tension disturbance at time t, and μDg represents the average tension disturbance value of the interfacial tension disturbance; The pollution feature set acquisition unit fits the acquired particle disturbance index FR1, emulsion stability degradation function FR2 and tension fluctuation intensity FR3 to acquire the pollution feature set FRw.

4. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 3 is characterized in that: The pollution trend evolution modeling module includes a characteristic dynamic integration unit and a pollution trend function calculation unit; The feature dynamic normalization unit performs standard normalization processing on the obtained pollution feature set FRw to obtain the pollution normalization set nFRw; The pollution normalization set nFRw is obtained by the following formula: Where nFRwi represents the i-th feature in the pollution normalization set nFRw, FRwi represents the i-th feature in the pollution feature set FRw, μFRwi represents the mean of the i-th feature in the pollution feature set FRw, and σFRwi represents the standard deviation of the i-th feature in the pollution feature set FRw.

5. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 4 is characterized in that: The pollution trend function calculation unit constructs indicators for the features in the pollution normalization set nFRw and calculates the pollution trend function Ψ(t); The pollution trend function Ψ(t) is obtained by the following formula: Where α1, α2 and α3 represent the factor calibration coefficients of the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, respectively; xt represents the integral variable; FR1(xt) represents the particle disturbance index at time xt; FR2(xt) represents the emulsion stability degradation function at time xt; and FR3(xt) represents the tension fluctuation intensity at time xt. The current pollution status is judged by the obtained pollution trend function Ψ(t); When 0<pollution trend function Ψ(t)<0.5, it means the current pollution state is normal; When 0.5≤pollution trend function Ψ(t)<1.0, it means that the current pollution state is abnormal and needs to be filtered.

6. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 5, characterized in that: The filter cycle adaptive scheduling module maps and transforms the pollution trend index Ψ(t) according to its numerical range and change form, and obtains the pollution normalization index nΨ according to the input scale required by the periodic control function; The pollution normalization index GYΨ is obtained by the following formula: Where Ψmin represents the valley value of the pollution trend index, and Ψmax represents the peak value of the pollution trend index; According to the obtained pollution normalization index GYΨ, a nonlinear periodic function is constructed to adjust the basic period Tbase to obtain the actual filtration period ΔTfter; The actual filtration period ΔTfter is obtained by the following formula: Where Tbase represents the base period, λ represents the period compression adjustment factor, and e represents a constant.

7. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 6, characterized in that: The intelligent control execution module includes a filtration cycle control unit and a post-filtration pollution trend feedback unit; The filtration cycle control unit compares the current running time Tnew with the last filtration time Tlast to determine whether the actual filtration cycle ΔTfter has been reached; The judgment formula is as follows: Where Pctr(t) represents the pump status instruction, control output: 1 for start, 0 for hold or stop After the filter pump completes one action, the post-filtration contamination trend feedback unit monitors the new contamination trend data in real time, including the new particle disturbance index nFR1, the new emulsification stability degradation function nFR2, and the new tension fluctuation intensity nFR3, and constructs the post-filtration contamination trend function nΨ; The pollution trend function nΨ after filtering is obtained by the following formula: Where Ts represents the short-term feedback window.

8. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 7, characterized in that: The feedback and learning correction module includes a pollution deviation judgment unit and a parameter adaptive correction unit; The pollution deviation determination unit compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ; The contamination deviation ΔΨ is obtained by the following formula: ΔΨ=Ψ(t)-nΨ; Compare the obtained pollution deviation ΔΨ with the preset pollution deviation threshold TΨ to judge the filtering effect; When the contamination deviation ΔΨ ≥ the contamination deviation threshold TΨ, it means that the filtering is effective; When the contamination deviation ΔΨ is less than the contamination deviation threshold TΨ, it means that the filtering is invalid, a correction signal is output, and the parameter correction process is started.

9. The intelligent control system for circulating and filtering cutting fluid of machine tools according to claim 8, characterized in that: The parameter adaptive correction unit corrects the parameters according to the correction signal, including correcting the factor calibration coefficients α1, α2 and α3 and correcting the periodic compression adjustment factor λ; Correcting the factor calibration coefficients α1, α2, and α3 of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3 to obtain new factor calibration coefficients nα1, nα2, and nα3 of the factor calibration coefficients of the particle disturbance index FR1, the emulsion stability degradation function FR2, and the tension fluctuation intensity FR3; The correction formula is as follows: Where, α i represents the calibration coefficient of the i-th factor, and η represents the sensitivity adjustment factor; Correct the periodic compression adjustment factor λ to obtain a new periodic compression adjustment factor nλ; The new period compression adjustment factor nλ is obtained by the following formula:

10. An intelligent control method for circulating and filtering cutting fluid of a machine tool, applied to the intelligent control system for circulating and filtering cutting fluid of a machine tool according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The data acquisition and time series preprocessing module collects the cutting fluid data through the sensor installed on the machine tool, fits it into the original data set W, and performs preprocessing to obtain the filtered data set WG; Step 2: The pollution state factor extraction module extracts features from the obtained filtered data set WG, obtains the particle disturbance index FR1, the emulsion stability degradation function FR2 and the tension fluctuation intensity FR3, and fits them into the pollution feature set FRw; Step 3: The pollution trend evolution modeling module analyzes the obtained pollution feature set FRw and constructs a pollution trend function Ψ(t) that comprehensively reflects the change of pollution severity over time; Step 4: The filtering cycle adaptive scheduling module converts the period of the pollution trend index Ψ(t) into the actual filtering period ΔTfter, replacing the fixed period logic; Step 5: The intelligent control execution module triggers the actual pump control action according to the actual filtration cycle ΔTfter, realizes the cycle control closed loop, and obtains the post-filtration pollution trend function nΨ; Step 6: The feedback and learning correction module compares the obtained pollution trend function Ψ(t) with the filtered pollution trend function nΨ to obtain the pollution deviation ΔΨ, and compares it with the preset pollution deviation threshold TΨ to determine the filtering effect.

Images