Multi-stage purification intelligent treatment method and system for cutting liquid of machine tool

Abstract

The invention discloses a machine tool cutting fluid multistage purification intelligent treatment method and system, and relates to the technical field of cutting fluid multistage purification. Five types of pollution parameters including optical ambiguity Bg, viscoelastic loss factor Bn, redox residual energy Bh, peculiar smell release frequency Bw, emulsified oil cluster density Br and the like are obtained by deploying multiple types of sensors; the system can master the pollution state of the cutting fluid in real time, and the problems that the precision is low and the misjudgment rate is high due to the fact that the pollution degree is judged only through single-point monitoring are solved. The main pollution change rate F1, the main change acceleration F2 and the chronic accumulation state indication F3 are integrated into a unified index, so that the system not only can identify the current pollution degree, but also can perceive the pollution change trend in advance, and the prospective adjustment capability of the system is remarkably improved. The system does not adopt fixed parameters to control the purification frequency any more, but automatically judges the current purification level by dynamically calculating the pollution threshold value and comparing the evolution index in real time, so that the resource utilization rate is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-stage purification of cutting fluid, and in particular to an intelligent processing method and system for multi-stage purification of cutting fluid of machine tools. Background Art

[0002] The cutting fluid is pumped through a circulation pump to the contact area between the tool and the workpiece for cooling, lubrication, chip flushing, and cleaning. However, as the machining process continues, suspended impurities, metal powder, microemulsified oil clusters, chemical reaction byproducts, and odorous gases accumulate in the cutting fluid, causing contamination to develop. This phenomenon not only affects machining accuracy but also equipment life and personnel health. Therefore, a multi-stage purification system must be used to periodically regenerate the cutting fluid to extend its service life and ensure stable system operation.

[0003] In existing multi-stage cleaning systems for chip fluids, the purification process is mostly controlled by fixed time intervals or manual judgment to control the start and stop frequency and treatment intensity, lacking the ability to perceive and respond to the actual contamination status in real time. This approach has obvious drawbacks: when the contamination has not yet accumulated to a significant level, the system prematurely initiates high-intensity purification, resulting in unnecessary energy consumption and material waste. Furthermore, when contamination evolves rapidly but is not identified in time, a delayed purification response may lead to excessive contamination, fluid failure, and even damage to key components.

[0004] The primary root cause of this situation is that existing systems lack the ability to accurately monitor and predict contamination evolution. Without a multi-source sensing mechanism and evolutionary modeling framework related to the cutting fluid contamination process, the system is unable to extract contamination trend information from multi-dimensional parameters such as liquid turbidity, viscosity changes, pH drift, residual oil concentration, or odor release. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a multi-stage purification intelligent treatment method and system for machine tool cutting fluid, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-stage purification intelligent processing system for machine tool cutting fluid, including a multi-source pollution state acquisition and preprocessing module, a pollution feature extraction and enhancement module, a pollution evolution index modeling module, a purification rhythm switching decision module, a multi-stage purification process control module and an operation feedback and learning module;

[0007] The multi-source pollution state acquisition and preprocessing module collects the data of the chip fluid pollution state through sensors, fits it into the original data set W, and performs preprocessing to obtain the purified data set WH;

[0008] The pollution feature extraction and enhancement module uses statistical filtering and principal component analysis methods to purify the data set WH and extract the three-dimensional pollution feature parameter set FW;

[0009] The pollution evolution index modeling module establishes the pollution evolution index Peo based on the obtained three-dimensional pollution characteristic parameter set FW;

[0010] The purification rhythm switching decision module compares the obtained pollution evolution index Peo with the dynamic pollution threshold to determine the state of the chip fluid purification mode Mh;

[0011] The multi-stage purification process control module dynamically adjusts the parameters of each level of purification unit according to the obtained purification mode Mh;

[0012] The operation feedback and learning module collects the state of the liquid after purification, and combines it with the pollution evolution index Peo and the purification mode Mh to adjust the future dynamic pollution threshold.

[0013] Preferably, the multi-source pollution status acquisition and pre-processing module includes a pollution status data acquisition unit and a pollution status data processing unit;

[0014] The pollution state data acquisition unit collects the data of the pollution state of the cutting fluid through sensors, including optical fuzziness Bg, viscoelastic loss factor Bn, redox residual energy Bh, odor release frequency Bw and emulsified oil cluster density Br, and fits them into the original data set W;

[0015] The optical blur Bg is obtained by using a laser transmittance sensor and an optical scatterometer. Specifically, it is obtained by measuring the intensity attenuation rate of the light source after passing through the liquid and calculating the scattering degree.

[0016] The viscoelastic loss factor Bn is acquired through microrheological sensors. Specifically, it is acquired by calculating the fluid loss ratio based on the energy loss and phase delay under micro-oscillation shear.

[0017] The redox residual energy Bh is collected and obtained through the electrochemical sensor ORP probe; specifically, it is obtained by detecting the rate of change of the redox potential and the residual trend of the reaction in the current liquid;

[0018] The odor release frequency Bw is acquired through a micro gas spectrum component detector; specifically, it is acquired by monitoring the gas volatilization rate and the frequency of composition changes, and extracting the frequency of odor volatilization events;

[0019] The density of emulsified oil clusters Br is obtained by fluorescence particle analyzer; specifically, the density per unit area of ​​oily clusters is obtained by fluorescence excitation identification;

[0020] The pollution state data processing unit removes outliers and performs standardization on the original data set W to obtain a purified data set WH;

[0021] Outlier removal was performed by using the triple median difference method to remove outliers in the original data set W;

[0022] Normalization processing is performed on the original data set W by using Z-score normalization to obtain the purified data set WH;

[0023] The cleaned dataset WH is obtained by the following formula:

[0024]

[0025] Where WHz represents the z-th data in the purified dataset WH, Wz represents the z-th data in the original dataset W, μWz represents the mean of the z-th data in the original dataset W, and σWz represents the standard deviation of the z-th data in the original dataset W.

[0026] Preferably, the pollution feature extraction and enhancement module includes a statistical filtering and denoising unit and a principal component analysis and feature generation unit;

[0027] The statistical filtering and denoising unit performs multi-order time smoothing on the purified data set WH to eliminate discrete jitter, measurement noise, and local abnormal fluctuations, ensuring the stability and physical continuity of the extracted features, and obtains the filtered data set LWH;

[0028] The filtered data set LWH is obtained by the following formula:

[0029] LWHz(t)=λ*WHz(t)+(1-λ)*LWHz(t-1);

[0030] Where LWHz(t) represents the zth data item in the filtered dataset LWH at time t, λ represents the filter smoothing factor, WHz(t) represents the zth data item in the purified dataset WH at time t, and LWHz(t-1) represents the zth data item in the filtered dataset LWH at time t-1.

[0031] Preferably, the principal component analysis and feature generation unit analyzes the filtered data set LWH using a principal component analysis method to construct a three-pollution feature parameter set FW, including the main pollution change rate F1, the main change acceleration F2, and the chronic accumulation state indicator F3;

[0032] The main pollution change rate F1 is obtained by the following formula:

[0033]

[0034] Where LWHi(t) represents the value of the i-th indicator in the filtered data set LWH at time t, αi represents the dynamic contribution factor of the i-th indicator to the overall pollution response, and d represents the derivative symbol;

[0035] The main change acceleration F2 is obtained by the following formula:

[0036]

[0037] Where, d 2 represents the symbol of the second-order derivative, and βi represents the amplification factor used to adjust the acceleration trend of each pollution source;

[0038] The chronic accumulation state indicator F3 is obtained by the following formula:

[0039]

[0040] Where γi represents the weight coefficient of the i-th indicator, log represents the logarithmic function, T represents the integration time window width, LWHi(tx) represents the value of the i-th indicator in the filtered data set LWH at time tx, and dtx represents the integral at time tx.

[0041] Preferably, the pollution evolution index modeling module includes a feature dynamic transformation unit and a pollution evolution index fusion calculation unit;

[0042] The feature dynamic transformation unit normalizes the acquired three-dimensional pollution feature parameter set FW to obtain a normalized parameter set GFW;

[0043] The normalized parameter set GFW is obtained by the following formula:

[0044]

[0045] Where GFWb represents the bth item of data in the normalized parameter set GFW, FWb represents the bth item of data in the three-dimensional pollution characteristic parameter set FW, minFWb represents the valley value of the bth item of data in the three-dimensional pollution characteristic parameter set FW, and maxFWb represents the peak value of the bth item of data in the three-dimensional pollution characteristic parameter set FW;

[0046] The pollution evolution index fusion calculation unit combines the three features in the normalized parameter set GFW into a pollution evolution index Peo, which is used to quantify the dynamic trend of the current pollution level evolution;

[0047] The pollution evolution index Peo is obtained by the following formula:

[0048]

[0049] Where F1(t) represents the main pollution change rate at time t, F2(t) represents the main change acceleration at time t, F3 represents the chronic accumulation state indicator at time t, and ln represents the logarithmic function with base e.

[0050] Preferably, the purification rhythm switching decision module includes a dynamic threshold generation unit and a multi-level rhythm state recognition unit;

[0051] The dynamic threshold generation unit constructs a dynamic threshold model based on the pollution evolution index Peo, and obtains dynamic pollution thresholds through the dynamic threshold model, including the upper switching threshold Thigh, the lower switching threshold Tlow, and the overload risk threshold Tcsu;

[0052] The upper switching threshold Thigh is obtained by the following formula:

[0053] Thigh=μPeo(t-Δ:t)+ω1*σPeo(t-Δ:t);

[0054] Where μPeo(t-Δ:t) represents the mean value of the pollution index in the past time window, σPeo(t-Δ:t) represents the standard deviation of the mean value of the pollution index in the past time window, and ω1 represents the adjustment factor.

[0055] The lower switching threshold Tlow is obtained by the following formula:

[0056] Tlow=μPeo(t-Δ:t)-ω2*σPeo(t-Δ:t);

[0057] Where ω2 represents the adjustment coefficient;

[0058] The overload risk threshold Tcsu is obtained using the following formula:

[0059] Tcsu=Thigh+CSD;

[0060] Where CSD is the overload protection constant, which is used to set the high-risk warning threshold and trigger the emergency mode.

[0061] Preferably, the multi-level rhythm state recognition unit compares the acquired dynamic pollution threshold with the pollution evolution index Peo in real time, determines the state of the current cutting fluid purification mode Mh, and outputs control suggestions;

[0062] The state of the chip fluid purification mode Mh is obtained by the following formula:

[0063] When the pollution evolution index Peo is less than the lower switching threshold Tlow, it indicates the basic maintenance mode, with small pollution fluctuations and a stable trend; it operates at the lowest frequency to maintain the filtration cycle;

[0064] When the lower switching threshold Tlow ≤ pollution evolution index Peo < upper switching threshold Thigh, it indicates medium-speed purification mode, pollution fluctuations are controllable, and an increasing trend appears; the intermediate purification layer is activated and activity monitoring is maintained;

[0065] When the upper switching threshold Thigh ≤ pollution evolution index Peo < overload risk threshold Tcsu, it indicates high-speed purification mode, pollution is aggravated, and there is concentrated load; all levels of purification equipment are activated to increase the flow rate and filtering frequency processing intensity;

[0066] When the pollution evolution index Peo ≥ overload risk threshold Tcsu, it indicates overload emergency mode, severe pollution, and exceeds the tolerance range; the bypass drainage, alarm mechanism and emergency treatment channel are triggered.

[0067] Preferably, the multi-stage purification process control module adjusts the purification unit parameters, including the fluid pump speed Qp, the filter activation level Nf and the ultraviolet disinfection energy PUV, according to the obtained purification mode Mh, including the basic maintenance mode, the medium speed purification mode, the high speed purification mode and the overload emergency mode;

[0068] The fluid pump speed Qp is obtained by the following formula:

[0069] Qp=Qbase+η1*SM(Mh);

[0070] Where Qbase represents the basic maintenance speed, the minimum flow rate; η1 represents the adjustment coefficient of the mode level on the pump speed; SM(Mh) represents the purification mode level, and the numerical mapping is 1, 2, 3, and 4; the basic maintenance mode corresponds to 1, the medium speed purification mode corresponds to 2, the high speed purification mode corresponds to 3, and the overload emergency mode corresponds to 4;

[0071] The number of filter enabled levels Nf is obtained by the following formula:

[0072]

[0073] Where η2 represents the mapping ratio (ensuring that the number of layers increases as the mode level increases), Nmax represents the maximum number of physical layers, such as 4 layers, and min represents the minimum value in the formula;

[0074] Ultraviolet disinfection energy PUV is obtained through the following formula:

[0075] PUV=Pbase*(1+η3*log(1+SM(Mh)));

[0076] Where Pbase represents the basic UV power, and η3 represents the amplification adjustment factor.

[0077] Preferably, the operation feedback and learning module includes a purification effect collection and evaluation unit and a dynamic threshold reconstruction unit;

[0078] The purification effect collection and evaluation unit collects the purified data set WH, obtains the new purified data set nWH, and compares it with the purified data set WH to obtain the reduction ratio Rb;

[0079] The reduction ratio Rb is obtained by the ratio of the difference between the z-th data item of the purified data set WH and the z-th data item of the new purified data set nWH to the z-th data item of the purified data set WH;

[0080] And the comprehensive purification response efficiency index Eeff is constructed through the reduction ratio Rb;

[0081] The purification response efficiency index Eeff is obtained by the following formula:

[0082]

[0083] In the formula, Rbz represents the reduction ratio of the z-th data, and kz represents the response weight of the z-th data;

[0084] The dynamic threshold reconstruction unit combines the purification response efficiency index Eeff, the pollution evolution index Peo and the purification mode Mh to obtain the response deviation index Δdev;

[0085] Δdev=Peo*Mt-Ho*Eeff;

[0086] Where, Ho represents the adjustment term;

[0087] The upper switching threshold Thigh and the lower switching threshold Tlow are adjusted by the obtained response deviation indicator Δdev to obtain the high threshold nThigh and the low threshold nTlow;

[0088] The high threshold nThigh and the low threshold nTlow are obtained by the following formula:

[0089] nThigh=Thigh+μ1*Δdev;

[0090] nTlow=Tlow+μ2*Δdev;

[0091] Where μ1 represents the dynamic correction rate of the high threshold, and μ2 represents the dynamic correction rate of the low threshold.

[0092] When the response deviation index Δdev>0, it means that the purification efficiency is insufficient, and the system will appropriately lower the threshold to enter the high purification mode in advance;

[0093] When the response deviation indicator Δdev is less than 0, it indicates that the purification reaction is excessive, and the threshold is increased to avoid resource waste.

[0094] A multi-stage purification intelligent treatment method for machine tool cutting fluid includes the following steps:

[0095] Step 1: The multi-source pollution state acquisition and preprocessing module collects the data of the chip fluid pollution state through sensors, fits it into the original data set W, and performs preprocessing to obtain the purified data set WH;

[0096] Step 2: The pollution feature extraction and enhancement module uses statistical filtering and principal component analysis methods to purify the data set WH and extract the three-dimensional pollution feature parameter set FW;

[0097] Step 3: The pollution evolution index modeling module establishes the pollution evolution index Peo based on the obtained three-dimensional pollution characteristic parameter set FW;

[0098] Step 4: The purification rhythm switching decision module compares the obtained pollution evolution index Peo with the dynamic pollution threshold to determine the state of the chip fluid purification mode Mh;

[0099] Step 5: The multi-stage purification process control module dynamically adjusts the parameters of each level of purification unit according to the obtained purification mode Mh;

[0100] Step 6: Run the feedback and learning module to collect the state of the liquid after purification, and combine it with the pollution evolution index Peo and the purification mode Mh to adjust the future dynamic pollution threshold.

[0101] The present invention provides a multi-stage purification intelligent treatment method and system for machine tool cutting fluid, which has the following beneficial effects:

[0102] (1) When the system is running, by deploying multiple types of sensors to obtain optical ambiguity Bg, viscoelastic loss factor Bn, redox residual energy Bh, odor release frequency Bw, and emulsified oil cluster density Br, the system can fully and real-time grasp the contamination status of the cutting fluid, solving the problem of low accuracy and high misjudgment rate of traditional systems that rely on single-point monitoring to judge the degree of contamination. The main contamination change rate F1, the main change acceleration F2, and the chronic accumulation state indicator F3 are integrated into a unified index, allowing the system to not only identify the current degree of contamination, but also perceive the trend of pollution changes in advance, realizing a functional leap from "state perception" to "dynamic prediction", and significantly improving the system's forward-looking adjustment capabilities.

[0103] The system no longer uses fixed parameters to control the purification frequency. Instead, it dynamically calculates the pollution threshold and compares the evolution index in real time to automatically determine the current purification level, thereby achieving on-demand response and hierarchical regulation, effectively improving resource utilization.

[0104] (2) By uniformly fitting various pollution parameters into the original data set W, a structural foundation is laid for subsequent data processing, modeling, and feature extraction, facilitating stable system operation and enhancing the consistency and repeatability of data processing. Compared with the traditional mean and triple standard deviation methods, the median difference method is more robust in dealing with asymmetric distributions and extreme value interference, effectively removing abnormal noise data, improving the reliability of pollution perception and the stability of subsequent model training. By performing Z-score standardization on pollution parameters, data of different physical dimensions are mapped to the same scale, making subsequent feature extraction and pollution evolution modeling universal and comparable, and solving the modeling obstacle problem caused by inconsistent dimensions of different parameters.

[0105] (3) The system no longer uses fixed thresholds as the basis for triggering purification strategies. Instead, it combines the historical mean, volatility, and system adjustment factors of the pollution evolution index to construct dynamic upper and lower thresholds and overload protection boundaries that evolve over time. This fundamentally solves the problem of traditional systems failing to respond when operating conditions change or pollution occurs suddenly. Based on the comparison results of the evolution index and the dynamic threshold, the system automatically identifies the current purification level state. Each level corresponds to a different equipment operating intensity and treatment strategy, realizing the "rhythmic, hierarchical" adjustment of the purification strategy for the first time.

[0106] The specially designed overload risk threshold Tcsu is not only used to identify situations where the severity of pollution has exceeded the system's carrying capacity, but also can jointly initiate bypass drainage, alarm and emergency handling procedures, building a system-level "pollution extreme state safety control mechanism" and improving the overall system robustness and engineering safety.

[0107] (4) After each round of purification, the system collects and compares the change in pollution parameters before and after purification in real time, quantifies it into a reduction ratio Rb, and further constructs a comprehensive purification response efficiency index. This mechanism ensures that the system can dynamically evaluate whether the current strategy is reasonable based on the "actual purification results" and has the basic ability of "reflection-adjustment". By integrating the actual purification efficiency, current pollution index and purification level, the "response deviation" index is calculated, and the threshold is dynamically corrected to avoid the problem of "unchanging threshold setting" and achieve long-term adaptive evolution of the system to uncertain factors in operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] Figure 1 This is a block diagram and flow chart of a multi-stage purification intelligent processing system for machine tool cutting fluid according to the present invention;

[0109] Figure 2 This is a schematic diagram of the steps of a multi-stage purification intelligent treatment method for machine tool cutting fluid according to the present invention;

[0110] Figure 3Schematic diagram of the acquisition process of the purification mode of the present invention;

[0111] Figure 4 It is a line graph of the pollution evolution index of the present invention. DETAILED DESCRIPTION

[0112] 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.

[0113] Example 1

[0114] The present invention provides a multi-stage purification intelligent processing system for machine tool cutting fluid, please refer to Figures 1 to 4 , including multi-source pollution status acquisition and preprocessing module, pollution feature extraction and enhancement module, pollution evolution index modeling module, purification rhythm switching decision module, multi-stage purification process control module and operation feedback and learning module.

[0115] The multi-source contamination state acquisition and preprocessing module collects data on the contamination state of the cutting fluid through sensors, fits it into a raw data set W, and performs preprocessing to obtain a purified data set WH. The contamination feature extraction and enhancement module uses statistical filtering and principal component analysis to extract the three-dimensional contamination feature parameter set FW from the purified data set WH. The contamination evolution index modeling module establishes the contamination evolution index Peo based on the acquired three-dimensional contamination feature parameter set FW. The purification rhythm switching decision module compares the acquired contamination evolution index Peo with the dynamic contamination threshold to determine the status of the cutting fluid purification mode Mh. The multi-stage purification process control module dynamically adjusts the parameters of each purification unit based on the acquired purification mode Mh. The operation feedback and learning module collects the state of the purified liquid and combines it with the contamination evolution index Peo and purification mode Mh to adjust the dynamic contamination threshold in the future.

[0116] This embodiment eliminates the traditional time-interval-based purification adjustment approach. Instead, it collects pollution status data in real time, enabling dynamic control based on the "actual pollution status," effectively avoiding blind purification or delayed response. The introduction of a pollution feature extraction and evolution index modeling module accurately characterizes the speed, acceleration, and long-term cumulative trends of pollution changes, enabling the system to proactively identify the possibility of pollution escalation, shifting from "post-event purification" to "predictive purification."

[0117] The purification rhythm switching decision module automatically determines and switches between basic maintenance, medium-speed enhancement, and high-speed response modes, ensuring a dynamic match between purification capacity and pollution severity, significantly improving system resource scheduling efficiency. The multi-stage purification process control module enables precise parameter control of multiple physical units, including fluid pumps, filters, microbubbles, UV treatment, and emulsification separation. These components operate in tandem, avoiding the efficiency limitations of a single purification unit in traditional systems.

[0118] The system rapidly switches purification strategies based on changing contamination trends, reducing processing lag time, lowering the risk of contamination exceeding standards, and improving coolant quality stability. By precisely controlling the purification rhythm, it avoids both the destruction of liquid components caused by overtreatment and the degradation of liquid caused by prolonged contamination accumulation, achieving efficient liquid recycling and regeneration. The system proactively identifies worsening contamination trends, preventing cooling and lubrication failures caused by degraded fluid performance and minimizing the likelihood of abnormalities such as tool ablation and surface damage.

[0119] Example 2

[0120] This embodiment is explained in Example 1, please refer to Figure 1 Specifically, the multi-source pollution status acquisition and preprocessing module includes a pollution status data acquisition unit and a pollution status data processing unit.

[0121] The pollution state data acquisition unit collects the data of the pollution state of the cutting fluid through sensors, including optical ambiguity Bg, viscoelastic loss factor Bn, redox residual energy Bh, odor release frequency Bw and emulsified oil cluster density Br, and fits them into the original data set W.

[0122] The optical blur (Bg) is acquired using a laser transmittance sensor and an optical scatterometer. The viscoelastic loss factor (Bn) is acquired using a microrheological sensor. The redox residual energy (Bh) is acquired using an electrochemical ORP probe. The odor release frequency (Bw) is acquired using a micro gas spectrum component detector. The emulsified oil cluster density (Br) is acquired using a fluorescent particle analyzer.

[0123] The contamination status data processing unit performs outlier removal and normalization on the acquired raw dataset W to obtain a cleaned dataset WH. Outlier removal is performed by using the triple median difference method to remove outliers from the raw dataset W. Normalization is performed by using the Z-score method to normalize the raw dataset W to obtain a cleaned dataset WH.

[0124] The cleaned dataset WH is obtained by the following formula:

[0125]

[0126] Where WHz represents the z-th data in the purified dataset WH, Wz represents the z-th data in the original dataset W, μWz represents the mean of the z-th data in the original dataset W, and σWz represents the standard deviation of the z-th data in the original dataset W.

[0127] This embodiment expands the contamination status parameter from the previous "single physical quantity" to five contamination characteristics with distinct physical and chemical properties: optical blur (Bg), viscoelastic loss factor (Bn), redox residual energy (Bh), odor release frequency (Bw), and emulsified oil cluster density (Br). Each parameter is accurately captured by a dedicated sensor, forming a multi-angle, three-dimensional perception model of the cutting fluid contamination status, significantly improving the system's comprehensiveness in identifying actual contamination conditions. The sensors used, such as a laser transmittance sensor, microrheological sensor, ORP electrochemical probe, gas spectrum detector, and fluorescent particle analyzer, are all highly sensitive and high-resolution professional equipment, ensuring that the collected data accurately reflects subtle changes in the contamination evolution process, addressing the large errors and significant hysteresis associated with traditional sensing mechanisms.

[0128] By uniformly fitting various pollution parameters into the original data set W, a structured foundation is laid for subsequent data processing, modeling, and feature extraction, facilitating stable system operation and enhancing the consistency and repeatability of data processing. Compared with the traditional mean + triple standard deviation method, the median difference method is more robust when dealing with asymmetric distributions and extreme value interference, effectively removing abnormal noise data, improving the reliability of pollution perception and the stability of subsequent model training. By performing Z-score standardization on pollution parameters, data of different physical dimensions are mapped to the same scale, making subsequent feature extraction and pollution evolution modeling universal and comparable, and solving the modeling obstacles caused by inconsistent dimensions of different parameters.

[0129] Through multi-dimensional sensing, anomaly rejection, and standardized preprocessing, the system can promptly identify subtle changes in pollution status, avoiding delayed purification responses due to perception delays. Multiple source parameters jointly reflect pollution patterns, enabling the system to shift from single-indicator judgment to multi-factor analysis. This allows for more accurate distinction between normal fluctuations and pollution evolution trends, reducing the probability of misjudgments and missed detections. The purified dataset WH not only has a more stable numerical distribution but also has been stripped of abnormal noise. As input to subsequent models, it effectively improves the accuracy of the evolution index Peo, ensuring a more scientific basis for decision-making when switching purification rhythms.

[0130] Example 3

[0131] This embodiment is explained in Example 2, please refer to Figure 3Specifically, the pollution feature extraction and enhancement module includes a statistical filtering and denoising unit and a principal component analysis and feature generation unit. The statistical filtering and denoising unit performs multi-order time smoothing on the cleaned data set WH to obtain a filtered data set LWH.

[0132] The filtered data set LWH is obtained by the following formula:

[0133] LWHz(t)=λ*WHz(t)+(1-λ)*LWHz(t-1);

[0134] Where LWHz(t) represents the zth data item in the filtered dataset LWH at time t, λ represents the filter smoothing factor, WHz(t) represents the zth data item in the purified dataset WH at time t, and LWHz(t-1) represents the zth data item in the filtered dataset LWH at time t-1.

[0135] The principal component analysis and feature generation unit uses the principal component analysis method to analyze the filtered data set LWH and constructs the three pollution feature parameter set FW, including the main pollution change rate F1, the main change acceleration F2 and the chronic accumulation state indicator F3.

[0136] The main pollution change rate F1 is obtained by the following formula:

[0137]

[0138] Where LWHi(t) represents the value of the i-th indicator in the filtered data set LWH at time t, αi represents the dynamic contribution factor of the i-th indicator to the overall pollution response, and d represents the derivative symbol.

[0139] The main change acceleration F2 is obtained by the following formula:

[0140]

[0141] Where, d 2 Represents the symbol of the second-order derivative, and βi represents the amplification factor used to adjust the acceleration trend of each pollution source.

[0142] The chronic accumulation state indicator F3 is obtained by the following formula:

[0143]

[0144] Where γi represents the weight coefficient of the i-th indicator, log represents the logarithmic function, T represents the integration time window width, LWHi(tx) represents the value of the i-th indicator in the filtered data set LWH at time tx, and dtx represents the integral at time tx.

[0145] This embodiment dynamically smooths time-series pollution data using statistical filtering and denoising before generating pollution signatures. This method suppresses high-frequency noise and short-term abnormal fluctuations in sensor acquisition, making the data more stable and continuous, and laying a solid foundation for subsequent modeling. The filtering process dynamically fuses new and old data using exponential weighting. This not only avoids misjudgments caused by extreme value disturbances, but also maintains a rapid response to sudden changes in pollution status, addressing the excessive hysteresis of traditional sliding mean filtering.

[0146] This embodiment uses the principal component analysis method to perform dimensionality reduction processing on the filtered multi-source pollution data, and extracts the three composite features that are most relevant to the evolution of pollution: pollution change rate, change acceleration and chronic accumulation index, which effectively reduces redundant information interference and improves modeling efficiency and generalization capabilities. The three extracted pollution characteristics not only include the rate of change of the current pollution level, but also include a second-order trend analysis of the pollution aggravation rate, as well as a comprehensive judgment of the long-term cumulative pollution risk, thus achieving a key leap from "state identification" to "trend prediction". The three characteristic values ​​all correspond to specific pollution physical processes, and through the introduction of dynamic contribution factors, acceleration factors and integral windows, the characteristic indicators are made adjustable and environmentally adaptable, supporting flexible configuration for different working conditions, reflecting the engineering practicality and scalability of model construction.

[0147] Through a two-stage process of filtering, denoising, and feature extraction, the system no longer relies on superficial numerical changes to determine pollution status. Instead, it systematically captures the internal evolutionary logic of pollution dynamics, enhancing the ability of subsequent decision-making modules to assess actual risks. The system can not only identify when pollution levels are already high, but also respond to drastic changes in pollution using acceleration indicators and accumulation items, enabling early intervention to prevent sudden pollution violations.

[0148] Example 4

[0149] This embodiment is explained in Example 3, please refer to Figure 1 and Figure 4 Specifically, the pollution evolution index modeling module includes a feature dynamic transformation unit and a pollution evolution index fusion calculation unit.

[0150] The feature dynamic transformation unit normalizes the acquired three-dimensional pollution feature parameter set FW to obtain a normalized parameter set GFW.

[0151] The normalized parameter set GFW is obtained by the following formula:

[0152]

[0153] Where GFWb represents the b-th data item in the normalized parameter set GFW, FWb represents the b-th data item in the three-dimensional pollution feature parameter set FW, minFWb represents the valley value of the b-th data item in the three-dimensional pollution feature parameter set FW, and maxFWb represents the peak value of the b-th data item in the three-dimensional pollution feature parameter set FW.

[0154] The pollution evolution index fusion calculation unit combines the three features in the normalized parameter set GFW into a pollution evolution index Peo. The pollution evolution index Peo is obtained using the following formula:

[0155]

[0156] Where F1(t) represents the main pollution change rate at time t, F2(t) represents the main change acceleration at time t, F3 represents the chronic accumulation state indicator at time t, and ln represents the logarithmic function with base e.

[0157] Specific examples:

[0158] Table 1: Calculation results of pollution evolution index;

[0159] Time t Main pollution change rate F1 Main change acceleration F2 Chronic accumulation state indication F3 Pollution Evolution Index Peo t1 0.45 0.30 0.25 0.5138 t2 0.60 0.50 0.40 0.6776 t3 0.35 0.20 0.15 0.3925 t4 0.50 0.40 0.35 0.6151 t5 0.55 0.45 0.38 0.6494

[0160] The purification rhythm switching decision module includes a dynamic threshold generation unit and a multi-level rhythm state recognition unit.

[0161] The dynamic threshold generation unit constructs a dynamic threshold model through the pollution evolution index Peo, and obtains the dynamic pollution threshold through the dynamic threshold model, including the upper switching threshold Thigh, the lower switching threshold Tlow and the overload risk threshold Tcsu.

[0162] The upper switching threshold Thigh is obtained by the following formula:

[0163] Thigh=μPeo(t-Δ:t)+ω1*σPeo(t-Δ:t);

[0164] Where μPeo(t-Δ:t) represents the mean of the pollution index in the past time window, σPeo(t-Δ:t) represents the standard deviation of the mean of the pollution index in the past time window, and ω1 represents the adjustment factor.

[0165] The lower switching threshold Tlow is obtained by the following formula:

[0166] Tlow=μPeo(t-Δ:t)-ω2*σPeo(t-Δ:t);

[0167] Where ω2 represents the adjustment coefficient.

[0168] The overload risk threshold Tcsu is obtained using the following formula:

[0169] Tcsu=Thigh+CSD;

[0170] Where, CSD represents the overload protection constant.

[0171] The multi-level rhythm state recognition unit compares the acquired dynamic pollution threshold with the pollution evolution index Peo in real time, determines the state of the current cutting fluid purification mode Mh, and outputs control suggestions.

[0172] The state of the chip fluid purification mode Mh is obtained by matching in the following ways:

[0173] When the pollution evolution index Peo is less than the lower switching threshold Tlow, it indicates the basic maintenance mode, with small pollution fluctuations and a stable trend; it operates at the lowest frequency to maintain the filtration cycle;

[0174] When the lower switching threshold Tlow ≤ pollution evolution index Peo < upper switching threshold Thigh, it indicates medium-speed purification mode, pollution fluctuations are controllable, and an increasing trend appears; the intermediate purification layer is activated and activity monitoring is maintained;

[0175] When the upper switching threshold Thigh ≤ pollution evolution index Peo < overload risk threshold Tcsu, it indicates high-speed purification mode, pollution is aggravated, and there is concentrated load; all levels of purification equipment are activated to increase the flow rate and filtering frequency processing intensity;

[0176] When the pollution evolution index Peo ≥ overload risk threshold Tcsu, it indicates overload emergency mode, severe pollution, and exceeds the tolerance range; the bypass drainage, alarm mechanism and emergency treatment channel are triggered.

[0177] This embodiment normalizes the three-dimensional pollution characteristic parameter set FW so that data of different dimensions and different physical meanings can be uniformly compared and integrated. This approach avoids the dominant influence of a certain characteristic value's absolute value on the model calculation results, and improves the fairness and stability of the subsequent evolution index fusion calculation. Through the construction of the pollution evolution index, the three pollution trend factors, namely the main pollution change rate F1, the main change acceleration F2 and the chronic accumulation state indicator F3, are integrated into a comprehensive indicator in a nonlinear manner. This approach breaks through the traditional coarse-grained judgment mechanism based on a single pollution concentration threshold, and significantly enhances the system's comprehensive recognition capability of the entire pollution evolution process.

[0178] This embodiment no longer uses fixed thresholds as the basis for triggering the purification strategy. Instead, it combines the historical mean and volatility of the pollution evolution index with the system adjustment factor to construct dynamic upper and lower thresholds and overload protection boundaries that evolve over time. This fundamentally solves the problem of traditional systems failing to respond to changes in operating conditions or sudden pollution outbreaks. Based on the comparison results of the evolution index and the dynamic threshold, the system automatically identifies the current purification level state. Each level corresponds to a different equipment operating intensity and treatment strategy, realizing the "rhythmic, hierarchical" adjustment of the purification strategy for the first time.

[0179] The specially designed overload risk threshold Tcsu is not only used to identify situations where the severity of pollution has exceeded the system's carrying capacity, but also can jointly initiate bypass drainage, alarm and emergency handling procedures, building a system-level "pollution extreme state safety control mechanism" and improving the overall system robustness and engineering safety.

[0180] The system no longer relies on static indicators to make purification decisions. Instead, it integrates the rate, acceleration, and long-term accumulation characteristics of dynamic changes to accurately model and predict pollution evolution trends, significantly enhancing the time sensitivity of pollution identification. The multi-level rhythm switching design breaks through the limitations of traditional "start / stop" binary control, enabling the system to select different treatment strategies based on the different stages of pollution, improving resource utilization efficiency and reducing energy consumption and equipment wear.

[0181] The dynamic threshold mechanism enables the system to adaptively adjust under different cutting conditions, material types, and pollution change rates, enhancing its cross-scenario applicability and making it suitable for the long-term stable operation of modern highly flexible manufacturing workshops. The setting of the overload emergency mode not only provides a response measure for extreme pollution situations, but also builds a "safety buffer" for the system, reducing the risk of equipment damage and production line interruption, and ensuring manufacturing continuity and safety. The quantitative expression of the dynamic evolution index and rhythm state provides basic and structured input conditions for the performance evaluation and parameter adaptive adjustment of the operation feedback module, forming a data closed-loop support.

[0182] Example 5

[0183] This embodiment is explained in Example 4. Please refer to Figure 1 and Figure 3 Specifically: the multi-stage purification process control module adjusts the purification unit parameters, including the fluid pump speed Qp, the filter activation level Nf and the ultraviolet disinfection energy PUV, according to the obtained purification mode Mh, including the basic maintenance mode, medium-speed purification mode, high-speed purification mode and overload emergency mode.

[0184] The fluid pump speed Qp is obtained by the following formula:

[0185] Qp=Qbase+η1*SM(Mh);

[0186] Where Qbase represents the basic maintenance speed, η1 represents the adjustment coefficient of the mode level on the pump speed, SM(Mh) represents the purification mode level, and the numerical mapping is 1, 2, 3, and 4.

[0187] The number of filter enabled levels Nf is obtained by the following formula:

[0188]

[0189] Where η2 represents the mapping ratio, Nmax represents the maximum number of physical layer levels, and min represents the minimum value in the formula.

[0190] Ultraviolet disinfection energy PUV is obtained through the following formula:

[0191] PUV=Pbase*(1+η3*log(1+SM(Mh)));

[0192] Where Pbase represents the basic UV power, and η3 represents the amplification adjustment factor.

[0193] The operation feedback and learning module includes a purification effect collection and evaluation unit and a dynamic threshold reconstruction unit.

[0194] The purification effect collection and evaluation unit collects the purified data set WH to obtain a new purified data set nWH, and compares it with the purified data set WH to obtain a reduction ratio Rb.

[0195] The reduction ratio Rb is obtained by the ratio of the difference between the z-th data item of the purified data set WH and the z-th data item of the new purified data set nWH to the z-th data item of the purified data set WH.

[0196] The comprehensive purification response efficiency index Eeff is constructed through the reduction ratio Rb.

[0197] The purification response efficiency index Eeff is obtained by the following formula:

[0198]

[0199] In the formula, Rbz represents the reduction ratio of the z-th data, and kz represents the response weight of the z-th data.

[0200] The dynamic threshold reconstruction unit combines the purification response efficiency index Eeff, the pollution evolution index Peo and the purification mode Mh to obtain the response deviation index Δdev.

[0201] Δdev=Peo*Mt-Ho*Eeff;

[0202] Where Ho represents the adjustment term.

[0203] The upper switching threshold Thigh and the lower switching threshold Tlow are adjusted using the obtained response deviation indicator Δdev to obtain a high threshold nThigh and a low threshold nTlow.

[0204] The high threshold nThigh and the low threshold nTlow are obtained by the following formula:

[0205] nThigh=Thigh+μ1*Δdev;

[0206] nTlow=Tlow+μ2*Δdev;

[0207] Where μ1 represents the dynamic correction rate of the high threshold, and μ2 represents the dynamic correction rate of the low threshold.

[0208] This embodiment obtains the current purification mode level, maps its numerical value and directly drives the changes in the parameters of the key purification units, thereby realizing a complete closed-loop logic chain from pollution perception - mode judgment - equipment response. The system no longer "passively starts and stops" according to the preset threshold, but actively adjusts according to the pollution trend. Unlike traditional systems that only adjust a single parameter, this system simultaneously regulates the three core processing units of pump speed, filter activation level and UV power, making the response means under different purification levels more accurate, efficient and resource-matched, solving the previous problems of "over-purification" or "under-treatment".

[0209] After each round of purification, the system collects and compares the change in pollution parameters before and after purification in real time, quantifying it as a "reduction ratio Rb" and further constructing a comprehensive purification response efficiency index. This mechanism ensures that the system can dynamically evaluate the rationality of the current strategy based on "actual purification results," providing the fundamental ability for "reflection and adjustment." By integrating actual purification efficiency, current pollution index, and purification level, a "response deviation" metric is calculated, allowing for dynamic adjustment of thresholds. This avoids the problem of "unchanging threshold settings" and enables the system to adapt to long-term operational uncertainties.

[0210] Threshold reconstruction no longer relies on empirical settings, but instead continuously fine-tunes the high and low switching thresholds through the system's own operating results, response efficiency, and pollution trends, enabling the system to "learn and optimize while running" and gradually reach the optimal control state.

[0211] Purification components at all levels can respond in a coordinated manner as pollution levels change, avoiding the problem of "too late" or "too frequent" purification behavior in traditional systems, improving processing efficiency while reducing energy consumption and equipment load. Through graded response methods such as medium speed and high speed, resources are enabled on demand, reducing unnecessary high-intensity operating frequencies and effectively extending the service life of key components such as pumps, filters, and UV modules. Real-time quantitative feedback on purification effects supports the system to actively correct control strategies, allowing the system to self-optimize according to actual operating performance, reduce manual intervention, and improve the level of intelligence. Dynamic correction of upper and lower threshold boundaries can ensure that the system is always in a reasonable sensitivity range, avoiding response delays due to overly high settings, or frequent switching due to overly low settings.

[0212] Example 6

[0213] A multi-stage purification intelligent treatment method for machine tool cutting fluid, please refer to Figure 2 , specifically: including the following steps:

[0214] Step 1: The multi-source pollution state acquisition and preprocessing module collects the data of the chip fluid pollution state through sensors, fits it into the original data set W, and performs preprocessing to obtain the purified data set WH;

[0215] Step 2: The pollution feature extraction and enhancement module uses statistical filtering and principal component analysis methods to purify the data set WH and extract the three-dimensional pollution feature parameter set FW;

[0216] Step 3: The pollution evolution index modeling module establishes the pollution evolution index Peo based on the obtained three-dimensional pollution characteristic parameter set FW;

[0217] Step 4: The purification rhythm switching decision module compares the obtained pollution evolution index Peo with the dynamic pollution threshold to determine the state of the chip fluid purification mode Mh;

[0218] Step 5: The multi-stage purification process control module dynamically adjusts the parameters of each level of purification unit according to the obtained purification mode Mh;

[0219] Step 6: Run the feedback and learning module to collect the state of the liquid after purification, and combine it with the pollution evolution index Peo and the purification mode Mh to adjust the future dynamic pollution threshold.

[0220] In this embodiment, by deploying multiple sensors to capture five types of contamination parameters, including optical fuzziness Bg, viscoelastic loss factor Bn, redox residual energy Bh, odor release frequency Bw, and emulsified oil cluster density Br, the system can comprehensively and real-timely assess the contamination status of the cutting fluid, resolving the low accuracy and high misjudgment rate of traditional systems that rely solely on single-point monitoring to determine contamination levels. By integrating the primary contamination change rate F1, primary change acceleration F2, and chronic accumulation state indicator F3 into a unified index, the system not only identifies the current contamination level but also proactively perceives pollution trends, achieving a functional leap from "state perception" to "dynamic prediction," significantly enhancing the system's proactive regulatory capabilities.

[0221] The system no longer uses fixed parameters to control purification frequency. Instead, it dynamically calculates pollution thresholds and compares them with evolution indices in real time to automatically determine the current purification level. This enables on-demand response and hierarchical regulation, effectively improving resource utilization. Based on the results of the purification mode switch, the system automatically adjusts key equipment parameters such as fluid pump speed, filter stages, and UV sterilization power, achieving coordinated regulation of multi-dimensional purification methods and achieving the optimal balance between treatment intensity, rhythm, and resource input.

[0222] This approach builds a full-link control system through six functional steps, from pollution detection to pattern determination, from strategy execution to feedback learning. This completely solves the problem of traditional systems being "only moving without thinking, knowing without adjustment." Purification no longer starts blindly or responds with hysteresis. Instead, it is tightly coupled with the pollution status, capable of both responding to sudden and rapid pollution escalations and adopting stable energy-saving strategies during the slow accumulation phase of pollution. Multi-stage purification units are dynamically activated on demand, effectively controlling unnecessary load operation, reducing energy consumption and equipment fatigue, and extending the overall life of the system.

[0223] 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. A multi-stage purification intelligent treatment system for machine tool cutting fluid, characterized in that: It includes multi-source pollution status acquisition and preprocessing module, pollution feature extraction and enhancement module, pollution evolution index modeling module, purification rhythm switching decision module, multi-stage purification process control module and operation feedback and learning module; The multi-source pollution state acquisition and preprocessing module collects the data of the chip fluid pollution state through sensors, fits it into the original data set W, and performs preprocessing to obtain the purified data set WH; The pollution feature extraction and enhancement module uses statistical filtering and principal component analysis methods to purify the data set WH and extract the three-dimensional pollution feature parameter set FW; The pollution evolution index modeling module establishes the pollution evolution index Peo based on the obtained three-dimensional pollution characteristic parameter set FW; The purification rhythm switching decision module compares the obtained pollution evolution index Peo with the dynamic pollution threshold to determine the state of the chip fluid purification mode Mh; The multi-stage purification process control module dynamically adjusts the parameters of each level of purification unit according to the obtained purification mode Mh; The operation feedback and learning module collects the state of the liquid after purification, and combines it with the pollution evolution index Peo and the purification mode Mh to adjust the future dynamic pollution threshold.

2. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 1 is characterized in that: The multi-source pollution status acquisition and pre-processing module includes a pollution status data acquisition unit and a pollution status data processing unit; The pollution state data acquisition unit collects the data of the pollution state of the cutting fluid through sensors, including optical fuzziness Bg, viscoelastic loss factor Bn, redox residual energy Bh, odor release frequency Bw and emulsified oil cluster density Br, and fits them into the original data set W; Among them, the optical ambiguity Bg is acquired by laser transmittance sensor and optical scatterometer; The viscoelastic loss factor Bn is acquired by microrheological sensor; The redox residual energy Bh is collected and obtained through the electrochemical sensor ORP probe; The odor release frequency Bw is acquired through a micro gas spectrum component detector; The density of emulsified oil clusters Br was obtained by fluorescence particle analyzer; The pollution state data processing unit removes outliers and performs standardization on the original data set W to obtain a purified data set WH; Outlier removal was performed by using the triple median difference method to remove outliers in the original data set W; Normalization processing is performed on the original data set W by using Z-score normalization to obtain the purified data set WH; The cleaned dataset WH is obtained by the following formula: Where WHz represents the z-th data in the purified dataset WH, Wz represents the z-th data in the original dataset W, μWz represents the mean of the z-th data in the original dataset W, and σWz represents the standard deviation of the z-th data in the original dataset W.

3. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 1 is characterized in that: The pollution feature extraction and enhancement module includes a statistical filtering and denoising unit and a principal component analysis and feature generation unit; The statistical filtering and denoising unit performs multi-order time smoothing on the purified data set WH to obtain the filtered data set LWH; The filtered data set LWH is obtained by the following formula: LWHz(t)=λ*WHz(t)+(1-λ)*LWHz(t-1); Where LWHz(t) represents the zth data item in the filtered dataset LWH at time t, λ represents the filter smoothing factor, WHz(t) represents the zth data item in the purified dataset WH at time t, and LWHz(t-1) represents the zth data item in the filtered dataset LWH at time t-1.

4. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 3 is characterized in that: The principal component analysis and feature generation unit uses the principal component analysis method to analyze the filtered data set LWH and constructs the three pollution feature parameter sets FW, including the main pollution change rate F1, the main change acceleration F2 and the chronic accumulation state indicator F3; The main pollution change rate F1 is obtained by the following formula: Where LWHi(t) represents the value of the i-th indicator in the filtered data set LWH at time t, αi represents the dynamic contribution factor of the i-th indicator to the overall pollution response, and d represents the derivative symbol; The main change acceleration F2 is obtained by the following formula: Where, d 2 represents the symbol of the second-order derivative, and βi represents the amplification factor used to adjust the acceleration trend of each pollution source; The chronic accumulation state indicator F3 is obtained by the following formula: Where γi represents the weight coefficient of the i-th indicator, log represents the logarithmic function, T represents the integration time window width, LWHi(tx) represents the value of the i-th indicator in the filtered data set LWH at time tx, and dtx represents the integral at time tx.

5. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 4 is characterized in that: The pollution evolution index modeling module includes a feature dynamic transformation unit and a pollution evolution index fusion calculation unit; The feature dynamic transformation unit normalizes the acquired three-dimensional pollution feature parameter set FW to obtain a normalized parameter set GFW; The normalized parameter set GFW is obtained by the following formula: Where GFWb represents the bth item of data in the normalized parameter set GFW, FWb represents the bth item of data in the three-dimensional pollution characteristic parameter set FW, minFWb represents the valley value of the bth item of data in the three-dimensional pollution characteristic parameter set FW, and maxFWb represents the peak value of the bth item of data in the three-dimensional pollution characteristic parameter set FW; The pollution evolution index fusion calculation unit fuses the three features in the normalized parameter set GFW into a pollution evolution index Peo; The pollution evolution index Peo is obtained by the following formula: Where F1(t) represents the main pollution change rate at time t, F2(t) represents the main change acceleration at time t, F3 represents the chronic accumulation state indicator at time t, and ln represents the logarithmic function with base e.

6. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 5 is characterized in that: The purification rhythm switching decision module includes a dynamic threshold generation unit and a multi-level rhythm state recognition unit; The dynamic threshold generation unit constructs a dynamic threshold model based on the pollution evolution index Peo, and obtains dynamic pollution thresholds through the dynamic threshold model, including the upper switching threshold Thigh, the lower switching threshold Tlow, and the overload risk threshold Tcsu; The upper switching threshold Thigh is obtained by the following formula: Thigh=μPeo(t-Δ:t)+ω1*σPeo(t-Δ:t); Where μPeo(t-Δ:t) represents the mean value of the pollution index in the past time window, σPeo(t-Δ:t) represents the standard deviation of the mean value of the pollution index in the past time window, and ω1 represents the adjustment factor. The lower switching threshold Tlow is obtained by the following formula: Tlow=μPeo(t-Δ:t)-ω2*σPeo(t-Δ:t); Where ω2 represents the adjustment coefficient; The overload risk threshold Tcsu is obtained using the following formula: Tcsu=Thigh+CSD; Where, CSD represents the overload protection constant.

7. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 6 is characterized in that: The multi-level rhythm state recognition unit compares the acquired dynamic pollution threshold with the pollution evolution index Peo in real time to determine the state of the current chip fluid purification mode Mh and output control suggestions; The state of the chip fluid purification mode Mh is obtained by the following formula: When the pollution evolution index Peo is less than the lower switching threshold Tlow, it indicates the basic maintenance mode, with small pollution fluctuations and a stable trend; it operates at the lowest frequency to maintain the filtration cycle; When the lower switching threshold Tlow ≤ pollution evolution index Peo < upper switching threshold Thigh, it indicates medium-speed purification mode, pollution fluctuations are controllable, and an increasing trend appears; Start the intermediate purification layer and maintain activity monitoring; When the upper switching threshold Thigh ≤ pollution evolution index Peo < overload risk threshold Tcsu, it indicates high-speed purification mode, pollution is aggravated, and there is concentrated load; Start all levels of purification equipment to increase the flow rate and filtration frequency; When the pollution evolution index Peo ≥ overload risk threshold Tcsu, it indicates overload emergency mode, severe pollution, and exceeds the tolerance range; Trigger bypass drain, alarm mechanism and emergency treatment channel.

8. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 7 is characterized in that: The multi-stage purification process control module adjusts the purification unit parameters, including the fluid pump speed Qp, the filter activation level Nf and the ultraviolet disinfection energy PUV, according to the obtained purification mode Mh, including the basic maintenance mode, medium-speed purification mode, high-speed purification mode and overload emergency mode; The fluid pump speed Qp is obtained by the following formula: Qp=Qbase+η1*SM(Mh); Where Qbase represents the basic maintenance speed, η1 represents the adjustment coefficient of the mode level on the pump speed, SM(Mh) represents the purification mode level, and the numerical mapping is 1, 2, 3, and 4; The number of filter enabled levels Nf is obtained by the following formula: Where η2 represents the mapping ratio, Nmax represents the maximum number of physical layer levels, and min represents the minimum value in the formula; Ultraviolet disinfection energy PUV is obtained through the following formula: PUV=Pbase*(1+η3*log(1+SM(Mh))); Where Pbase represents the basic UV power, and η3 represents the amplification adjustment factor.

9. The multi-stage purification intelligent treatment system for machine tool cutting fluid according to claim 8, characterized in that: The operation feedback and learning module includes a purification effect collection and evaluation unit and a dynamic threshold reconstruction unit; The purification effect collection and evaluation unit collects the purified data set WH, obtains the new purified data set nWH, and compares it with the purified data set WH to obtain the reduction ratio Rb; The reduction ratio Rb is obtained by the ratio of the difference between the zth item data of the purified data set WH and the zth item data of the new purified data set nWH to the zth item data of the purified data set WH; And the comprehensive purification response efficiency index Eeff is constructed through the reduction ratio Rb; The purification response efficiency index Eeff is obtained by the following formula: In the formula, Rbz represents the reduction ratio of the z-th data, and kz represents the response weight of the z-th data; The dynamic threshold reconstruction unit combines the purification response efficiency index Eeff, the pollution evolution index Peo and the purification mode Mh to obtain the response deviation index Δdev; Δdev=Peo*Mt-Ho*Eeff; Where Ho represents the adjustment term; The upper switching threshold Thigh and the lower switching threshold Tlow are adjusted by the obtained response deviation indicator Δdev to obtain the high threshold nThigh and the low threshold nTlow; The high threshold nThigh and the low threshold nTlow are obtained by the following formula: nThigh=Thigh+μ1*Δdev; nTlow=Tlow+μ2*Δdev; Where μ1 represents the dynamic correction rate of the high threshold, and μ2 represents the dynamic correction rate of the low threshold.

10. A multi-stage purification intelligent treatment method for machine tool cutting fluid, applied to the multi-stage purification intelligent treatment system for machine tool cutting fluid according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The multi-source pollution state acquisition and preprocessing module collects the data of the chip fluid pollution state through sensors, fits it into the original data set W, and performs preprocessing to obtain the purified data set WH; Step 2: The pollution feature extraction and enhancement module uses statistical filtering and principal component analysis methods to purify the data set WH and extract the three-dimensional pollution feature parameter set FW; Step 3: The pollution evolution index modeling module establishes the pollution evolution index Peo based on the obtained three-dimensional pollution characteristic parameter set FW; Step 4: The purification rhythm switching decision module compares the obtained pollution evolution index Peo with the dynamic pollution threshold to determine the state of the chip fluid purification mode Mh; Step 5: The multi-stage purification process control module dynamically adjusts the parameters of each level of purification unit according to the obtained purification mode Mh; Step 6: Run the feedback and learning module to collect the state of the liquid after purification, and combine it with the pollution evolution index Peo and the purification mode Mh to adjust the future dynamic pollution threshold.

Images