A fault automatic early warning method and system for a cutting fluid filling device

Abstract

The present application relates to the technical field of fault early warning, more specifically, the present application relates to a kind of cutting fluid filling equipment's fault automatic early warning method and system, comprising: the acquisition cutting fluid filling loop's multiple-source flow variable data, include pressure, flow and fluid density data;Based on the pressure data and flow data in the multiple-source flow variable data.The present application constructs impedance factor and the discrete significance of statistics by combining the principle of fluid mechanics, quantifies the severity of failure, and establishes a negative correlation mapping between fault significance and algorithm calculation dimension, improves the extended isolation forest algorithm, automatically reduces the feature dimension to reduce the calculation redundancy when facing serious failure, maintains high-precision in weak failure, thereby solving the problem of slow response and wasted computing power of existing fixed dimension algorithm in high-speed filling scene, realizing real-time, low-latency and accurate early warning.

Description

Technical Field

[0001] This invention relates to the field of fault early warning technology. More specifically, this invention relates to an automatic fault early warning method and system for cutting fluid filling equipment. Background Technology

[0002] Cutting fluid filling equipment is widely used in precision machining, primarily for delivering emulsions and other non-Newtonian fluid cooling and lubricating media to the machining area. Because the filling process involves high-frequency pressure shocks, and the fluid often contains trace amounts of metal chips, the equipment circuit is highly susceptible to malfunctions such as nozzle filter clogging or pump wear. Traditional monitoring methods often rely on low-frequency sampling data or single pressure thresholds for judgment, making it difficult to capture transient impedance changes in the fluid under the influence of viscosity and impurities, resulting in insufficient ability to identify early, subtle faults.

[0003] To improve the accuracy of fault detection, existing technologies often employ anomaly detection algorithms based on multidimensional data. Among these, the Extended Isolation Forest (EIF) algorithm has attracted widespread attention due to its ability to effectively handle multidimensional feature spaces. This algorithm identifies outliers by generating random hyperplanes in the feature space to partition data points. However, existing EIF algorithms have limitations when applied to high-speed filling scenarios: their algorithm structure typically uses fixed feature dimensions to generate normal vectors, lacking the ability to perceive the severity of faults. When equipment experiences severe faults with extremely strong characteristic signals (such as pipe bursts or complete blockages), the algorithm still needs to perform complex matrix operations across all dimensions. This computational redundancy not only wastes processor power but also limits the system's response speed to severe faults, making it difficult to meet the needs of industrial sites for real-time, low-latency early warning of streaming data. Summary of the Invention

[0004] This invention provides an automatic fault early warning method and system for cutting fluid filling equipment, aiming to solve the problem that the existing extended isolated forest algorithm in related technologies, when applied to high-speed filling scenarios, usually uses a fixed feature dimension to generate normal vectors and lacks the ability to perceive the significance of faults.

[0005] In a first aspect, the present invention provides an automatic fault early warning method for a cutting fluid filling equipment, comprising: collecting multi-source rheological data of the cutting fluid filling circuit, including pressure, flow rate, and fluid density data; calculating an impedance factor characterizing the transient impedance characteristics of the fluid at the current moment based on the pressure and flow rate data in the multi-source rheological data; calculating a discrete saliency characterizing the deviation of the current data point by combining the impedance factor with the statistical distribution characteristics of the multi-source rheological data under preset normal operating conditions; adjusting the value of a dynamic dimensionality reduction factor based on the discrete saliency within the range of the number of feature dimensions, and using the value of the dynamic dimensionality reduction factor as the number of feature dimensions used for fault detection at the current moment, wherein the dynamic dimensionality reduction factor is negatively correlated with the discrete saliency, and the value range is from the minimum number of dimensions to the number of dimensions of the input data; processing the rheological state data selected according to the number of feature dimensions using an anomaly detection algorithm based on sparse projection of feature saliency to obtain an anomaly score; and generating a fault early warning signal when the anomaly score meets preset early warning conditions. By calculating the impedance factor, which characterizes the fluid's physical properties, and the discrete significance of its statistical characteristics, a dynamic adjustment mechanism was established that shows a negative correlation between the number of feature dimensions and the severity of the fault. Compared with existing technologies, this method can automatically reduce the computational dimension when the fault characteristics are obvious (severe faults), while maintaining full-dimensional detection for weak faults. This significantly reduces the computational complexity of the algorithm while ensuring detection accuracy, and improves the system's real-time response speed to severe faults in high-speed filling scenarios.

[0006] Furthermore, the impedance factor is calculated using the following formula: In the formula, Let be the impedance factor at time i; For sliding windows The pressure value at the i-th time step; For sliding windows The flow rate at time i; and These represent sliding windows. The standardized pressure and standardized flow values ​​at the (i-1)th time interval; For sliding windows The fluid density at the i-th time step; Preset parameters to prevent the denominator from being zero; The impedance factor represents the length of the sliding window. Compared to traditional single pressure threshold monitoring, this impedance factor quantifies the imbalance state of transient flow resistance, enabling more sensitive identification of complex fault characteristics such as high pressure and low flow or low pressure and high flow caused by filter blockage or pump wear. Furthermore, it utilizes fluid density changes to correct and amplify the impedance factor, making the fluctuations more significant in fault scenarios such as chip blockage leading to localized fluid density changes. This improves the algorithm's sensitivity in identifying specific fault modes in fluids containing impurities.

[0007] Furthermore, the calculation of discrete significance includes: calculating the absolute value of the difference between each dimension and its corresponding mean in the multi-source rheological data; standardizing the absolute value of the difference between each dimension, and summing the processed results to obtain the total deviation; multiplying the total deviation by the impedance factor to obtain the discrete significance. This index essentially quantifies the degree to which data points deviate from the normal manifold structure. Compared to simple statistical distance, it utilizes physical impedance characteristics as a gain coefficient, exponentially amplifying the signal when a fault occurs. This allows for clear and accurate quantification of the severity of the fault, providing a reliable basis for subsequent dynamic dimensionality reduction.

[0008] Furthermore, the value of the dynamic dimensionality reduction factor is used as the number of feature dimensions for fault detection at the current moment, wherein the calculation formula for the dynamic dimensionality reduction factor is: In the formula, Let be the dynamic dimensionality reduction factor at time t. Let be the discrete significance at time t; The number of dimensions in the input data; This is the minimum number of dimensions required for the algorithm to run; the minimum number of dimensions is 2. This represents the floor function.

[0009] Furthermore, the anomaly detection algorithm is an extended isolated forest algorithm, and the rheological state data selected based on the number of feature dimensions is processed.

[0010] Furthermore, the processing includes: when constructing isolated tree nodes, generating a sparse random normal vector matching the number of feature dimensions to partition the data space. By generating a sparse random normal vector matching the number of feature dimensions, the weights of unselected dimensions are forced to be reset to zero when constructing isolated tree nodes. This specific processing method reduces the amount of matrix multiplication and dot product operations during hyperplane partitioning from the algorithm's underlying level, causing the computational load to decrease geometrically as the severity of the fault increases, ensuring the system's millisecond-level response capability to sudden severe faults.

[0011] Furthermore, after collecting multi-source rheological data, the method further includes a preprocessing step, which includes filling missing values ​​using regression interpolation, filtering high-frequency noise using a moving average filter, and standardizing the data using the Z-score method. This ensures that the data input into the model has high quality and statistical consistency, avoids false alarms caused by noise interference, and improves the accuracy of subsequent feature calculation and anomaly scoring.

[0012] Furthermore, the length of the sliding window is set to 50 to 100 sampling points.

[0013] In a second aspect, an automatic fault warning system for a cutting fluid filling equipment is also provided, comprising a processor and a memory, the memory storing a computer program, the processor executing the computer program to implement the automatic fault warning method for the cutting fluid filling equipment described in any of the above embodiments.

[0014] Beneficial effects: By combining fluid dynamics principles to construct an impedance factor and statistical discrete significance, the severity of faults is quantified, and a negative correlation mapping between fault significance and algorithm computational dimension is established, thus improving the extended isolated forest algorithm. It automatically reduces feature dimensions to minimize computational redundancy when facing severe faults, while maintaining high-dimensional accuracy for minor faults. This solves the problems of slow response and wasted computing power in existing fixed-dimensional algorithms in high-speed filling scenarios, achieving real-time, low-latency, and accurate early warning. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating a fault detection flowchart according to an embodiment of the present invention;

[0016] Figure 2 This is a schematic diagram illustrating the visualization of detection results according to an embodiment of the present invention;

[0017] Figure 3 This is a schematic diagram showing a comparison of the detection time between the traditional algorithm and the improved algorithm. Detailed Implementation

[0018] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] like Figures 1 to 3 As shown, S101: High-frequency acquisition and cleaning of multi-source rheological data.

[0020] This step aims to acquire physical quantities that reflect the microscopic state of the cutting fluid filling circuit. Because cutting fluids (such as emulsions) have non-Newtonian fluid properties and the filling process is accompanied by high-frequency pressure shocks, conventional low-frequency sampling cannot capture the transient characteristics of the early stages of a fault.

[0021] In this embodiment, a dedicated sensor array is deployed at key locations in the cutting fluid filling circuit: Pressure acquisition: A high-frequency pressure transmitter is installed at the outlet of the volumetric pump to acquire the static pressure of the fluid within the pipeline. Flow rate and fluid density acquisition: A Coriolis mass flow meter is installed at the nozzle tip to simultaneously acquire instantaneous flow rate and fluid density. Compared to turbine flow meters, Coriolis flow meters have better tolerance to cutting fluids containing trace amounts of metal chips and can directly measure fluid density. Drive monitoring: A current transformer is configured at the motor drive end to acquire current data. Environmental monitoring: A temperature sensor is installed at the liquid tank to acquire fluid temperature. The detection equipment used in step S101 is connected to a high-speed data acquisition module via a shielded cable. In this embodiment, the sampling frequency is... Preferred setting is .

[0022] The collected multi-source rheological data were preprocessed as follows: First, the missing values ​​were filled by regression interpolation; then, the high-frequency electromagnetic noise caused by the mechanical vibration of the pump body was filtered out by the moving average filter; finally, the heterogeneous data such as pressure, flow rate, fluid density, current, and temperature were mapped to the same dimension by the Z-score normalization method to form a multi-dimensional condition monitoring time series.

[0023] S102: Calculate the impedance factor at each time step.

[0024] In cutting fluid filling scenarios, the fluid exhibits high viscosity and surface tension. When the filter becomes gradually clogged by chips (increased flow resistance) or the pump wears (increased internal leakage), changes in fluid pressure exhibit a physical lag relative to changes in flow rate. To quantify this impedance imbalance caused by the viscosity and impurity characteristics of the medium, this embodiment introduces physical features. First, the pre-processed pressure sequence... and flow sequence Perform time alignment and truncate the current time. The length before time is The sliding window, denoted as In this embodiment, the window length Preferred A total of 10 sampling points are used to cover at least one complete fluid pulsation cycle.

[0025] Next, calculate The impedance factor at each time step is constructed based on the principles of energy conservation in fluid mechanics and the weighted effect of fluid density. The impedance factor at each time step can be calculated using the following formula: In the formula, Let be the impedance factor at time i; For sliding windows The pressure value at the i-th time step; For sliding windows The flow rate at time i; and These represent sliding windows. The standardized pressure and standardized flow values ​​at the (i-1)th time interval; For sliding windows The fluid density at the i-th time step; To prevent the preset parameter from having a denominator of zero, the value is set to 0.1 in this embodiment. The first part of the formula uses a variant of Bernoulli's principle in fluid mechanics to capture the square law deviation of transient flow resistance; the second part uses the fluid density term for weighting to amplify the impedance fluctuation caused by chip incorporation.

[0026] As can be seen from the above formula, the first part of the formula... This captures the square-law deviation of the fluid's pressure potential energy to kinetic energy conversion within a short time window. Under normal filling conditions, the cutting fluid flow is smooth, and this deviation is small. When chips clog the nozzle, the flow channel cross-section narrows, causing a sudden increase in pressure and a sudden decrease in flow rate. Furthermore, because the chips alter the local fluid density, the fluid density weighted term in the latter part of the formula becomes more significant. Changes occur. The combination of the two leads to... The value increases significantly. Compared to a single pressure threshold alarm, this impedance factor... It has higher sensitivity to combined fault modes of high pressure and low flow or low pressure and high flow.

[0027] S103: Calculate the discrete significance at each time point.

[0028] Based on the impedance factor obtained from the above steps, further analysis revealed that when a fault occurs, the data trajectory diverges towards a low-density region in the multidimensional phase space. To measure the degree of escape of the current data point relative to the normal manifold structure, this embodiment constructs a discrete significance index. Calculation The discrete significance at time point is given by the following formula: In the formula, Let be the discrete significance at time t; Let be the impedance factor at time t; This represents the total number of feature dimensions. For time t, the first Monitoring values ​​in each dimension; and These are the mean and standard deviation of this dimension within a historical normal window, respectively. To prevent the preset parameter from having a denominator of zero, the value in this embodiment is set to 0.1.

[0029] In the above formula The term is essentially a simplified variant for calculating the Mahalanobis distance of data points, reflecting the... The degree to which each dimension deviates from the center. When the cutting fluid equipment is running smoothly, the data points cluster around... Nearby, and the fluid resistance is stable ( (smaller), calculated Approaching When the signs of a malfunction are obvious (such as a burst pipe or severe blockage), not only does the fluid resistance imbalance lead to... The summation term grows rapidly as the summation terms increase and each physical quantity deviates significantly from the mean. As a gain factor, this deviation is further amplified, making It increases exponentially. This indicator can clearly quantify the severity of the fault.

[0030] S104: Calculate the dynamic dimensionality reduction factor at each time step.

[0031] Based on the discrete significance obtained from the above steps, the core hyperparameters of the algorithm need to be determined. In Extended Isolation Forest (EIF), the feature space needs to be segmented to distinguish data points. The uniqueness of the failure of cutting fluid filling equipment lies in the fact that when a serious failure occurs (such as a burst pipe or complete blockage), the signal characteristics are extremely strong. (Very large), at this point in the feature space, the fault point and the normal point are clearly distinguishable. In fact, it is not necessary to call all dimensions (such as temperature, small vibrations, etc.) for complex hyperplane calculations; it is sufficient to judge based on the rough relationship between flow rate and pressure. Only in cases of minor faults or suspected faults ( Only when the fault is relatively small does a full-dimensional fine-grained segmentation become necessary. Therefore, we need to construct a dynamic dimensionality reduction factor to control the number of feature dimensions involved in the algorithm's computation based on the severity of the fault.

[0032] The dynamic dimensionality reduction factor is calculated using the following formula: In the formula, Let be the dynamic dimensionality reduction factor at time t. Let be the discrete significance at time t; The number of dimensions of the input data is the same as the number of dimensions of the multi-source rheological data collected in step S101. This is the minimum number of dimensions required for the algorithm to run; the minimum number of dimensions is 2. This represents the floor function.

[0033] When the cutting fluid equipment shows obvious signs of failure, leading to When the denominator increases dramatically, it rapidly increases, causing the fractional terms to approach 0, and eventually... The value dropped directly to This means that the more obvious the fault characteristics (i.e., the easier they are to identify), the fewer computational dimensions we force the algorithm to use. Conversely, when... When the value is very small (the equipment is close to normal or in the early stages of a minor fault), the denominator is close to 1. Maintain at Nearby, this ensures the algorithm enables high-precision detection across all dimensions. This metric directly maps the physical layer fault intensity to the computational complexity parameter of the algorithm layer.

[0034] S105: Lightweight algorithm improvements and early warning execution.

[0035] Based on the aforementioned dynamic dimensionality reduction factor, this embodiment presents a lightweight improvement to the traditional Extended Isolated Forest (EIF) algorithm.

[0036] In the traditional EIF algorithm, the construction of each tree node requires generating a random slope vector that follows a multidimensional normal distribution. Its dimension is fixed as the full feature dimension. The improved method adopts a feature saliency sparse projection strategy. The specific implementation steps are as follows: Data stream input: The system loads the streaming data of the cutting fluid filling equipment in real time. Dynamic parameter determination: For each sample point to be detected, the real-time dynamic dimensionality reduction factor is first calculated according to steps S103 and S104. Then, when constructing or traversing isolated tree nodes, the algorithm no longer generates full-dimensional normal vectors, but instead randomly selects them. One feature dimension is used to generate the normal vector of the segmenting hyperplane, and the weights of the remaining unselected dimensions are forcibly set to zero. This is because during periods of high failure incidence... Automatic reduction results in a geometric decrease in the amount of computation involved in matrix multiplication and dot product operations.

[0037] The specific early warning process includes: the system loads streaming data from the cutting fluid filling equipment in real time, preprocesses it, and then inputs it into the improved EIF model. The model outputs anomaly scores (0~1). Due to the introduction of a dynamic dimensionality reduction mechanism, the model's response speed when processing significant fault data is improved several times. When the anomaly scores are continuous... When a time step exceeds a preset warning threshold, the system determines that a definite equipment fault has occurred, immediately sends a stop command to the PLC to prevent coolant splashing, and displays the fault type through the human-machine interface, such as abnormal flow resistance - blockage. In this embodiment... The value is 5, and the preset warning line value is 0.85.

[0038] The present invention also provides an automatic fault warning system for cutting fluid filling equipment. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the automatic fault warning method for cutting fluid filling equipment according to the first aspect of the present invention.

[0039] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0040] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0041] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for automatically warning of a failure of a cutting fluid filling apparatus, characterized by, The method comprises: Collecting multi-source rheological data of the cutting fluid filling circuit, including pressure, flow rate and fluid density data; An impedance factor representing the fluid transient impedance characteristic at the current time is calculated based on the pressure data and the flow data in the multi-source flow variation data, and the calculation formula is: , is the impedance factor at the i th time, is the pressure value at the i th time in a sliding window is the flow value at the i th time in a sliding window and respectively represent the normalized pressure value and the normalized flow value at the (i-1) th time in the sliding window is the fluid density at the i th time in a sliding window is a preset parameter for preventing the denominator from being zero, is the length of the sliding window, is the sliding window with the length of before the current time;​​​​ Combining the impedance factor with the statistical distribution characteristics of the multi-source rheological data under a preset normal working condition to calculate a discrete significance representing the deviation degree of a current data point, including: calculating the absolute value of the difference between each dimension of the multi-source rheological data and the corresponding mean value; standardizing the absolute value of each dimension difference and accumulating the processed results to obtain a total deviation, multiplying the total deviation by the impedance factor to obtain the discrete significance; In the value range of the number of feature dimensions, the value of the dynamic dimension reduction factor is adjusted based on the discrete salience, and the value of the dynamic dimension reduction factor is used as the number of feature dimensions for fault detection at the current time; wherein the calculation formula of the dynamic dimension reduction factor is: , is the dynamic dimension reduction factor at time t, is the discrete salience at time t, is the number of dimensions of input data; is the minimum dimension number required for algorithm running, and the value of the minimum dimension number is 2; represents a floor function; the value range is the minimum dimension number to the number of dimensions of input data; Using an anomaly detection algorithm based on feature significance sparse projection to process the rheological state data selected according to the number of feature dimensions to obtain an anomaly score; when the anomaly score meets a preset warning condition, a fault warning signal is generated.

2. The automatic failure early warning method for cutting fluid filling equipment according to claim 1, characterized in that, The anomaly detection algorithm is an extended isolation forest algorithm, and the rheological state data selected according to the number of feature dimensions is processed.

3. The automatic failure early warning method for cutting fluid filling equipment according to claim 2, characterized in that, The processing includes: when constructing an isolated tree node, a sparse random normal vector matching the number of feature dimensions is generated to divide the data space.

4. The automatic failure early warning method for cutting fluid filling equipment according to claim 1, characterized by, After collecting the multi-source rheological data, the method further comprises: A preprocessing step is performed on the multi-source rheological data, including filling missing values using a regression interpolation method, filtering high-frequency noise using a moving average filter, and standardizing data using a Z-score method.

5. The automatic failure early warning method for cutting fluid filling equipment according to claim 1, characterized by, The length of the sliding window is set to 50 to 100 sampling points.

6. A failure automatic early warning system for a cutting fluid filling apparatus, comprising a processor and a memory, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the automatic fault warning method of the cutting fluid filling equipment according to any one of claims 1-5.

Images