Method and system for evaluating monitoring state of processing machine tool based on multi-source sensor

Abstract

The invention discloses a monitoring state evaluation method and system of a machining tool based on a multi-source sensor. The method comprises the steps that a multi-source sensor is adopted to collect operation data of a machine tool; calculating a mixed reaction index HRI of the bed according to the monitoring data; carrying out FFT (Fast Fourier Transform) analysis on the vibration data to generate a change curve of each parameter; according to the calculation result of the mixed reaction index HRI and the change curve of each parameter, a combined diagnosis model is adopted to carry out monitoring and early warning; and performing combined alarm according to a preset alarm threshold value and automatic monitoring early warning, and outputting a maintenance suggestion. The comprehensive, real-time and intelligent functions are realized, potential problems are warned in advance, the downtime is reduced, the operation efficiency and reliability of the machine tool are improved, and powerful technical support is provided for intelligent manufacturing.

Description

Technical Field

[0001] The present invention relates to the technical field of machine tool status assessment, and in particular to a monitoring status assessment method and system for a processing machine tool based on a multi-source sensor. Background Art

[0002] Gear machining CNC machine tools refer to high-precision mechanical equipment specially used for manufacturing and processing various types of gears. This type of machine tool is controlled by a CNC system and can realize automated and precise processing of gears. It can automatically complete the entire processing process from blank to finished product according to pre-programmed instructions. Gear machining CNC machine tools are widely used in many industries such as automobile manufacturing, aerospace, railway transportation, engineering machinery, wind power equipment, etc., and are indispensable key equipment in modern manufacturing. Gear machining CNC machine tools usually include hardware and software parts. The hardware part includes the mechanical structure and control system of the machine tool. The mechanical structure includes the machine tool itself, the workpiece table, the spindle system, the feed system, the guide rail and the lead screw, etc. The software system includes CNC system software, CNC programming software, network communication and other software parts. The hardware and software systems in the mainstream gear machining CNC systems at home and abroad are mostly upper and lower computer structures, in which the upper computer is responsible for the non-real-time tasks of the system, and the lower computer is responsible for the real-time motion control and logic control tasks of the system. The two interact with each other through the internal bus or network. Operators can directly read the real-time data of processed parts and control the CNC machine processing process through the human-machine interface (HMI) of the host computer.

[0003] During the machining process, gear machining CNC machine tools often face the problem of maintenance or tool replacement. However, the machining status display of gear machining CNC machine tools cannot fully reflect the machine tool status. In addition, since machine tools are generally not equipped with large-capacity storage hardware, historical machining curve parameters are usually not stored. This makes it difficult to determine whether the machine tool status is abnormal simply by the machine tool itself. At the same time, whether there are quality problems with the workpiece cannot be discovered and traced back in time during the machine tool machining process.

[0004] During the gear machining process, CNC machine tools often require regular maintenance and tool replacement. However, traditional gear machining CNC machine tools face many challenges in these aspects, especially in real-time monitoring and historical data storage. These problems not only affect the operating efficiency of the machine tools, but also bring difficulties to the quality control of the machined parts.

[0005] The specific problems are as follows: 1. Challenges of maintenance and tool replacement Frequent maintenance needs: Gear machining involves high-precision cutting operations, which places strict requirements on the mechanical components and tools of the machine tool. As the machining time increases, problems such as tool wear, spindle vibration, and guide rail wear gradually emerge, resulting in reduced machining accuracy and even possible failures. Therefore, regular maintenance and tool replacement are the key to ensuring stable operation of machine tools and product quality.

[0006] Lack of comprehensive status display: Although modern CNC machine tools are equipped with some sensors and monitoring systems, the existing machining status display is usually limited to real-time parameters (such as speed, feed rate, temperature, etc.), which cannot fully reflect the overall health of the machine tool. For example, some potential problems (such as early wear and small vibrations) may not be immediately reflected in these parameters, making it difficult to detect and deal with them in time.

[0007] 2. Limitations of Historical Data Storage and Analysis Limited storage capacity: Most gear processing CNC machine tools are not equipped with large-capacity storage hardware, which makes it impossible for them to save detailed processing curves and parameter records for a long time. Even if some high-end models have certain data storage functions, their capacity and storage time are very limited, and usually only the current processing data is displayed.

[0008] Lack of traceability: Due to the lack of historical data, when processing quality problems or equipment failures occur, companies often find it difficult to conduct effective traceback analysis through existing data. This not only increases the difficulty of troubleshooting, but may also lead to production delays and cost increases. In addition, the lack of historical data also limits the optimization and improvement of the processing process, making it difficult to achieve continuous quality improvement. Summary of the invention

[0009] Therefore, the purpose of the present invention is to provide a monitoring status evaluation method and system for processing machine tools based on multi-source sensors, to achieve comprehensive, real-time and intelligent machine tool status monitoring, to provide early warning of potential problems, to reduce downtime, to improve machine tool operation efficiency and reliability, and to provide strong technical support for intelligent manufacturing.

[0010] In order to achieve the above object, the present invention provides a monitoring state evaluation method for a processing machine tool based on a multi-source sensor, comprising the following steps: Using multi-source sensors to collect the operation data of the machine tool; the operation data includes vibration data, temperature data, speed data, motor pressure data, and motor load data of different axes of the CNC machine tool; Based on the monitored data, the mixed reaction index HRI of the tool machine is calculated; ; in, represents the temperature index, Indicates the pressure index; It represents the vibration index; Perform FFT analysis on the vibration data to generate a variation curve of the vibration parameters; perform curve fitting on other data to obtain a time variation curve of other data; Based on the calculation results of the mixed reaction index HRI and the change curves of various data, a joint diagnosis model is used to conduct monitoring and early warning; Combined alarms are issued based on preset alarm thresholds and automatic monitoring warnings, and maintenance recommendations are output.

[0011] Further, in the mixed reaction index HRI, Temperature Index +...; Stress Index +...; Vibration Index +...; in, is the axis number / tool ​​number selected according to the working condition, x is 1, 2, 3...; , represents the temperature of the selected axis and / or tool, Indicates the temperature offset of the selected axis / tool ​​No.1; Indicates the temperature offset of the selected axis / tool ​​No. 2; All represent the pressure correction value of the drive motor of the selected axis; , , , All are numbers of drive motors; They respectively represent the vibration correction values ​​of the selected No. 1 axis / tool, No. 2 axis / tool, and No. 3 axis / tool.

[0012] Further, the vibration data is subjected to FFT analysis to generate a variation curve of vibration parameters; and the method also includes generating a single-piece frequency domain diagram, a multi-piece frequency domain diagram and a multi-piece Campbell diagram according to the FFT analysis results.

[0013] Further, the curve fitting of other data to obtain the time variation curve of other data includes: Collect speed data, temperature data and pressure data at preset time intervals; calculate the root mean square values ​​respectively; use edge computing architecture to generate a time change curve for the data collected at preset time intervals.

[0014] Furthermore, the joint diagnosis model includes: The first-level warning based on real-time data collection and analysis includes: the real-time monitored operation data is divided into specific sensitive data and general data. The specific sensitive data includes key parameters such as vibration, temperature, speed, tool life, motor drive load, motor pressure, etc., which have a direct impact on the operation status of the machine tool. General data includes auxiliary parameters such as workpiece coordinates, current processing part number, current number of processing parts, and start-up device control switch. When the specific sensitive data in the real-time monitored data exceeds the preset threshold, a first-level warning is issued; Secondary warning based on mixed reaction index: During the gear processing, vibration, temperature and motor pressure are important processing parameters that can comprehensively reflect the current gear processing status. Therefore, we extracted and analyzed multi-dimensional features of real-time monitoring data. When the calculated mixed reaction index HRI does not meet the preset range, a secondary warning is issued; and the change curve formed by the current moment and the previous processing part is extracted to compare the faults. Three-level warning of intelligent model based on multi-dimensional features: the real-time monitored operation data is formed into a multi-dimensional vector, and the operation data collected each time is selected as samples according to the sampling frequency, the joint distribution function of the selected sample data is calculated, and the abnormal score of the sample data is output to perform a three-level warning.

[0015] Further, the calculating of the joint distribution function of the selected sample data and outputting the abnormality score of the sample data includes: For each feature in each sample data, calculate the left tail joint distribution value of the current feature and the right-tailed joint distribution value ; Calculate the left-tail anomaly score, right-tail anomaly score, and automatic detection anomaly score for each sample; The left-tail anomaly score, right-tail anomaly score, and automatic detection anomaly score are fused to return the anomaly scores of all samples.

[0016] The present invention also provides a monitoring state evaluation system for a gear CNC machining machine tool based on a multi-source sensor, comprising a client and a server; the client is used for multi-source data acquisition and local calculation, and sends the data to the server; the server performs cloud-based calculation of a hybrid reaction index HRI based on the original data and the local calculation results, and sends the calculation results to the client for early warning; the client comprises a multi-source data acquisition module, an FFT calculation module, a real-time monitoring module, and an abnormal alarm module; the server comprises a hybrid reaction calculation module, a real-time monitoring module, an FFT analysis module, and an abnormal alarm module; The multi-source data acquisition module is used to collect the operation data of the machine tool; the operation data includes vibration data, temperature data, speed data, motor pressure data, and motor load data of different axes of the CNC machining machine tool; The FFT calculation / analysis module performs FFT calculation and analysis on the vibration monitoring data to generate the change curve of each parameter; The real-time monitoring module performs FFT analysis on the vibration data to generate a change curve of the vibration parameters; performs curve fitting on other data to obtain a time change curve of other data. When displaying the monitoring data in the time domain, it can be filtered according to processing date, processing parts, processing workpieces, processing steps, etc. The mixed reaction calculation module calculates the mixed reaction index HRI of the tool according to the monitoring data; ； in, represents the temperature index, Indicates the pressure index; It represents the vibration index; The abnormal alarm module uses a joint diagnosis model to monitor and warn based on the calculation results of the mixed reaction index HRI and the change curves of various parameters; it conducts joint alarms based on preset alarm thresholds and automatic monitoring and warnings, and outputs maintenance suggestions.

[0017] Furthermore, it also includes edge computing and cloud storage modules, which use edge computing technology to perform preliminary processing and compression on the collected data locally and upload important data to cloud storage.

[0018] Furthermore, it also includes an edge computing module, which is used to collect vibration data, rotation speed data, temperature data and pressure data at preset time intervals; perform root mean square value calculations respectively; and use an edge computing architecture to generate a time change curve for the data collected at preset time intervals.

[0019] Furthermore, it also includes a visualization module, which is used to display the change curves of various parameters generated by FFT analysis, real-time monitoring data and fault data.

[0020] The present application discloses a monitoring status assessment method and system for machining machine tools based on multi-source sensors. Compared with the existing technology, it has significant improvements in comprehensiveness, real-time and intelligence, can warn of potential problems in advance, reduce downtime, improve the operating efficiency and reliability of machine tools, and provide strong technical support for intelligent manufacturing. Through multi-source sensor integration and multi-protocol support: the operating data of machine tools can be collected more comprehensively. These sensors can not only monitor key parameters, but also capture subtle changes and warn of potential problems in advance. The system supports multiple communication protocols, which not only improves the compatibility and scalability of the system, but also simplifies integration and maintenance. Using edge computing technology, the collected data is preliminarily processed and compressed locally to reduce the transmission burden. At the same time, important data is uploaded to cloud storage to ensure long-term preservation and access at any time.

[0021] The artificial intelligence-based real-time monitoring system developed by this application can automatically identify abnormal patterns and issue alarms, and send preset instructions to the machine tool to complete automatic control. By comparing historical data and current status, the system can quickly determine whether there are potential risks in the machine tool and provide corresponding maintenance suggestions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic flow chart of a method for monitoring state evaluation of a machining tool based on a multi-source sensor according to the present invention; Figure 2 It is a schematic diagram of the C / S architecture of the monitoring state evaluation system of the gear CNC machining machine tool based on multi-source sensors of the present invention; Figure 3 It is the status display interface of each factory area under real-time monitoring of the present invention; Figure 4 The present invention is a real-time monitoring - status data display interface of each machine tool in a certain factory area; Figure 5 A time domain monitoring data diagram of a machine tool monitored in real time by the present invention; Figure 6 The processing curve of a machine tool monitored in real time by the present invention; Figure 7 It is the statistical result of abnormal situation of the present invention; Figure 8 Record results for the unusual problems of the present invention; Fig. 9 It is a frequency domain diagram of the machine tool spindle when the machine tool is a toothed machine in the embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described in detail below through the accompanying drawings and specific embodiments.

[0024] like Figure 1 As shown, an embodiment of the present invention provides a monitoring state evaluation method for a processing machine tool based on a multi-source sensor, comprising the following steps: S1. Use multi-source sensors to collect the operation data of the machine tool; the operation data includes vibration data, temperature data, speed data, motor pressure data, and motor load data of different axes of the CNC machine tool; First, temperature sensors are used to monitor the temperature changes of various components during machine tool processing, including key parts such as the spindle and tool. These temperature data are crucial to determine whether the equipment is overheating and whether there is a potential failure. High temperature may cause deformation or damage to mechanical parts, affecting processing accuracy, surface quality and equipment life. In order to monitor these temperature changes, temperature sensors are installed inside the motor, and the various temperature values ​​are provided as parameters by the BOSCH Rexroth control system or Siemens control system. The temperature can be monitored individually, and if it exceeds the set value, the corresponding error response will be triggered. For example, if the temperature reaches the limit of 50°C or 55°C, the machine will stop running through the "StopCycle" instruction to prevent further damage or quality problems.

[0025] Secondly, vibration sensors are used to detect vibrations of tools, spindles, centers, and other components. Abnormal vibrations may be an early warning sign of tool wear, bearing failure, or other mechanical problems. By monitoring these vibration data in real time, potential problems can be discovered and dealt with in a timely manner to avoid greater losses.

[0026] In addition, the speed sensor is used to monitor the spindle speed information in real time. Spindle speed is one of the important indicators to measure the performance of machine tools. Ensuring its stability and accuracy is crucial to ensuring machining quality. The speed sensor can provide accurate speed data to help operators and maintenance personnel to effectively monitor and adjust.

[0027] In addition to the above information, operators and quality control personnel need more processing information to monitor the processing process, including: workpiece coordinate information (xyz axis), starter control switch, current processing part number, current processing number of pieces, tool life, number of pieces to be processed, number of tool rotations, slide penetration, slide pressure, spindle drive load, spindle drive torque, etc. This information can be collected through a variety of protocols, such as PLC, OPC UA, and NC (numerical control) program collection.

[0028] The vibration monitoring data is subjected to FFT analysis to generate the variation curves of each vibration parameter; including: performing FFT analysis on the vibration signals of different axes of the CNC machining machine tool; generating a single-piece frequency domain diagram, a multi-piece frequency domain diagram and a multi-piece Campbell diagram based on the FFT analysis results.

[0029] Further, the curve fitting of other data to obtain the time variation curve of other data includes: Collect speed data, temperature data and pressure data at preset time intervals; calculate the root mean square values ​​respectively; use edge computing architecture to generate a time change curve for the data collected at preset time intervals.

[0030] The collected information will be pre-processed to reduce the transmission burden, and then stored in the database and uploaded to the cloud database to facilitate the subsequent real-time display, analysis and early warning on the server side. Through the integration of the above-mentioned multi-source sensors and multi-protocol support, the data acquisition and analysis module can fully and real-timely obtain various operating parameters of the machine tool, and transmit these data to the data analysis and storage module for processing. This not only helps to timely discover and solve abnormal situations in the operation of the machine tool, but also provides important data support for subsequent quality control and process optimization.

[0031] Figure 3 The real-time monitoring module shows the machine tool status monitoring and alarm information for multiple factory areas. The factory status is divided into: normal, warning, out of tolerance and shutdown. For each factory area, the monitoring status of the machine tools in the factory area will scroll in real time below. Figure 4 The monitoring status of machine tools in a single factory is displayed. The key parts of the machine tools are numbered, and different colors are used to mark them when warnings are issued based on the monitoring data of the spindle or tool. For example, red and yellow indicate that the warning value has been exceeded, and green indicates normal. Figure 5 The time domain monitoring data of a single machine tool is shown. The monitoring parameters include: vibration and speed of multiple spindles, current percentage of each motor, vibration information of the top spindle and tool spindle, HRI index status (this index will be described below), tool life, and processing curve, etc. Figure 6 A machine tool processing curve diagram is given. This curve is the processing curve displayed by the machine tool host computer, which shows the tool feed amount, slide pressure, spindle drive load and torque of each processing step during the processing. However, due to the storage capacity of the machine tool itself, the curve usually only displays the curve data of the current workpiece, and does not store historical data, making it impossible for operators to trace back historical processing information. Based on this pain point, the system of the present invention performs historical storage and real-time monitoring of the machine tool processing curve, which is convenient for operators and quality control personnel to monitor the machine tool processing status and trace back historical processing part status information.

[0032] The hybrid reaction index HRI of the machine tool is calculated based on the monitoring data; a new index is created by combining three process parameters using a formula. This index represents the process without units and is used to reflect the overall status of the machine tool. The three key process parameters are: temperature, vibration and pressure. The index is synthesized based on the temperature, current pressure, vibration and other information of each axis of the machine tool processing, and uses one value to reflect the machine tool processing process.

[0033] ； in, represents the temperature index, Indicates the pressure index; It represents the vibration index; Temperature Index +...; Stress Index +...; Vibration Index +...; in, is the axis number / tool ​​number selected according to the working condition, x is 1, 2, 3...; , represents the temperature of the selected axis / tool, Indicates the temperature offset of the selected axis / tool ​​No.1; Indicates the temperature offset of the selected axis / tool ​​No. 2; All represent the pressure correction value of the drive motor of the selected axis; , , , All are numbers of drive motors; They respectively represent the vibration correction values ​​of the selected No. 1 axis / tool, No. 2 axis / tool, and No. 3 axis / tool.

[0034] Regarding the calculation of the above-mentioned temperature, pressure and vibration parameters, the specific formulas for gear machining machines of different brands and with different processes will vary, depending on the axes and / or tools to be monitored by the gear machining machines of different brands and with different processes.

[0035] It should be noted that in this application, the shaft and / or tool to be tested can be selected according to the actual working conditions. For example, in the following embodiment, if the machine tool is a toothing machine, the sensor is determined according to the test requirements of the toothing machine, that is, according to the shaft. When the machine tool is a gear rolling machine, the number of sensors is determined according to both the shaft and the tool. For example, if there are three tools on the same tool shaft, three sensors are added to monitor each tool. That is, the addition of sensors is highly related to the type of machine tool, and the number of sensors is not calculated based solely on the shaft.

[0036] When the machine tool is a gouge machine, the temperature, pressure, and vibration index in the mixed reaction index HRI are calculated using the following formula: Temperature Index ; Stress Index ; Vibration Index ; in, Indicates the temperature of the workpiece spindle, Indicates the temperature offset of the workpiece spindle; Indicates the temperature of the tool spindle, Indicates the temperature offset of the tool spindle; ForC, ForB, ForX, ForZ Respectively represent the pressure correction value of the drive motor of the workpiece spindle, the pressure correction value of the drive motor of the tool spindle, the pressure correction value of the drive motor of the feed axis, and the pressure correction value of the drive motor of the swing axis; , , Respectively represent the vibration correction values ​​of the workpiece spindle, tool spindle and tailstock spindle.

[0037] The vibration value of a single axis is calculated as ; During measurement, the vibration sensor can set the offset Offset and factor Factor to calibrate the sensor value and change the dimension.

[0038] In this embodiment, The axis No. 1 selected by C is the workpiece spindle. The second axis selected by B is the tool spindle.

[0039] Based on the calculation results of the mixed reaction index HRI and the change curves of various parameters, a joint diagnosis model is used to perform monitoring and early warning; based on the preset alarm threshold and automatic monitoring and early warning, a joint alarm is performed and maintenance suggestions are output.

[0040] The combined diagnosis model includes: Level 1 warning for real-time data collection and analysis: The real-time monitored operating data is divided into specific sensitive data and general data. Specific sensitive data include key parameters such as vibration, temperature, rotation speed, tool life, motor drive load, motor pressure, etc., which have a direct impact on the operating status of the machine tool. General data include auxiliary parameters such as workpiece coordinates, current processing part number, current number of processed parts, and starting device control switch. When specific sensitive data in the real-time monitored data exceeds the preset threshold, a level 1 warning is issued; Level 2 warning based on mixed reaction index: During the gear processing process, vibration, temperature, and motor pressure are important processing parameters that can comprehensively reflect the current gear processing status. Therefore, we performed multi-dimensional feature extraction and analysis on the real-time monitoring data. When the calculated mixed reaction index HRI does not meet the preset range, a level 2 warning is issued; and the change curve formed by the current moment and the previous processing part is extracted for fault comparison and display; Three-level warning of intelligent model based on multi-dimensional features: the real-time monitored operation data is formed into a multi-dimensional vector, each monitoring data is selected as a sample in turn, the joint distribution function of the selected sample data is calculated, the abnormal score of the sample data is output, and a three-level warning is issued.

[0041] The above three levels of warnings can be executed in parallel. When any warning condition is met, a joint alarm will be issued and maintenance suggestions will be output.

[0042] Based on the above multi-source sensor acquisition, the characteristics of the multi-dimensional vector are mainly composed of the following key parameters: Vibration parameters: Vibration correction values ​​for the selected monitored axis and / or tool.

[0043] Temperature parameters: Temperature correction values ​​for the selected monitoring axis and / or tool.

[0044] Pressure parameter: The pressure correction value of the selected monitoring drive motor.

[0045] Speed ​​and load: The speed of the selected monitoring shaft, the motor drive torque, and the correction value of the load.

[0046] Tool life: The life value of the selected monitoring tool.

[0047] Timestamp: The sensor acquisition timestamp.

[0048] Further, the calculating of the joint distribution function of the selected sample data and outputting the abnormality score of the sample data includes: For each multidimensional feature in the sample data, calculate the left tail joint distribution value of the current sample and the right-tailed joint distribution value ; Calculate the left-tail anomaly score, right-tail anomaly score, and automatic detection anomaly score for each sample; The left-tail anomaly score, the right-tail anomaly score and the automatic detection anomaly score are fused to return the anomaly scores of all samples; specifically, the left-tail anomaly score, the right-tail anomaly score and the automatic detection anomaly score of each sample are maximized, and then fused in the order of the samples to finally obtain the anomaly scores of all samples.

[0049] ECOD algorithm is an unsupervised anomaly detection method that identifies anomalies in machining process data based on the empirical cumulative distribution function (ECDF). Anomalies usually correspond to rare events that occur in the low-density part of the probability distribution. Therefore, points at the "extreme" of the tail probability can be considered as anomalies. The following is a detailed explanation of the anomaly score with a real case.

[0050] First, the collected temperature parameters, vibration parameters, speed, voltage, pressure values, and other types of parameters are formed into a multidimensional random vector. Each monitoring parameter represents a dimension, forming a multidimensional random vector X=(X (1) ,X (2) ,…,X (d) ), where X (j) represents the jth feature. For example, X (1) represents the vibration vector of sensor 1, X (2)Represents the temperature vector of sensor 2. Use F:R d →[0,1] represents the joint cumulative distribution function (CDF) of all d features. Assume X1,X2,…,X n are samples drawn independently from the same distribution and have a joint CDF. d , use z (j) represents its jth element. Similarly, sample X i The jth element of i (j) The random variable X represents the i Universal random variables with the same distribution.

[0051] According to the definition of joint CDF, for any x∈R d , the joint CDF F(x) represents the probability that all eigenvalues ​​of the random vector X are less than or equal to x at the same time, that is: This probability is a measure of X i An indicator of how "extreme" the left tail is. F(X i ) is smaller, from i Points X drawn from the same distribution satisfy the inequality X≤X i The lower the possibility.

[0052] Assume that the input data , where n is the number of samples and d is the number of features. i (j) represents the jth eigenvalue of the i-th sample. After the ECOD unsupervised anomaly detection algorithm, the output is the anomaly score of the sample data X

[0053] The following are the steps of the anomaly detection algorithm: 1. Calculate the left and right tail ECDF of each feature: For each feature j, the left-tail and right-tail empirical cumulative distribution functions (ECDFs) are calculated: This step aims to quantify the empirical cumulative distribution of each feature in the left tail (low value area) and right tail (high value area). Left-tail ECDF ( ) represents the cumulative probability that the observed value of feature j is less than or equal to z, reflecting the extreme degree of the data in the left tail; the right-tail ECDF ( ) represents the cumulative probability that the observed value of feature j is greater than or equal to z, which is used to capture right-tail anomalies. By constructing the left and right tail distribution functions, a probability basis is provided for the subsequent calculation of anomaly scores, which directly supports the logarithmic transformation of anomaly scores in step 3.

[0054] 2. Calculate the skewness coefficient for each feature: Calculate the sample skewness coefficient of the j-th feature distribution: in is the sample mean of the jth feature. When calculating the sample skewness coefficient, using n-1 as the denominator is to correct the bias of the biased estimate by unbiased estimation. In statistics, when calculating the sample variance, if 1 / n is used to calculate the sample variance, the result will systematically underestimate the population variance, especially when the sample size is small. Using 1 / n-1 can correct this deviation, making the sample variance closer to the true value of the population variance.

[0055] Skewness coefficient ( ) is used to determine the symmetry of the feature distribution. , the distribution is left-skewed, and the left tail abnormality is more significant; if , the distribution is right-skewed or symmetrical, and the right-tail anomaly is more critical. This step quantifies the skew direction of the distribution morphology to provide a decision basis for automatically selecting the left / right tail ECDF in step 3, ensuring that the anomaly detection strategy is adaptively matched to the data distribution characteristics.

[0056] 3. Calculate the anomaly score for each sample: Left tail score (O left-only ): Accumulate the negative logarithm of the left-tail ECDF of each feature, reflecting the global abnormality of the sample in the left tail; right-tail score (O right-only ): Accumulate the negative logarithm of the right-tail ECDF of each feature to capture the right-tail anomaly; automatic scoring (O auto ): Dynamically select left / right tail ECDF according to skewness, taking into account both distribution characteristics and abnormal direction.

[0057] This step integrates the multi-dimensional feature anomaly information into a single sample score, providing a multi-dimensional feature fusion result for the anomaly judgment in step 4.

[0058] 4. Determine the final anomaly score: By taking the maximum value of the left tail, right tail, and automatic score, we ensure that samples with extreme values ​​in any dimension or direction are marked as abnormal. This strategy enhances the sensitivity to compound anomalies, avoids missed detection, and provides a unified quantitative indicator for the anomaly ranking and threshold determination in step 5.

[0059] 5. Return anomaly score: The anomaly scores of all samples returned are O=(O1,…,O n), the higher the score, the greater the probability of abnormality. The higher the abnormality score of the sample point, the higher the probability of abnormality of the sample. Based on the empirical data of gear processing CNC machine tools, we set the abnormality score standard line. Sample points exceeding the abnormality score standard line will be marked as abnormal values, so as to perform subsequent abnormal alarm actions. The joint distribution function has time complexity and expands linearly in both the number of samples and the dimension. Unlike other algorithms, the amount of calculation will not increase sharply with the increase of the dimension.

[0060] In the above embodiment, various monitoring data of the same machine tool can be considered to conform to the same distribution. When monitoring multiple different machine tools, the monitoring data do not conform to the same distribution due to the different working times and types of the monitored machine tools. At this time, the empirical cumulative distribution function (ECDF) is used to directly monitor outliers, which may lead to deviations in the detected abnormal points.

[0061] In order to solve the above problems, when performing a three-level warning of an intelligent model based on multi-dimensional features, the present application provides another embodiment, including the following contents: After forming multiple sets of multi-dimensional vectors from the real-time monitoring operation data of different machine tools, the data are standardized; Gaussian mixture model (GMM) was used for multimodal fitting, and clustering results were obtained after fitting; For each sample x i , calculate its logarithmic probability density logp(x i ). To improve numerical stability, the logarithmic density can be used directly instead of the raw density value.

[0062] Sort the logarithmic density values ​​of all samples, and calculate their ECDF using the above embodiment. For sample x i The ECDF value is ECDF(logp(x i ))= where rank(logp(x i )) is in ascending order (the smallest value is ranked 1).

[0063] Setting Thresholds α (such as 0.05), the ECDF value is less than α The samples are considered abnormal. The density values ​​of these samples are at the lowest level in the entire data set. α At this time, a separate alarm is issued for the machine tool with abnormal values, realizing joint monitoring of multiple machine tools.

[0064] The present invention also provides a monitoring state evaluation system for a gear CNC machining tool based on a multi-source sensor, which is used to implement the steps of the monitoring state evaluation method for a machining tool based on a multi-source sensor, including a client and a server; the client is used for multi-source data acquisition and local calculation, and sends it to the server; the server performs cloud-based calculation of the hybrid reaction index HRI based on the original data and the local calculation results, and sends the calculation results to the client for early warning; the client includes a multi-source data acquisition module, a real-time monitoring module, and an abnormal alarm module; the server includes a hybrid reaction calculation module; The multi-source data acquisition module is used to collect the operation data of the machine tool; the operation data includes the vibration, temperature, speed, motor pressure and motor load of different axes of the CNC machine tool; The real-time monitoring module performs FFT analysis on the monitoring data to generate the change curve of each parameter; The mixed reaction calculation module calculates the mixed reaction index HRI of the tool according to the monitoring data; ； in, represents the temperature index, Indicates the pressure index; It represents the vibration index; The abnormal alarm module uses a joint diagnosis model to monitor and warn based on the calculation results of the mixed reaction index HRI and the change curves of various parameters; it conducts joint alarms based on preset alarm thresholds and automatic monitoring and warnings, and outputs maintenance suggestions.

[0065] Furthermore, it also includes edge computing and cloud storage modules, which use edge computing technology to perform preliminary processing and compression on the collected data locally and upload important data to cloud storage.

[0066] Furthermore, it also includes an edge computing module, which is used to collect vibration data, rotation speed data, temperature data and pressure data at preset time intervals; perform root mean square value calculations respectively; and use an edge computing architecture to perform FFT conversion on the data collected at preset time intervals.

[0067] Furthermore, it also includes a visualization module, which is used to display the change curves of various parameters generated by FFT analysis, real-time monitoring data and fault data.

[0068] The system adopts C / S architecture ( Figure 2), including data acquisition and analysis module, real-time monitoring module, abnormal alarm module, FFT analysis module, and hybrid reaction index (HRI) to comprehensively guide the status monitoring process of machine tools. The main functions of the system include data storage, global monitoring (including time domain and frequency domain information), HRI indicator calculation, status warning and status report, which are mainly implemented on the server side. The client side is responsible for multi-protocol parsing, data preprocessing, local storage, instruction issuance, FFT calculation and status warning. The terminal device is responsible for data acquisition and execution decision-making. The overall architecture shows the process of data being collected from the terminal device, processed by the client, and finally stored and monitored on the server side.

[0069] The system collects machining signals such as vibration, temperature, speed, motor pressure, etc. of different axes of CNC machining machine tools, performs preliminary calculations and analysis, and then uploads the data to the cloud for monitoring, analysis and alarm. Users can use the system software platform to monitor the overall status of the machine tool and the processing status of the machining process, or configure the diagnostic items to be monitored and manually set the alarm threshold and subsequent operations for certain monitoring parameters. The system software platform provides real-time monitoring module, abnormal alarm module, FFT analysis module, and hybrid reaction index (HRI) module to comprehensively guide the state monitoring process of the machine tool. In terms of time domain monitoring and analysis, the system monitors parameters such as vibration, temperature, speed, motor drive load and motor pressure in real time. In terms of frequency domain monitoring and analysis, the system generates single-piece frequency domain diagrams, multi-piece frequency domain diagrams and multi-piece Campbell diagrams to deeply analyze the operating status of the machine tool. Fig. 9 The frequency domain diagram of the C1 spindle of a certain brand of toothed machine is shown. Users can conveniently determine whether there is a potential problem and avoid catastrophic resonance during processing by checking the amplitude at a certain frequency in the frequency domain. The system also has an abnormal alarm function, supports manual setting of alarm thresholds and operations, as well as automatic calculation and alarm operations through algorithms. Once an alarm occurs, the pre-set instructions will be sent to the machine tool in real time. Overall, the system realizes comprehensive monitoring and intelligent evaluation of the machine tool status through multi-dimensional data collection and analysis, which helps to improve the operating efficiency and reliability of the machine tool.

[0070] First, the collected data is preliminarily analyzed on the client side (C side) through the data acquisition and analysis module and uploaded to the system software platform in real time. Then, in the real-time monitoring module on the server side (S side), users can view detailed information of the machining process, such as machining curves, vibration of multiple spindles, speed, current percentage of each motor, vibration information of the top spindle and tool axis, etc. Users can also manually set time domain curve diagnostic items and upper and lower alarm limits.

[0071] In the FFT analysis module on the server side, users can view the processing spectrum information and manually set the frequency / order diagnostic items and alarm upper and lower limits. Then, in the HRI module, users can view the mixed reaction index and manually set the HRI diagnostic items and alarm upper and lower limits.

[0072] Finally, in the abnormal alarm module, the system supports manual settings and intelligent alarm functions, and issues instructions based on the analysis results, and the process ends. The entire system realizes a complete closed loop from data collection, real-time monitoring, analysis to abnormal alarm and instruction issuance. Figure 7 and Figure 8 It is the abnormal situation statistical result of the present invention. Figure 8 Record the results for the unusual problems of the present invention.

[0073] When collecting vibration information of each axis, due to the high collection frequency (sampling rate is 50000 / s), the original collected data is usually compressed. The compression method is to calculate the vibration value within the time period at a higher time interval. For example, the root mean square value (RMS) of the original collected data within the time period is calculated every 86ms, that is, Wherein, N represents the total number of originally collected data.

[0074] In the vibration monitoring system of rotating machinery, in order to effectively reduce the pressure of data transmission and processing, the original collected vibration data is first compressed. Although the amount of compressed data has been significantly reduced compared to the original data, the computational pressure cannot be ignored when performing fast Fourier transform (FFT). Therefore, the system adopts an edge computing architecture (data acquisition and analysis module) to deploy the computationally intensive task of FFT conversion directly on the edge device. After the data acquisition is completed, the edge device immediately performs FFT conversion according to the preset frequency resolution and bandwidth requirements to ensure the real-time and accuracy of frequency domain analysis. The frequency domain data after FFT conversion not only greatly reduces the amount of data, but also retains key frequency information, which is convenient for subsequent fault diagnosis and trend analysis. These frequency domain data are then stored in the local database and uploaded to the server through an efficient data synchronization mechanism for further centralized management and function display. The server is responsible for the integration, visualization and advanced analysis of global data. Through this hierarchical processing method, the system not only improves the efficiency of data processing, but also reduces the demand for network bandwidth, ensuring the real-time response capability and reliability of the system. In addition, the FFT calculation at the edge can dynamically adjust the frequency resolution and bandwidth settings according to actual needs, flexibly respond to different application scenarios, and further improve the adaptability and intelligence of the system.

[0075] Obviously, the above embodiments are only examples for clear explanation, and are not intended to limit the implementation methods. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from them are still within the protection scope of the invention.

Claims

1. A monitoring state assessment method for a machining tool based on a multi-source sensor, characterized in that: The following steps are involved: Using multi-source sensors to collect the operation data of the machine tool; the operation data includes vibration data, temperature data, speed data, motor pressure data, and motor load data of different axes of the CNC machine tool; Based on the monitored data, the mixed reaction index HRI of the tool machine is calculated; ； in, represents the temperature index, Indicates the pressure index; It represents the vibration index; Perform FFT analysis on the vibration data to generate a variation curve of the vibration parameters; perform curve fitting on other data to obtain a time variation curve of other data; Based on the calculation results of the mixed reaction index HRI and the change curves of various data, a joint diagnosis model is used to conduct monitoring and early warning; Combined alarms are issued based on preset alarm thresholds and automatic monitoring warnings, and maintenance recommendations are output.

2. The monitoring state evaluation method for a processing machine tool based on a multi-source sensor according to claim 1 is characterized in that: In the mixed reaction index HRI, Temperature Index +...; Stress Index +...; Vibration Index +...; in, is the axis number and / or tool number selected according to the working condition, x is 1, 2, 3, ...; , represents the temperature of the selected axis / tool, Indicates the temperature offset of the selected axis / tool ​​No.1; Indicates the temperature offset of the selected axis / tool ​​No. 2; All represent the pressure correction value of the drive motor of the selected axis; , , , All are numbers of drive motors; They respectively represent the vibration correction values ​​of the selected No. 1 axis / tool, No. 2 axis / tool, and No. 3 axis / tool.

3. The monitoring state assessment method for a processing machine tool based on a multi-source sensor according to claim 1 is characterized in that: The vibration data is subjected to FFT analysis to generate a variation curve of vibration parameters; and the method further includes generating a single-piece frequency domain diagram, a multi-piece frequency domain diagram and a multi-piece Campbell diagram according to the FFT analysis result.

4. The monitoring state assessment method for a processing machine tool based on a multi-source sensor according to claim 3 is characterized in that: The curve fitting of other data to obtain the time variation curve of other data includes: Collect speed data, temperature data and pressure data at preset time intervals; calculate the root mean square values ​​respectively; use edge computing architecture to generate a time change curve for the data collected at preset time intervals.

5. The monitoring state assessment method for a processing machine tool based on a multi-source sensor according to claim 3 is characterized in that: The combined diagnosis model includes: Level 1 warning based on real-time data collection and analysis, including: real-time monitored operation data is divided into specific sensitive data and general data. When specific sensitive data in the real-time monitored data exceeds the preset threshold, a level 1 warning is issued; Secondary warning based on hybrid reaction index, including: multi-dimensional feature extraction and analysis of real-time monitoring data, when the calculated hybrid reaction index HRI does not meet the preset range, secondary warning is issued; and the change curve formed by the current moment and the previous processing part is extracted to compare the faults; The three-level warning of the intelligent model based on multi-dimensional features includes: forming a multi-dimensional vector of the real-time monitored operation data, selecting the operation data collected each time as samples in turn according to the sampling frequency, calculating the joint distribution function of the selected sample data, outputting the abnormal score of the sample data, and performing a three-level warning.

6. The monitoring state assessment method for a processing machine tool based on a multi-source sensor according to claim 5 is characterized in that: The step of calculating the joint distribution function of the selected sample data and outputting the abnormality score of the sample data includes: For each feature in each sample data, calculate the left tail joint distribution value of the current feature and the right-tailed joint distribution value ; Calculate the left-tail anomaly score, right-tail anomaly score, and automatic detection anomaly score for each sample; The left-tail anomaly score, right-tail anomaly score, and automatic detection anomaly score are fused to return the anomaly scores of all samples.

7. A monitoring state evaluation system for a gear CNC machining machine tool based on multi-source sensors, characterized in that: It includes a client and a server; the client is used for multi-source data acquisition and local calculation, and sends it to the server; The server performs cloud-based calculation of the hybrid reaction index HRI based on the original data and local calculation results, and sends the calculation results to the client for early warning; the client includes a multi-source data acquisition module, a real-time monitoring module, and an abnormal alarm module; The server side includes a mixed reaction calculation module; The multi-source data acquisition module is used to collect the operation data of the machine tool; the operation data includes vibration data, temperature data, speed data, motor pressure data, and motor load data of different axes of the CNC machining machine tool; The real-time monitoring module performs FFT analysis on the vibration data to generate a change curve of the vibration parameters; performs curve fitting on other data to obtain a time change curve of other data; The mixed reaction calculation module calculates the mixed reaction index HRI of the tool according to the monitoring data; ； in, represents the temperature index, Indicates the pressure index; It represents the vibration index; The abnormal alarm module uses a joint diagnosis model to monitor and warn based on the calculation results of the mixed reaction index HRI and the change curves of various data. It also conducts joint alarms based on preset alarm thresholds and automatic monitoring and warnings, and outputs maintenance recommendations.

8. The monitoring state evaluation system for gear CNC machining machine tools based on multi-source sensors according to claim 7 is characterized in that: It also includes edge computing and cloud storage modules, which use edge computing technology to perform preliminary processing and compression on the collected data locally and upload important data to cloud storage.

9. The monitoring state evaluation system for gear CNC machining machine tools based on multi-source sensors according to claim 8, characterized in that: The edge computing module is used to collect rotation speed data, temperature data and pressure data at preset time intervals; perform root mean square value calculations respectively; and use the edge computing architecture to generate a time change curve for the data collected at preset time intervals.

10. The monitoring state evaluation system for gear CNC machining machine tools based on multi-source sensors according to claim 7, characterized in that: It also includes a visualization module, which is used to display the change curves of various parameters generated by FFT analysis, real-time monitoring data and fault data.

Images