Adaptive real-time tool breakage detection method based on spindle current in CNC machine tools

Abstract

The adaptive real-time tool breakage detection method based on spindle current in CNC machine tools proposed in the present invention is suitable for economical CNC systems that do not have high-end sensor configurations. This method obtains the spindle current signal through a low-sampling rate current sensor, and combines processing state identification, historical template comparison and normalized correlation coefficient calculation to achieve real-time online detection of tool breakage events. The system accurately identifies the starting point of the processing section through the processing state alignment mechanism of "coarse alignment + fine correction", and identifies two typical tool breakage behaviors of "unloaded" and "contact abnormality" in combination with the current fluctuation characteristics. At the same time, adaptive temperature drift calibration and historical template self-update strategies are designed to improve the long-term stability and environmental adaptability of the system. This method has the advantages of low cost, strong robustness and short response time, and is particularly suitable for tool breakage fault detection and intelligent upgrading in small and medium-sized processing workshops.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and fault identification in a numerical control machining process, and in particular to an adaptive real-time tool breakage detection method based on spindle current in a numerical control machine tool. Background Art

[0002] Traditional broken tool detection methods often rely on high-frequency vibration sensors, force sensors, or vision systems. While highly accurate, these methods are expensive and complex to maintain, making them difficult to adopt in resource-limited machining scenarios. Some broken tool detection technologies that rely on current signals suffer from issues such as inaccurate machining state recognition, fixed thresholds, and insufficient compensation for temperature drift. These issues lead to frequent misjudgments and missed detections, making them incapable of meeting actual industrial needs. Therefore, a broken tool detection method that is highly robust, low-cost, automatically adapts to new working conditions, and can accommodate fluctuating machining conditions is urgently needed.

[0003] The present application is compared with the prior art as follows:

[0004] Technical comparison with application document CN105619178A “A real-time detection method for broken tool in CNC machine tools”;

[0005] Application CN105619178A discloses a real-time tool breakage detection method based on the average current variation of the spindle motor. Specifically, the method comprises the following steps: Step 1: The CNC system reads the average normal machining current and the average no-load current stored in the tool process database; Step 2: A current sampling timer is set; Step 3: Each L sampling points is considered an analysis segment, and the average spindle current is calculated for each analysis segment; Step 4: The current average variation of three adjacent analysis segments is continuously compared using a sliding window method, and a fixed proportional coefficient is used to determine whether the tool breakage condition is met; Step 5: If the tool breakage condition is met, an alarm is triggered and the machining process is terminated. This method, integrated into the CNC system, utilizes the time-domain current characteristics of the spindle motor for detection. It is suitable for rough machining scenarios with large spindle current variations and has the advantage of not requiring additional sensors. Its focus is on real-time detection of tool breakage events by analyzing the average current variation trend.

[0006] 2. This application proposes an adaptive real-time self-breaking tool detection method based on spindle current. This method collects low-frequency spindle current signals by connecting low-cost current transformers in series, and combines modules such as spindle start-stop mutation recognition, historical template matching alignment, and fluctuation trend judgment to achieve high-precision time alignment of the current signal and historical data, identify the starting point of the processing segment under multi-tool and multi-process conditions, and construct the aligned current sequence. Two types of judgment models are used to detect "unloaded" and "contact abnormality" tool breaking behaviors respectively; at the same time, temperature drift calibration and historical template self-update mechanisms are designed to adapt to changes in factors such as current drift and tool wear during the processing process, and achieve long-term stable operation of the system. This method does not rely on high-frequency sampling or complex sensor systems, and has the characteristics of high robustness, self-updateability, and low-cost deployment. It focuses on high-precision alignment at the processing segment level and classification and identification of typical tool breakage types, and improves environmental adaptability under complex working conditions.

[0007] There are essential differences between the two in broken knife identification mechanism, real-time data alignment granularity and system design path.

[0008] Technical comparison with application document CN114700802A "Broken Knife Detection Method and Device";

[0009] 1. Application CN114700802A discloses a tool breakage detection method and device. Specifically, during the machine tool cutting process, the current tool identification, machine current, and machining position are periodically acquired, and the sampled data for each tool is recorded in a corresponding file. At the end of machining, for each tool file, the corresponding machining process is determined based on the machining position, and the load threshold corresponding to each process is calculated and recorded based on the machine current sampled during each process. Tool breakage detection is performed based on the recorded load thresholds for each tool and each process. This patent considers the automated calculation of load thresholds in multi-tool and multi-process scenarios, aiming to improve the efficiency and accuracy of tool breakage detection. It focuses on automatically dividing processes based on machining position and using historical sampled data (primarily weighted by the average and maximum current values) to calculate a static load threshold for each process, enabling tool breakage detection by tool and process.

[0010] 2. This application proposes an adaptive real-time self-breaking tool detection method based on spindle current. This method collects low-frequency spindle current signals by connecting low-cost current transformers in series, and combines modules such as spindle start-stop mutation recognition, historical template matching alignment, and fluctuation trend judgment to achieve high-precision time alignment of the current signal and historical data, identify the starting point of the processing segment under multi-tool and multi-process conditions, and construct the aligned current sequence. Two types of judgment models are used to detect "unloaded" and "contact abnormality" tool breaking behaviors respectively; at the same time, temperature drift calibration and historical template self-update mechanisms are designed to adapt to changes in factors such as current drift and tool wear during the processing process, and achieve long-term stable operation of the system. This method does not rely on high-frequency sampling or complex sensor systems, and has the characteristics of high robustness, self-updateability, and low-cost deployment. It focuses on high-precision alignment at the processing segment level and classification and identification of typical tool breakage types, and improves environmental adaptability under complex working conditions.

[0011] There are essential differences between the two in terms of broken tool identification mechanism, process division principle and environmental adaptation mechanism.

[0012] Technical comparison with application document CN115781411A "Online tool breakage detection method, device and system"

[0013] 1. Application CN115781411A discloses an online tool breakage detection method, specifically: detecting the current no-load voltage value corresponding to the spindle no-load current under current machining conditions and the real-time machining voltage value corresponding to the real-time machining current; calculating the difference between the two to obtain a real-time voltage difference; obtaining a pre-set reference voltage threshold (determined based on the difference between the normal machining voltage and the no-load voltage); and determining tool breakage by comparing the real-time voltage difference with the reference voltage threshold. This patent focuses on a voltage difference comparison mechanism, utilizing a dynamic threshold for the voltage difference between the no-load and machining states to determine tool breakage. Its core technology ensures threshold reliability through high-precision voltage sampling and calibration, and optimizes detection stability for multi-spindle and multi-plate scenarios in PCB machining. It focuses on addressing voltage signal acquisition accuracy issues, using a pre-set fixed reference voltage threshold for binary judgment (yes / no tool breakage).

[0014] 2. This application proposes an adaptive real-time self-breaking tool detection method based on spindle current. This method collects low-frequency spindle current signals by connecting low-cost current transformers in series, and combines modules such as spindle start-stop mutation recognition, historical template matching alignment, and fluctuation trend judgment to achieve high-precision time alignment of the current signal and historical data, identify the starting point of the processing segment under multi-tool and multi-process conditions, and construct the aligned current sequence. Two types of judgment models are used to detect "unloaded" and "contact abnormality" tool breaking behaviors respectively; at the same time, temperature drift calibration and historical template self-update mechanisms are designed to adapt to changes in factors such as current drift and tool wear during the processing process, and achieve long-term stable operation of the system. This method does not rely on high-frequency sampling or complex sensor systems, and has the characteristics of high robustness, self-updateability, and low-cost deployment. It focuses on high-precision alignment at the processing segment level and classification and identification of typical tool breakage types, and improves environmental adaptability under complex working conditions.

[0015] There are essential differences between the two in core detection principles, real-time data alignment granularity, and environmental adaptation mechanisms. Summary of the Invention

[0016] To address the challenges of existing CNC machining tool breakage detection methods, such as high deployment costs, low machining state recognition accuracy, poor environmental interference resistance, and a lack of adaptability, this paper proposes an adaptive, real-time tool breakage detection method based on spindle current for CNC machine tools. This method aims to achieve efficient and reliable identification of tool breakage at low cost and low sampling rates. By integrating the current variation characteristics of the spindle start-up and shutdown processes, a template matching and alignment mechanism, current fluctuation analysis and tool breakage anomaly pattern recognition, and a self-updating mechanism for historical current templates, the system achieves rapid and accurate detection of two typical tool breakage types ("no-load tool breakage" and "contact anomaly"), enhancing the intelligent monitoring capabilities of CNC systems in complex machining environments.

[0017] To achieve the above object, the technical solution adopted by the present invention is:

[0018] The method for detecting a tool breakage in a CNC machine tool based on spindle current in real time and adaptively includes the following specific steps:

[0019] 1) Obtain the current signal of the spindle motor and transmit it to the industrial control system in real time;

[0020] 2) Obtain the temperature drift correction value in the template library and calculate the calibrated current;

[0021] 3) Identify the start time of rough machining based on the current mutation during the spindle start and stop process ;

[0022] 4) Extract the machining state recognition template sequence of the corresponding tool in the historical machining data and historical processing current ;

[0023] 5) In Perform sliding matching search nearby, calculate the normalized correlation coefficient NCC, and obtain the optimal alignment time , and construct the calibrated and aligned machining current sequence ;

[0024] 6) Use the current fluctuation trend to identify the broken knife state, including load loss and contact abnormality;

[0025] 7) In the absence of abnormal cycles, the historical template library is automatically updated to improve the model's adaptability.

[0026] As a further improvement of the present invention, the calculation formula of the normalized correlation coefficient NCC in step 5) is:

[0027]

[0028] in is the template sequence, is a sliding window sequence, and Sequence and No. elements, and are their means respectively, N is the length of T and S sequences, when the maximum normalized correlation coefficient NCC is lower than the threshold When the system fails to recognize the starting point, it will record the failure and skip the detection of this cycle.

[0029] As a further improvement of the present invention, the step 6) detecting a broken knife during the process of identifying a broken knife state by using a current fluctuation trend includes:

[0030] Loss-of-load identification: If the absolute value of the difference between the calibrated real-time current and the historical no-load value is continuously smaller than the first threshold, and the absolute value of the difference between the calibrated real-time current and the historical no-load value is continuously smaller than the second threshold, it is determined to be a loss-of-load knife break;

[0031] Abnormal contact identification: If the difference between the calibrated real-time current and the historical processing current exceeds the third threshold continuously, it is determined to be an abnormal contact tool breakage.

[0032] As a further improvement of the present invention, the self-update mechanism steps of the template library in step 7) updating the historical template library are as follows:

[0033] During the first machining under new working conditions, the current of one no-load machining and one normal machining is recorded, and the no-load and load machining intervals of the historical current are obtained by difference comparison and included in the historical template library; when there is no tool breakage abnormality in the current machining cycle, the mechanism automatically includes the current data of the current machining segment in the historical template library for reference in the next machining cycle.

[0034] The present invention includes the following key steps and modules:

[0035] 1. Current acquisition and preprocessing;

[0036] Low-cost current transformers are installed in series with the spindle input circuit of a CNC machine tool. The acquisition frequency is approximately 30 Hz, and the analog signal is transmitted to the industrial control host via an ADC module, creating a real-time current data stream. A ring buffer structure is used to achieve efficient data reading and storage, avoiding frame loss or processing delays.

[0037] 2. Processing segment starting point identification and alignment mechanism;

[0038] Combined with the sudden change characteristics of the spindle start-stop current, the system uses the "current sudden rise point" as the basis for rough identification of the initial starting point to obtain the processing start time . Construct a sliding window sequence around this time point and compare it with the processing state recognition template sequence Calculate the normalized correlation coefficient (NCC) and select the maximum correlation point as the starting point for final precise processing This mechanism effectively eliminates misalignment problems caused by spindle response delays and unstable machining cycles, and the alignment accuracy can be controlled within one sampling cycle.

[0039] 3. Dynamic historical template modeling and update mechanism;

[0040] In the initial stage, the system first collects the no-load and normal processing data of each tool, builds a processing current template library, and divides the load range into and no-load interval , reference current average and processing waveform To adapt to current drift caused by temperature drift in the machining environment and tool wear, a dynamic update mechanism for historical templates is introduced: if no abnormality occurs in the current cycle, the machining current data of that period is used to update the template, improving the robustness and accuracy of the system in long-term operation.

[0041] 4. Broken blade state recognition model;

[0042] Based on the difference between the real-time current of the machining section and the historical baseline value, the system identifies the following two types of tool breakage behaviors:

[0043] (1) Type I: Unloaded (the tool breaks and is thrown away). The current differential fluctuation is significantly reduced, and the overall value is close to the no-load reference current, indicating that the tool breaks and loses the load.

[0044] (2) Type II: Contact abnormality (the broken part of the tool remains in the workpiece). The real-time current is significantly higher than the historical processing value and fluctuates violently, indicating that the remaining parts of the broken tool are still in contact with the material and generate additional cutting resistance. This is a hidden high-risk fault.

[0045] 5. Current correction parameter calibration mechanism;

[0046] During a long continuous machining cycle, the collected current value is greatly affected by temperature. Therefore, a temperature drift correction is introduced to calibrate the current value: when a tool machining cycle is completed normally, the average current value of the no-load segment of the current machining cycle is calculated based on the historical no-load interval. and The difference is used as the new temperature drift calibration correction , achieving continuous dynamic calibration of the current reference.

[0047] 6. Alarm and machine tool reverse control interface linkage;

[0048] When the system detects a broken tool, it can send an alarm signal to the CNC control unit, triggering a shutdown operation, achieving equipment-level linkage, and avoiding further processing errors or equipment damage.

[0049] Beneficial effects: Compared with the prior art, the present invention has the following advantages and practical value:

[0050] (1) Low-cost deployment: No high-frequency current sensors, vibration sensors, or vision systems are required. Broken knife detection can be achieved by relying only on low-speed current transformers, significantly reducing hardware costs.

[0051] (2) High alignment accuracy and response speed: Combining event-driven and sliding template matching, the alignment error of the processing segment can be as low as one sampling cycle, ensuring the accuracy of subsequent judgment logic;

[0052] (3) Strong adaptive anti-interference capability: By introducing a dynamic history template library and self-update mechanism, the system can adapt to the influence of temperature drift, tool wear and load changes, and has long-term operational stability;

[0053] (4) Supports classification and identification of typical tool breakage situations: It can accurately identify two types of tool breakage behaviors: load loss type and contact load type, with wide coverage, and the misjudgment rate and missed judgment rate are both less than 1%;

[0054] (5) Strong adaptability to industrial sites: The system has been successfully deployed on a conventional three-axis CNC platform, has good adaptability and real-time response performance, and is suitable for a variety of small-batch high-frequency tool change scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flow chart of the method of the present invention;

[0056] Figure 2 This is a flow chart of the state identification module of the present invention;

[0057] Figure 3 This is a flow chart of the broken knife detection module of the present invention;

[0058] Figure 4 This is the current curve when the present invention detects a broken knife in engineering applications;

[0059] Figure (a) shows the current curve when the abnormal state is the load-loss type cutter breakage, and Figure (b) shows the current curve when the abnormal state is the contact abnormal type cutter breakage.

[0060] Figure 5 This is a flow chart of the template library update module of the present invention. DETAILED DESCRIPTION

[0061] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0062] Example 1: System composition and deployment method;

[0063] As attached Figure 1 The broken tool detection system constructed by the present invention is applicable to the standard CNC machine tool platform. The system as a whole includes the following six core functional modules:

[0064] (1) Current acquisition module: A current transformer is installed non-invasively on the spindle input side of the CNC machine tool, with a sampling frequency set to 30 Hz. The sampled current signal is transmitted to the industrial control host in real time via an analog-to-digital converter (ADC), achieving high-efficiency, low-latency data acquisition. The data structure uses a ring buffer mechanism to ensure continuity and real-time performance during the acquisition process.

[0065] (2) State Identification Module: By detecting the spindle start and stop current change trend, the start and end time of the tool from "tool change → processing → tool change" can be accurately identified. Among them, the spindle current will suddenly change at the moment of starting processing, which serves as the key feature for identifying the starting point of processing.

[0066] (3) Data matching and alignment module: Combining the dual strategies of start-stop mutation point identification and sliding template matching, a "rough alignment + fine correction" processing start time identification mechanism is constructed. The optimal matching point is selected by calculating the normalized correlation coefficient (NCC). , achieving high-precision time alignment of the spindle machining current.

[0067] (4) Tool breakage detection module: Based on the real-time current fluctuation characteristics and historical template deviation during the machining process, it identifies tool breakage behavior. It can identify "unloaded" and "abnormal contact" tool breakage, improving detection sensitivity and risk response capabilities.

[0068] (5) Dynamic template library update module: At the beginning of processing, the system needs to continuously collect two sets of reference data: no-load rotation current and normal load processing current, and extract the load range and no-load interval , establish a processing status recognition template , no-load benchmark And historical processing current The system automatically updates the historical template using machining current data, assuming no tool breakage is detected during the current machining cycle. A current correction parameter calibration mechanism compensates for slow-changing factors such as temperature drift and machine aging, maintaining the model's stability and accuracy over long-term use.

[0069] The above modules can be integrated into existing industrial control systems with low deployment costs and small renovation workload, and have good engineering feasibility and on-site promotion value.

[0070] Example 2: Processing segment starting point identification and time alignment method;

[0071] The present invention proposes a fusion spindle current alignment mechanism, the steps of which are as follows:

[0072] Rough alignment stage: as shown in the attached Figure 2 During the machining process, detect the sudden rise point of the spindle current. Set a number of consecutive sampling points. When the current exceeds the "state recognition threshold" for the first time, it will be recognized as the starting point of rough machining. .

[0073] Fine alignment stage:

[0074] (1) Extracting fixed-length template sequences from historical data (e.g. 30 sampling points);

[0075] (2) For the current sequence ,by As the starting point, build a sliding window sequence , in the range of 10 sampling points and Calculate the normalized correlation coefficient (NCC):

[0076]

[0077] in is the template sequence, is a sliding window sequence, and Sequence and No. elements, and are their means respectively, N is the length of T and S sequences, when the maximum normalized correlation coefficient NCC is lower than the threshold When the system fails to recognize the starting point, it will record the failure and skip the detection of this cycle.

[0078] (3) Find the matching position with the largest correlation coefficient as the final processing starting point The current value at this point is used as the machining current sequence after the current tool is aligned. The first value of

[0079] (4) If the maximum NCC is less than the preset threshold If the value is 0.75, it is recorded as alignment failure and the cycle detection can be skipped or manual intervention can be triggered.

[0080] This mechanism controls the alignment error of the machining start point to within 1 sampling cycle, which is significantly better than the traditional machining segment identification method based on CNC status codes. The latter is limited by communication delays and system refresh cycles, and the error is often as high as hundreds of milliseconds.

[0081] Example 3: Broken knife state identification method;

[0082] As attached Figure 3 After the machining section is identified, when the cutting process enters the load range, belong , system compares real-time current With history template , based on the current change trend and deviation characteristics, the broken knife behavior is identified and divided into the following two typical modes:

[0083] (1) Type I: Unloaded blade breakage (the broken part of the tool is thrown away);

[0084] The manifestation is that the spindle load suddenly drops, and the current drops rapidly and tends to be stable. The judgment criteria are as follows:

[0085] ① Current differential m consecutive times below the first threshold ;

[0086] ② and and The absolute value of the difference is lower than the second threshold .

[0087] If the above conditions are met, it is determined that the tool has broken and lost load.

[0088] (2) Type II: abnormal contact and tool breakage (the broken part of the tool remains in the workpiece);

[0089] This is manifested as the remaining tool remaining in contact with the workpiece surface after the break, causing an abnormal increase in current. The judgment logic is:

[0090] like - continuous above the third threshold ; It is judged that there is abnormal contact load, which is a high-risk hidden tool breakage.

[0091] like Figure 4 , where Figure (a) is the current curve when the abnormal state is the load-loss type tool breakage, and Figure (b) is the current curve when the abnormal state is the contact abnormal type tool breakage. According to the attached figure, the real-time current and reference current curves of the two types of tool breakage can be compared; once the above two types of conditions are detected, the system will trigger a tool breakage alarm signal and link the CNC system to execute a shutdown operation.

[0092] Example 4: Template update and temperature drift calibration mechanism;

[0093] Since the performance of electronic components such as ADC and current transformer is affected by temperature, the measured current value also tends to change with temperature. In order to improve the long-term stability of the system and avoid finding the complex nonlinear relationship between multiple components, temperature and current offset to calibrate the current value, the present invention establishes the following adaptive update strategy, as shown in the attached figure. Figure 5 :

[0094] (1) At the beginning of processing, the system needs to collect two sets of reference data: no-load rotation current and normal load processing current, and complete them within 2 hours. Extract the processing section and no-load section data respectively, and establish a processing status recognition template , no-load benchmark , load range , no-load section , Historical cutting temperature drift correction and historical processing current Standard current library.

[0095] (2) If the machining cycle of a single tool is within 1 hour, it is considered that the ambient temperature change is extremely small and the current offset value can be ignored. The current value calibration within the current cycle can use the same temperature drift calibration correction value. ;

[0096] (3) If the current processing cycle is completed normally, the current data of the current processing section will be automatically included in the template library to replace the old value; the average current of the no-load section of the current processing cycle will be calculated. and The difference is used as the new temperature drift calibration correction .

[0097] This mechanism can continuously adapt to current drift problems caused by room temperature changes, tool wear, workpiece hardness changes, etc., and reduce false alarm rates.

[0098] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent variation based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. An adaptive real-time self-breaking tool detection method based on spindle current for CNC machine tools, characterized by: The specific steps include: 1) Obtain the current signal of the spindle motor and transmit it to the industrial control system in real time; At the beginning of processing, the system needs to continuously collect two sets of reference data: no-load rotation current and normal load processing current, and extract the load range. and no-load interval , establish a processing status recognition template , no-load benchmark and historical processing current sequence Standard current library; 2) Obtain the temperature drift correction value in the template library and calculate the calibrated current; 3) Identify the start time of rough machining based on the current mutation during the spindle start and stop process ; 4) Extract the machining status identification template of the corresponding tool from the historical machining data and historical processing current sequence ; 5) In Perform sliding matching search nearby, calculate the normalized correlation coefficient NCC, and obtain the optimal alignment time , and construct the calibrated and aligned machining current sequence , where the fine alignment stage is as follows; (1) Extracting fixed-length template sequences from historical data ; (2) For the current machining current sequence ,by As the starting point, build a sliding window sequence , in the range of 10 sampling points and Calculate the normalized correlation coefficient NCC. The calculation formula of the normalized correlation coefficient NCC is: ； in is the template sequence, is a sliding window sequence, and Sequence and No. elements, and are their means respectively, N is the length of T and S sequences, when the maximum normalized correlation coefficient NCC is lower than the threshold When the system records the failure of starting point recognition and skips the detection of this cycle; 6) Use the current fluctuation trend to identify the broken knife state, including load loss and contact abnormality; After the machining section is identified, when the cutting process enters the load range , the system compares the real-time processing current and historical processing current ,identify the broken knife behavior based on the current change trend and deviation characteristics; 7) In the absence of abnormal cycles, the historical template library is automatically updated to improve the model's adaptability.

2. The method for detecting tool breakage in a CNC machine tool based on spindle current in real time according to claim 1, characterized in that: The step 6) of identifying the broken knife state by using the current fluctuation trend includes the following steps: Loss-of-load identification: If the absolute value of the difference between the calibrated real-time current and the historical no-load value is continuously smaller than the first threshold, and the absolute value of the difference between the calibrated real-time current and the historical no-load value is continuously smaller than the second threshold, it is determined to be a loss-of-load knife break; Abnormal contact identification: If the difference between the calibrated real-time current and the historical processing current exceeds the third threshold continuously, it is determined to be an abnormal contact tool breakage.

3. The method for detecting tool breakage in a CNC machine tool based on spindle current in real time according to claim 1, wherein: The steps of the self-update mechanism of the template library in step 7) updating the historical template library are as follows: During the first machining under new working conditions, the current of one no-load machining and one normal machining is recorded, and the no-load and load machining intervals of the historical current are obtained by difference comparison and included in the historical template library; when there is no tool breakage abnormality in the current machining cycle, the mechanism automatically includes the current data of the current machining segment in the historical template library for reference in the next machining cycle.

Images