An intelligent diagnosis method and system for equipment failure of a forging production line

Abstract

The present application belongs to the technical field of industrial data processing, and particularly relates to a kind of equipment fault intelligent diagnosis method and system for forging production line, and the method comprises: collecting the current signal of sawing machine main shaft motor to extract effective cutting segment data sequence;Intercept the initial stable segment data of effective cutting segment, obtain refined data set by eliminating abnormal data points, and calculate the refined material benchmark and standard deviation of refined data set;Based on the refined material benchmark and standard deviation, a material cutting instability index for evaluating the inherent fluctuation characteristics of material cutting is constructed;Based on the refined material benchmark, standard deviation and material cutting instability index, an adaptive threshold is generated to realize the diagnosis of equipment failure.The present application can automatically adapt to the load benchmark and fluctuation characteristics of different materials, solves the false alarm and missed alarm problem of fixed threshold under multi-material working condition, and improves the accuracy and reliability of diagnosis.

Description

Technical Field

[0001] This invention relates to the field of industrial data processing technology. More specifically, this invention relates to an intelligent diagnostic method and system for equipment faults in forging production lines. Background Technology

[0002] In the forging industry, raw material preparation is fundamental to ensuring the smooth operation of subsequent high-cost processes, which rely on large saws to precisely cut bars or profiles. The health of the saw blade directly determines the quality of the cut section and the material utilization rate, while saw blade wear, especially dulling and tooth breakage, is the most common and highly destructive failure mode in this process.

[0003] Currently, to ensure production, in addition to traditional manual inspections and experience-based periodic replacements, existing technology also uses data acquisition devices to collect the current signal of the saw spindle motor in real time. This is because the signal can directly reflect changes in the cutting load. Specifically, the time-domain statistical characteristics of this current signal are calculated. For example, a simple moving average is used to reflect the average load to detect passivation, and kurtosis is used to detect signal spikes to identify tooth breakage. These calculated statistical values ​​are then compared with pre-set fixed thresholds to determine whether a fault has occurred.

[0004] However, the production of high-value precision forgings often requires handling multiple metals with different physical properties, such as high-hardness high-temperature alloys, titanium alloys, and low-hardness aluminum alloys. When cutting high-hardness titanium alloys, the average current in its healthy state may be much higher than the average current in its passivated state when cutting low-hardness aluminum alloys. If the system uses a fixed threshold calibrated for common steels, it will cause the system to falsely and frequently alarm for saw blade passivation when cutting healthy titanium alloys, resulting in false alarms. Conversely, when cutting passivated aluminum alloys, the system will fail to detect the fault, resulting in serious missed alarms. This coexistence of false alarms and missed alarms makes fixed-threshold-based diagnostic systems difficult to apply on multi-material production lines. Summary of the Invention

[0005] To address the technical problem that existing diagnostic methods based on fixed thresholds cannot adapt to various material conditions and thus fail to provide diagnostic solutions, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides an intelligent equipment fault diagnosis method for forging production lines, comprising: acquiring the current signal of the saw spindle motor, automatically identifying the start and end times of the cutting task based on the current signal, and extracting an effective cutting segment data sequence; extracting the initial stable segment data of the effective cutting segment data sequence, removing abnormal data points in the initial stable segment data through a self-filtering method based on robust statistics to obtain a refined dataset, and calculating the mean of the refined dataset as a refined material benchmark and calculating the standard deviation of the refined dataset; constructing a material cutting instability index for evaluating the inherent fluctuation characteristics of material cutting based on the refined material benchmark and the standard deviation; generating an adaptive threshold based on the refined material benchmark, the standard deviation, and the material cutting instability index, the adaptive threshold including an adaptive passivation threshold and an adaptive spike threshold, and realizing equipment fault diagnosis for forging production lines by comparing real-time current characteristics with the adaptive threshold.

[0007] This invention extracts a highly robust material reference current using a self-filtering method and constructs a new material cutting instability index. Based on this, an adaptive passivation threshold is generated that considers both the reference and instability, while an adaptive spike threshold considers both the reference and the initial fluctuation standard deviation. This allows the threshold to automatically adapt to different material properties, effectively solving the problem of false alarms and missed alarms under multiple material conditions with fixed thresholds, and significantly improving the accuracy and reliability of saw blade wear diagnosis.

[0008] Preferably, the automatic identification of the start and end times of the cutting task based on the current signal includes: offline calibration of the idle current baseline and idle current standard deviation when the saw is idling; when the real-time current is continuously higher than the sum of the idle current baseline and the idle current standard deviation by a preset multiple for a first preset duration, the cutting is determined to have started; when the real-time current is continuously lower than the sum of the idle current baseline and the idle current standard deviation by a preset multiple for a second preset duration, the cutting is determined to have ended.

[0009] In this way, the idle and load phases can be automatically distinguished, avoiding contamination of subsequent statistical analysis by low-amplitude idle data.

[0010] Preferably, the step of removing outlier data points from the initial stable segment data using a self-filtering method based on robust statistics to obtain a refined dataset includes: calculating the median and interquartile range of the initial stable segment data; constructing an upper threshold and a lower threshold for a statistical confidence interval based on the median and the interquartile range; traversing the initial stable segment data and retaining data points falling within the statistical confidence interval based on the upper threshold and the lower threshold to form the refined dataset.

[0011] By constructing a filtering interval using the median and interquartile range, which are insensitive to outliers, it is possible to effectively remove any abnormal spikes that may exist in the initial stable segment of data, making the benchmark for subsequent calculations very stable and accurately reflecting the true basic load of the material.

[0012] Preferably, the material cutting instability index satisfies the following relationship: ;in, This is an index of material cutting instability. Let be the standard deviation of the refined dataset. This serves as the benchmark for the refined materials.

[0013] It can simultaneously capture the basic load and cutting stability of materials, and quantitatively reflect the cutting difficulty and inherent fluctuation characteristics of different materials.

[0014] Preferably, the adaptive passivation threshold satisfies the following relationship: ;in, To adaptive passivation threshold, Based on the refined material, This is an index of material cutting instability. Based on passivation factor, It is an instability passivation factor.

[0015] The passivation threshold is dynamic, adapting to both the baseline and stability, making passivation detection both sensitive and reliable under various material conditions.

[0016] Preferably, the adaptive peak threshold satisfies the following relationship: ;in, To adapt to peak threshold, Based on the refined material, Let n be the standard deviation of the refined dataset, and n be the multiple of the standard deviation.

[0017] This allows the adaptive peak threshold to automatically adapt to the normal fluctuation range of different materials, improving the reliability of tooth breakage detection.

[0018] Preferably, the initial stable segment data of the effective cutting segment is specifically obtained by: skipping a preset entry impact time after the start time of the effective cutting segment data sequence, and extracting data of a subsequent preset base time period as the initial stable segment data.

[0019] By skipping the preset entry impact time, the initial entry impact of the cutting task can be actively avoided, preventing this inevitable but non-faulty drastic fluctuation from contaminating subsequent baseline calculations.

[0020] Preferably, the step of comparing real-time current characteristics with the adaptive threshold to achieve equipment fault diagnosis for forging production lines includes: calculating the operating average value of the current signal during the real-time monitoring phase; and determining that a saw blade passivation fault has occurred when the operating average value is greater than the adaptive passivation threshold and the duration exceeds a preset passivation confirmation time.

[0021] Preferably, the method of diagnosing equipment faults for forging production lines by comparing real-time current characteristics with the adaptive threshold further includes: monitoring the instantaneous value of the current signal during the real-time monitoring phase; determining that an abnormal peak has been detected when the instantaneous value is greater than the adaptive peak threshold; and confirming that a saw blade tooth breakage fault has occurred when the number of abnormal peaks detected within a preset peak counting time window exceeds a preset peak count value.

[0022] Secondly, the present invention provides an intelligent equipment fault diagnosis system for forging production lines, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned intelligent equipment fault diagnosis method for forging production lines is implemented.

[0023] By adopting the above technical solution, a computer program is generated from the above-mentioned intelligent equipment fault diagnosis method for forging production lines and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of terminal equipment based on the memory and processor, making it convenient to use.

[0024] This invention extracts a highly robust material reference current through a self-filtering method and constructs a new material cutting instability index. Based on this, a passivation and spike threshold that dynamically adjusts with working conditions is generated, which can automatically adapt to the load and fluctuation characteristics of different materials. This solves the problem of false alarms and missed alarms under multiple material working conditions with fixed thresholds, and significantly improves the accuracy and reliability of saw blade wear diagnosis.

[0025] Furthermore, by automatically identifying effective cutting segments, interference from idle data was eliminated, and by skipping the initial part when extracting the initial stable segment data, the impact of the entry impact was eliminated. The entire diagnostic process enhanced its robustness against noise and interference at multiple stages, ensuring the stability of the diagnostic results in complex production environments. Attached Figure Description

[0026] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0027] Figure 1This is a flowchart illustrating an intelligent equipment fault diagnosis method for a forging production line according to the present invention.

[0028] Figure 2 This is a schematic diagram comparing the diagnostic effects of adaptive passivation threshold and fixed threshold in an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram illustrating the adaptive spike threshold diagnostic effect on tooth breakage in an embodiment of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] This invention discloses an intelligent fault diagnosis method for equipment in a forging production line, referring to... Figure 1 This includes steps S1-S4:

[0033] S1. Acquire the current signal of the saw spindle motor, and automatically identify the start and end times of the cutting task based on the current signal, and extract the effective cutting segment data sequence.

[0034] In an optional embodiment, the current signal of the saw spindle motor can be acquired in real time at a high frequency, such as 10 kHz, by a data acquisition device, and the start and end times of the cutting task can be automatically identified based on the current signal to extract the effective cutting segment data sequence.

[0035] Specifically, current signals are collected for a period of time while the saw is idling, and their average value and standard deviation are calculated to calibrate a reliable idle current baseline and idle current standard deviation offline. Then, the collected current signals are monitored in real time. When the real-time current remains above the idle current baseline and a preset multiple (e.g., 0.5 seconds) for a first preset duration, the system determines that cutting has begun and records this time as the start time. When the real-time current remains below the idle current baseline and a preset multiple of the idle current standard deviation for a second preset duration, the system determines that cutting has ended and records this time as the end time.

[0036] For example, assuming the offline calibration yields an idle speed baseline of 5A and an idle current standard deviation of 0.5A, the upper limit of idle speed fluctuation is calculated to be 5A + 3 × 0.5A = 6.5A. In real-time monitoring, when the collected current is greater than 6.5A for 0.5 seconds, the system marks this moment as the start time; subsequently, when the current is lower than 6.5A for 1 second, the system marks this moment as the end time.

[0037] Furthermore, the system can accurately extract the valid segment data sequence corresponding to this cutting task from the memory buffer or real-time data stream based on the identified start and end times.

[0038] In this way, through automated operating condition identification and data segmentation, it can be ensured that subsequent diagnostic analysis is only performed on valid load data, eliminating the interference of no-load idling signals and improving the relevance and accuracy of the analysis.

[0039] S2. Extract the initial stable segment data from the effective segment data sequence, remove outlier data points from the initial stable segment data using a self-filtering method based on robust statistics to obtain the refined dataset, and calculate the mean of the refined dataset as the benchmark for refined materials and the standard deviation of the refined dataset.

[0040] In an optional embodiment, a data segment that reflects the working condition while avoiding interference is extracted from the beginning of the obtained effective cut segment data sequence as the initial stable segment data. Specifically, after the start time of the effective cut segment data sequence, a preset entry impact time is skipped, for example, 1 second, and subsequent preset base time period lengths, for example, 10 seconds, are extracted as the initial stable segment data.

[0041] Furthermore, firstly, the median and interquartile range of the initial stable segment data are calculated. The median is insensitive to outliers, while the interquartile range can assess the main dispersion range of the data. Then, based on the obtained robust statistics, upper and lower thresholds for the statistical confidence interval are constructed. A self-filtering method is then used to identify and remove any abnormal spikes that may exist in the initial stable segment data. Specifically:

[0042] ;

[0043] ;

[0044] in, and These are the upper and lower thresholds of the statistical confidence interval, respectively. This represents the median of the initial stable segment data. is the interquartile range of the initial stable segment data, and c is a statistical coefficient, for example, with a value of 1.5.

[0045] In this optional embodiment, after traversing the initial stable segment data, all data points falling within the statistical confidence interval are retained as the refined dataset based on upper and lower thresholds. Finally, after removing potential outliers, the clean refined dataset is used to calculate the final benchmark load as the refined material benchmark. In this embodiment, the mean of the refined dataset is used as the refined material benchmark, and the standard deviation of the refined dataset is calculated and stored.

[0046] Thus, by using a self-filtering method based on robust statistics, a highly robust benchmark for refined materials can be extracted from initial data containing noise and shocks, providing a stable and reliable anchor point for the subsequent generation of dynamic thresholds.

[0047] S3. Construct a material cutting instability index based on the refined material benchmark and standard deviation to evaluate the inherent fluctuation characteristics of material cutting.

[0048] In an alternative embodiment, simply knowing the material's baseline load is insufficient, as the normal fluctuation ranges of different materials, such as stable aluminum alloys and volatile titanium alloys, vary significantly, directly impacting the threshold settings for passivation and tooth chipping. Therefore, the inherent fluctuation characteristics of material cutting can be assessed by constructing a material cutting instability index to reflect the relative fluctuation amplitude of the current signal and its relationship to the baseline load level. Specifically, the material cutting instability index satisfies the following relationship:

[0049]

[0050] in, This is an index of material cutting instability. To refine the standard deviation of the dataset, It serves as a benchmark for refining materials.

[0051] Thus, by constructing a composite material cutting instability index, we can not only know the average load of the material, but also assess the stability of its cutting process, comprehensively reflecting the cutting difficulty and inherent fluctuation characteristics of different materials.

[0052] S4. Based on the refined material benchmark, standard deviation and material cutting instability index, an adaptive threshold is generated. The adaptive threshold includes an adaptive passivation threshold and an adaptive spike threshold. Equipment fault diagnosis for forging production lines is achieved by comparing real-time current characteristics with the adaptive threshold.

[0053] In an optional embodiment, since passivation is a relative process, its threshold should be related to both the material's baseline load and inherent instability. Therefore, an adaptive passivation threshold can be constructed using a fixed proportional quantity related to the baseline, namely the refined material baseline and standard deviation, and a dynamic compensation margin related to instability, namely the material cutting instability index. The adaptive passivation threshold satisfies the following relationship:

[0054]

[0055] in, To adaptive passivation threshold, Based on passivation factor, As an instability passivation factor, for example, The value is 0.3. The value is set to 50. This implements a dual adaptive passivation threshold: the threshold is automatically lowered when cutting aluminum alloys to ensure sensitive detection of passivation; and the threshold is automatically raised when cutting titanium alloys to avoid false alarms due to high loads under healthy conditions.

[0056] In this optional embodiment, during the real-time monitoring phase, the average operating value of the current signal within a certain time window, such as a 1-second sliding window, is calculated. If the average operating value is greater than the adaptive passivation threshold and the duration exceeds the preset passivation confirmation time, such as 3 seconds, it can be determined that a saw blade passivation fault has occurred, and a corresponding alarm is triggered.

[0057] like Figure 2 The diagram shows a comparison of the diagnostic effects of adaptive passivation threshold and fixed threshold in an embodiment of the present invention. The X-axis represents time, and the Y-axis represents the average current. It can be seen that the adaptive threshold generated by this scheme for aluminum alloy is 32.7A, and for titanium alloy it is 135.1A, while the fixed threshold generated by the prior art is 50A. Healthy titanium alloy fluctuates around 90A, but is higher than the fixed threshold of 50A, thus leading to false alarms. Setting the threshold below the adaptive threshold of 135.1A for titanium alloy generated by this invention avoids such false alarms. Simultaneously, the passivated aluminum alloy slowly climbs from 20A, passing through the adaptive threshold of 32.7A, but never exceeding the fixed threshold of 50A, thus also avoiding the missed alarms caused by the existing fixed threshold.

[0058] Furthermore, for tooth breakage, since it is an instantaneous impact, the corresponding adaptive spike threshold should be determined based on the statistical characteristics of the initial stable period to determine the upper limit of normal fluctuations. Therefore, the adaptive spike threshold satisfies the following relationship:

[0059]

[0060] in, For adaptive peak thresholding, n is a multiple of the standard deviation, for example, a value of 6. This also achieves dynamic response of adaptive peak thresholding. For example, when cutting aluminum alloy, the threshold is automatically lowered to sensitively capture relatively small chipping peaks; when cutting titanium alloy, the threshold is automatically raised to automatically ignore its already large normal fluctuation range and only alarm for true abnormal impacts.

[0061] In this optional implementation, the system directly monitors the instantaneous value of the original high-frequency current signal during the real-time monitoring phase. When an instantaneous value exceeds the adaptive spike threshold, an abnormal spike is detected. Simultaneously, to improve anti-interference capabilities, if the number of detected abnormal spikes exceeds the preset spike count value within a preset spike counting time window (e.g., 5 seconds), e.g., 3 times, a saw blade tooth breakage fault is confirmed, and a high-level alarm is triggered.

[0062] like Figure 3 The diagram shown is a schematic of the adaptive spike threshold diagnosis effect of tooth breakage in an embodiment of the present invention. The X-axis represents time and the Y-axis represents instantaneous motor current. It can be seen that when tooth breakage occurs, the instantaneous motor current will generate a spike that far exceeds the normal fluctuation, thus passing through the adaptive spike threshold and being reliably detected, achieving timely alarm.

[0063] In this way, the sensitivity of the diagnosis can be automatically adjusted according to the benchmark and stability of the refined materials, thereby achieving high-precision and high-reliability diagnosis of passivation and tooth breakage faults in complex production scenarios with multiple materials and variable working conditions.

[0064] This invention also discloses an intelligent equipment fault diagnosis system for forging production lines, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent equipment fault diagnosis method for forging production lines according to the present invention is implemented.

[0065] 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 will not be described in detail here.

[0066] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

[0067] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. An apparatus failure intelligent diagnosis method for a forging production line, characterized by, The method comprises the following steps: Collecting a current signal of a main shaft motor of a sawing machine, and automatically identifying a start time and an end time of a cutting task according to the current signal, and extracting an effective cutting segment data sequence; Cutting initial stable segment data of the effective cutting segment data sequence, removing abnormal data points in the initial stable segment data by a self-filtering method based on robust statistics to obtain a refined data set, and calculating a mean value of the refined data set as a refined material reference and calculating a standard deviation of the refined data set; Constructing a material cutting instability index for evaluating inherent fluctuation characteristics of material cutting based on the refined material reference and the standard deviation; Generating an adaptive threshold based on the refined material reference, the standard deviation and the material cutting instability index, the adaptive threshold comprising an adaptive dulling threshold and an adaptive spike threshold, and realizing equipment fault diagnosis facing a forging production line by comparing real-time current characteristics with the adaptive threshold; The adaptive dulling threshold satisfies: for an adaptive passivation threshold, for the refined material reference, for a material cutting instability indicator, for a base passivation factor, for an instability passivation factor; The adaptive spike threshold satisfies: for an adaptive spike threshold, for the standard deviation of the refined dataset, n is the number of standard deviations.

2. The method according to claim 1, wherein The automatic identification of the start time and the end time of the cutting task according to the current signal comprises: Offline calibration of an idle current baseline and an idle current standard deviation when the sawing machine is at idle speed; When the real-time current continuously and continuously exceeds the sum of the idle current baseline and a preset multiple of the idle current standard deviation for a first preset time length, it is determined that the cutting starts; When the real-time current continuously and continuously is lower than the sum of the idle current baseline and a preset multiple of the idle current standard deviation for a second preset time length, it is determined that the cutting ends.

3. The method according to claim 1, wherein The removal of the abnormal data points in the initial stable segment data by the self-filtering method based on robust statistics comprises: Calculating a median and a quartile range of the initial stable segment data; Based on the median and the quartile range, constructing an upper threshold and a lower threshold of a statistical confidence interval; Traversing the initial stable segment data, based on the upper threshold and the lower threshold, retaining data points falling within the statistical confidence interval to form the refined data set.

4. The method of claim 1, wherein the method is characterized by, The material cutting instability index satisfies the relationship: wherein, is an index of material cutting instability, is a standard deviation of the refined data set, is the refined material benchmark.

5. The method of claim 1, wherein the method is characterized by: The initial stable segment data of the effective cutting segment is specifically: from the start time of the effective cutting segment data sequence, skipping a preset tool entering impact time length, and cutting data of a subsequent preset basic time length as the initial stable segment data.

6. The method of claim 1, wherein the method is characterized by: The realization of the equipment fault diagnosis facing the forging production line by comparing the real-time current characteristics with the adaptive threshold comprises: Calculating a running average of the current signal in a real-time monitoring stage; When the running average is greater than the adaptive dulling threshold and the duration exceeds a preset dulling confirmation time length, it is determined that a saw blade dulling fault occurs.

7. The method of claim 1, wherein the method is characterized by, The realization of the equipment fault diagnosis facing the forging production line by comparing the real-time current characteristics with the adaptive threshold further comprises: Monitoring an instantaneous value of the current signal in a real-time monitoring stage; When there is an instantaneous value greater than the adaptive spike threshold, it is determined that an abnormal spike is detected; When the number of abnormal spikes detected within a preset spike counting time window exceeds a preset spike counting value, it is confirmed that a saw blade tooth collapse fault occurs.

8. An apparatus failure intelligent diagnosis system for a forging production line, characterized by, The method comprises the following steps: A processor and a memory, the memory storing computer program instructions which, when executed by the processor, implement a method for intelligent diagnosis of equipment faults of a forging line according to any one of claims 1-7.

Images