Riveting connection quality abnormity diagnosis method and system

Abstract

The invention provides a riveting connection quality abnormity diagnosis method. The method comprises the following steps: S1, collecting data and generating a quality curve; s2, initializing a sliding window length L and an area threshold value T; s3, preprocessing the data; s4, sliding window scanning is carried out, and a current maximum window area Smax and a mass curve of the maximum window area Smax are calculated; s5, if Smax is larger than or equal to T, it is judged that abnormity exists, and the quality curve with the maximum contribution degree to the area is removed; s6, repeating the steps S4 and S5 until all abnormal quality curves are removed; and S7, outputting classification results: a normal quality curve set and an abnormal quality curve set. The riveting connection quality abnormity diagnosis system provided by the invention is used for realizing the method. According to the method, manual feature definition is not needed, classification is completed directly through the area threshold value and the contribution degree, the anomaly judgment basis is clear, and engineers can reproduce problems conveniently.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of riveting quality management, and particularly relates to a riveting quality abnormality diagnosis method and system. BACKGROUND

[0002] The SPR riveting process is mainly used for connecting metal plates, and is widely used in the automobile industry. The working principle is to directly press the rivet into the plate to be connected by hydraulic pressure or a servo motor, and the rivet is plastically deformed under the pressure to form a stable connection.

[0003] For the SPR riveting process, quality curve feature analysis is the main means to diagnose riveting quality abnormalities. Quality curve feature analysis needs to find the features and feature points of the quality curve. The current method relies on manual experience to select features (such as mean, variance, and inflection point) or compares with a standard curve. This method has the following problems:

[0004] 1. Feature selection depends on experience and is difficult to adapt to different working conditions;

[0005] 2. The standard curve needs to be frequently updated according to the process parameters, which consumes manpower. SUMMARY

[0006] In order to solve the technical problems in the background art, the present application provides a riveting quality abnormality diagnosis method and system.

[0007] The riveting quality abnormality diagnosis method provided by the present application comprises the following steps:

[0008] S1, collecting parameter data of the riveting production process and generating a corresponding quality curve according to the collected parameter data;

[0009] S2, initializing the sliding window length L and the area threshold T;

[0010] S3, pre-processing the data collected in step S1, which includes data cleaning and curve alignment;

[0011] S4, sliding window scanning, and calculating the current maximum window area S max and the quality curve surrounding the maximum window area S max ;

[0012] S5, abnormality determination and classification: if S max ≥ T, it is determined to be abnormal, and the quality curve with the largest contribution to the area is removed;

[0013] S6, repeating steps S4 and S5 until all abnormal quality curves are removed;

[0014] S7, outputting the classification results: a normal quality curve set and an abnormal quality curve set.

[0015] Preferably, in step S6, if the quality curve surrounding the maximum window area S max equals 2, and S max ≥ T, output a warning: "abnormal quality curve too much".

[0016] Preferably, when outputting the warning "abnormal quality curve too much", manually review or adjust the sliding window parameters.

[0017] Preferably, the quality curve elimination operation comprises:

[0018] From the quality curve surrounding the maximum area S max , eliminate each quality curve one by one and recalculate the window area; select the quality curve whose window area after elimination drops to the minimum as the abnormal quality curve with the greatest contribution degree, and eliminate it.

[0019] Preferably, the sliding window length L is 5%-10% of the total length of the quality curve.

[0020] Preferably, the quality curve alignment method adopts an interpolation method or a truncation method.

[0021] Preferably, the window area calculation is divided into the following two cases:

[0022] When the quality curve is a continuous curve, the following formula is used:

[0023]

[0024] φ(x): the value set of each quality curve at x;

[0025] When the quality curve is a discrete curve, the following formula is used:

[0026]

[0027] φ(k): the value set of the kth point of each quality curve.

[0028] The riveting quality abnormality diagnosis system proposed by the present application is used to realize the above method, comprising:

[0029] A data acquisition module: used to acquire parameters of the riveting process in real time and generate quality curves;

[0030] A preprocessing module: used for data cleaning and curve alignment;

[0031] A sliding window processing module: used to calculate the area of the sliding window and determine abnormalities;

[0032] A classification output module: used to output the classification results of normal quality curves and abnormal quality curves.

[0033] The present application reflects the difference between the abnormal mass curve and the normal mass curve by the size of the area through the sliding window method, and gradually removes the mass curve with the largest contribution to the current maximum window area S max The contribution of the mass curve is the most important one. The unsupervised classification is completed directly through the area threshold and the contribution, without manual definition of the characteristics, and the abnormal judgment is clear (area exceeding the standard + contribution analysis), which is convenient for engineers to reproduce the problem. In addition, by adjusting the area threshold T, the sensitivity requirement of different production scenes can be adapted. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 A flow chart of a rivet connection quality abnormality diagnosis method proposed by the present application. DETAILED DESCRIPTION

[0035] The rivet connection quality abnormality diagnosis method proposed by the present application comprises the following steps:

[0036] S1, collecting the pressure, displacement and other parameters of the riveting process in real time through a sensor and generating a mass curve.

[0037] S2, initializing the sliding window length L and the area threshold T, specifically: the sliding window length L is dynamically adjusted according to the total length of the mass curve, and generally takes a value range of 5%-10% of the total length of the mass curve; the area threshold T is obtained through historical data statistics or experimental calibration.

[0038] S3, pre-processing the collected data, including data cleaning and curve alignment, to remove missing values and noise data through data cleaning; and aligning the generated mass curve through interpolation method or truncation method to unify the length and coordinate reference of each mass curve and ensure the effectiveness of subsequent sliding window analysis.

[0039] S4, sliding window scanning the mass curve, calculating the current maximum window area S max (maximum quadrilateral area formed by the mass curve in the sliding window and the upper and lower boundaries) and the mass curve surrounding the maximum window area S max , specifically: the window area calculation is divided into the following two cases:

[0040] When the mass curve is a continuous curve, the following formula is used:

[0041]

[0042] φ(x): the value set of each mass curve at x;

[0043] When the mass curve is a discrete curve (each mass curve in the sliding window has N points), the following formula is used:

[0044]

[0045] φ(k): the value set of the kth point of each mass curve;

[0046] Record the maximum window area S max , the minimum window area S min , and the mass curve enclosing the maximum window area S max .

[0047] S5, Abnormality determination and classification: if S max ≥ T, determine as abnormal, and eliminate the mass curve with the largest contribution to the area;

[0048] S6, repeat steps S4 and S5 until all abnormal mass curves are eliminated;

[0049] S7, output the classification results: the set of normal mass curves and the set of abnormal mass curves, the set of normal mass curves being the remaining mass curves after eliminating abnormal mass curves, and the set of abnormal mass curves being all the eliminated abnormal mass curves. For example, input 10 mass curves, eliminate 2 abnormal curves with the largest contribution after sliding window scanning, and classify the remaining 8 as normal, then the output result is: 8 normal curves and 2 abnormal curves.

[0050] In step S6, if the mass curve enclosing the maximum window area S max is equal to 2, and S max ≥ T, output a warning: "too many abnormal mass curves", for manual review or adjustment of the sliding window parameters.

[0051] In further embodiments, the curve with the largest contribution to S max is eliminated as follows: from the mass curves enclosing the maximum area S max , eliminate each mass curve one by one and recalculate the area, select the mass curve whose area decreases the most after elimination as the abnormal mass curve with the largest contribution, and eliminate it. The specific steps are as follows:

[0052] Step 1: from the mass curves enclosing S max , eliminate each mass curve one by one, and recalculate the window area S new after elimination each time.

[0053] Step 2: select the mass curve that makes S max -S new the largest (i.e., S new is the smallest after eliminating the mass curve), as the abnormal curve to be eliminated.

[0054] The riveting quality abnormality diagnosis system is used for realizing the method, and comprises a data acquisition module, a preprocessing module, a sliding window processing module, and a classification output module.

[0055] The data acquisition module collects data of the riveting process in real time and generates a quality curve. The preprocessing module performs data cleaning and curve alignment on the data collected by the data acquisition module. The sliding window processing module calculates an area through a sliding window and determines an abnormality. The classification output module outputs classification results of normal quality curves and abnormal quality curves.

[0056] In a further embodiment, a dynamic parameter adjustment module is further included, which is used for adaptively optimizing the sliding window length L and the area threshold T according to a production condition.

[0057] As known from the above, the difference between the abnormal quality curve and the normal quality curve is reflected by the area size, and the quality curve with the greatest contribution degree is gradually removed to perform unsupervised classification, without manually defining features, and the classification is directly completed through the area threshold and the contribution degree, and the abnormality determination is clear (area exceeding the threshold + contribution degree analysis), which is convenient for engineers to reproduce problems; in addition, the sensitivity requirement of different production scenes can be adapted by adjusting the area threshold T. max As known from the above, the difference between the abnormal quality curve and the normal quality curve is reflected by the area size, and the quality curve with the greatest contribution degree is gradually removed to perform unsupervised classification, without manually defining features, and the classification is directly completed through the area threshold and the contribution degree, and the abnormality determination is clear (area exceeding the threshold + contribution degree analysis), which is convenient for engineers to reproduce problems; in addition, the sensitivity requirement of different production scenes can be adapted by adjusting the area threshold T.

[0058] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can make equivalent replacements or changes to the technical range disclosed in the present application according to the technical solution and the inventive concept of the present application, which should be covered in the protection scope of the present application.

Claims

1. A rivet connection quality abnormality diagnosis method characterized by comprising: The method comprises the following steps: S1, collecting parameter data of a riveting production process and generating corresponding quality curves according to the collected parameter data; S2, initializing a sliding window length L and an area threshold T; S3, pre-processing the data collected in step S1, wherein the pre-processing comprises data cleaning and curve alignment; S4, sliding window scanning and calculating the current maximum window area S max and the mass curve of the maximum window area S max ​ S5, Abnormality determination and classification: if S max ≥ T, it is determined as abnormal, and the mass curve with the largest contribution to the area is eliminated; S6, repeating steps S4 and S5 until all abnormal quality curves are removed; S7, outputting classification results: a normal quality curve set and an abnormal quality curve set.

2. The rivet connection quality abnormality diagnosis method according to claim 1, characterized by, In step S6, if the maximum window area S max equals 2, and S max ≥ T, output a warning "Too many abnormal quality curves".

3. The rivet connection quality abnormality diagnosis method according to claim 2, characterized by, When an alert "too many abnormal quality curves" is output, manual review or adjustment of the sliding window parameters is performed.

4. The rivet connection quality abnormality diagnosis method according to claim 1, characterized by The quality curve removal operation comprises: From the quality curves enclosing the largest area S max , remove each single quality curve and recalculate the window area; select the quality curve whose window area drops to the minimum after removal as the abnormal quality curve with the largest contribution degree, and remove it.

5. The rivet connection quality abnormality diagnosis method according to claim 1, characterized by, The sliding window length L is 5%-10% of the total length of the quality curve.

6. The rivet connection quality abnormality diagnosis method according to claim 1, characterized by The method of curve alignment adopts an interpolation method or a truncation method.

7. The rivet connection quality abnormality diagnosis method according to claim 1, characterized by, The window area calculation is divided into the following two cases: When the quality curve is a continuous curve, the following formula is used: φ(x): a value set of each quality curve at x; When the quality curve is a discrete curve, the following formula is used: φ(k): a value set of the kth point of each quality curve.

8. A rivet connection quality abnormality diagnosis system for implementing the method according to any one of claims 1 to 7, characterized by The method comprises the following steps: A data collection module for collecting parameters of a riveting process in real time and generating quality curves; A pre-processing module for data cleaning and curve alignment; A sliding window processing module for calculating the area of a sliding window and determining abnormalities; A classification output module for outputting classification results of normal quality curves and abnormal quality curves.

Images