A thread tightening quality judgment method based on curve similarity and clustering

Abstract

A method for judging thread tightening quality based on curve similarity and clustering, characterized in that it comprises the following steps: (1) using curve similarity to preprocess displacement-tightening torque curve thread tightening quality data; (2) using a binary k-means clustering algorithm combined with a Spark distributed processing engine to extract feature points from each classification data after preprocessing; (3) for each classification, connecting the extracted feature points to obtain a thread tightening quality characteristic curve of each classification; (4) for a displacement-tightening torque curve to be tested, clarifying the category of the curve classification to which the tested curve belongs; (5) based on the clear classification, performing real-time thread tightening quality data point judgment. The present invention judges thread tightening quality by generating a characteristic curve of thread tightening quality data, thereby achieving effective control of thread tightening quality. In combination with the Spark distributed processing engine, the ability to process thread tightening quality data is guaranteed.

Description

Technical Field

[0001] The invention relates to an assembly quality control technology, in particular to a thread tightening quality judgment method, specifically a thread tightening quality judgment method based on curve similarity and clustering. Background Art

[0002] The tightening gun can automatically collect displacement-tightening torque curve data, and the displacement-tightening torque curve data is used to reflect the quality of thread tightening. The changing trend of the curve can reflect the stage of thread tightening. The value of the tightening torque at the displacement on the curve can reflect the quality of thread tightening. Accurately judging the abnormality of the tightening torque value can accurately judge whether the thread tightening quality is abnormal, thereby achieving effective control of the thread tightening quality. The present invention uses the above-mentioned displacement-tightening torque quality data to propose a thread tightening quality judgment method based on curve similarity and clustering, so as to achieve the purpose of effectively controlling the thread tightening quality.

[0003] The thread tightening quality data is collected in real time, and the real-time collected quality data is massive. In the present invention, clustering the massive data is a prerequisite for quality judgment. Therefore, in order to meet the processing capacity of massive data, the Spark distributed computing engine is integrated into the process of quality judgment, and a thread tightening quality judgment method based on curve similarity and clustering is proposed, thereby ensuring the processing capacity of massive thread tightening data, ensuring data processing efficiency, meeting enterprise application needs, and having high application value. Summary of the invention

[0004] The purpose of the present invention is to provide a thread tightening quality judgment method based on curve similarity and clustering, which uses multiple sets of qualified displacement-tightening torque thread tightening quality data to accurately control the thread tightening quality, in order to solve the problem of large data processing volume of tightening guns.

[0005] The technical solution of the present invention is:

[0006] A method for judging thread tightening quality based on curve similarity and clustering is characterized in that it comprises the following steps: (1) using curve similarity to preprocess displacement-tightening torque curve thread tightening quality data to obtain curve classification; (2) using a binary k-means clustering algorithm combined with a Spark distributed processing engine to extract feature points from each classification data after preprocessing; (3) for each classification, connecting the extracted feature points to obtain a tightening torque quality characteristic curve of each classification; (4) for a displacement-tightening torque curve to be tested, clarifying the category of the curve classification to which the test curve belongs based on curve similarity; and (5) based on the clarified classification, performing real-time tightening torque quality data point judgment.

[0007] Furthermore, in the step (1), the displacement-tightening torque curve is still classified based on the distance in the vertical direction of the displacement-tightening torque curve to obtain the curves of each classification. The thread tightening curve is denoted as O = (o1, o2, ..., o m ), n is the number of displacement-tightening torque data points collected during the thread tightening process, o k is the kth data point collected during the thread tightening process, and o k =(x k ,y k ), x is the displacement value of point o, y is the tightening torque value of point o, k=1,2,…,m. Then O=((x1,y1),(x2,y2),…,(x m ,y m )). Then the thread tightening curve O u With O v The distance formula is as follows:

[0008]

[0009] Let the curve class Q = {O1, O2, ..., O n}, where O i =((x i1 ,y i1 ),(x i2 ,y i2 ),…,(x im ,y im )). The similarity accuracy calculation formula of thread tightening curve type Q is as follows:

[0010]

[0011] Furthermore, in step (2), Spark is used to perform a parallel binary k-means clustering operation on the obtained curve of each classification to obtain the characteristic data points of each classification.

[0012] Furthermore, in step (3), the characteristic data points of each classification are connected to obtain a characteristic curve corresponding to each classification. t =(o t1 ,o t2 ,…,o tm ).

[0013] Furthermore, in step (4), the distance between the displacement-tightening torque curve in the vertical direction is used as a judgment basis to calculate the distance between the curve to be tested and each classification characteristic curve, and determine the category of the curve classification to which the curve to be tested belongs. r =(o r1 ,o r2 ,…,o rm), o = (x, y), for the characteristic curve O of any curve class t =(o t1 ,o t2 ,…,o tm ), the distance calculation formula from the curve to be measured to the characteristic curve is:

[0014]

[0015] Then L rt The curve class corresponding to the minimum value is determined as the curve class to which the curve to be tested belongs.

[0016] Furthermore, in step (5), the data points on the curve to be tested are compared with the characteristic curve, and the change rate Δ(c) is calculated. The change rate is used to determine whether the data points on the curve to be tested are abnormal. f =(x f ,y f ), O t and O r The corresponding tightening torque values ​​are y tf and rf , then the thread tightening curve to be tested is O r Relative to the characteristic curve O t The change rate calculation formula is as follows. The change rate is used as the degree of abnormal thread tightening, which can accurately judge the thread tightening quality.

[0017] Δ(c)=(y rc -y tc ) / y tc ×100%.

[0018] The beneficial effects of the present invention are:

[0019] 1. By using curve similarity to pre-process the classification of thread tightening quality, the classified curve is more in line with the actual application situation and provides a better data basis for clustering; 2. The characteristic curve obtained by the clustering algorithm can clearly deduce the degree of quality abnormality and realize precise control of thread tightening quality; 3. Using Spark for parallel clustering operation can meet the efficiency of processing massive thread tightening quality data. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of the thread tightening quality judgment method of the present invention.

[0021] Figure 2 It is a diagram of multiple groups of displacement-tightening torque data points of the present invention.

[0022] Figure 3 This is a diagram of the thread tightening feature curve extraction process of binary k-means in the present invention.

[0023] Figure 4 Schematic diagram of parallelized binary k-means operation in the present invention.

[0024] Figure 5 These are four types of characteristic curves obtained using the method of the present invention.

[0025] Figure 6 This is a quality judgment comparison chart of the present invention and the k-means algorithm using only binary division and k-means. DETAILED DESCRIPTION

[0026] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0027] The thread tightening quality judgment method of the present invention is used to realize the generation of a characteristic curve based on the qualified displacement-tightening torque thread tightening quality data of an existing tightening gun. Even for a tightening gun, there are differences in the curve trend. Based on the similarity of the curve, it is first pre-processed to obtain various classifications of the curve. Then, a clustering algorithm is used to obtain the characteristic curve of the classified pre-processed curve, which is used as the basis for judging the quality of a tightening gun.

[0028] Figure 1 The flow chart for generating the envelope of the present invention specifically includes the following steps:

[0029] Step 1: Classify the displacement-tightening torque curves based on the vertical distance of the displacement-tightening torque curves to obtain curves of each classification. Figure 2 shown.

[0030] Step 2: Use Spark to perform parallel binary k-means clustering operations on each classification curve to obtain the characteristic data points of each classification. Figure 3 The figure shows the process of extracting thread tightening feature curves using binary k-means. Figure 4 The figure shows a schematic diagram of parallel binary k-means operation using Spark.

[0031] Step 3: The characteristic data points of each classification can be connected by interpolation to obtain the characteristic curve corresponding to each classification. t =(o t1 ,o t2 ,…,o tm ).like Figure 5 Shown are 4 types of characteristic curve diagrams.

[0032] Step 4: Based on the vertical distance of the displacement-tightening torque curve, the distance between the curve to be tested and each classification characteristic curve is calculated to determine the category of the curve classification to which the curve to be tested belongs. r , any characteristic curve O t , calculate the distance L between the two curves rt , get the curve class to which the curve to be tested belongs.

[0033] Feature 5: Compare the data points on the curve to be tested with the characteristic curve, calculate the rate of change, and use the rate of change to determine whether the data points on the curve to be tested are abnormal. Figure 6 The figure shows some abnormal data in the displacement-tightening torque curve to be tested. After manual judgment, the displacements 101, 102, 103, and 104 are abnormal values, and 99, 100, 105, and 106 are normal values. By calculating the change rate Δ(c) by the method of the present invention, it can be effectively judged that the tightening torque is abnormal, and the abnormalities are accurately judged to be the four data points 101-104.

[0034] In order to further verify the accuracy of the thread tightening quality judgment method based on curve similarity and clustering, the method of the present invention is compared with the method using only binary k-means clustering processing and the k-means clustering algorithm, and the following results are obtained: Figure 6 The comparison results shown clearly show that the three methods can identify the abnormalities in the four data points 101-104. However, the method of the present invention is more accurate in judging the four normal data points 99, 100, 105, and 106. Therefore, it can be concluded that the method of the present invention can not only accurately identify quality abnormalities, but also has a small judgment deviation on normal data.

[0035] The present invention mainly judges the displacement-tightening torque quality data, thereby judging the tightening torque quality, and provides a feasible way for timely control of abnormal tightening torque quality, so as to further improve the manufacturing quality of assembly enterprises.

[0036] The present invention does not involve parts that are the same as the prior art or can be implemented by using the prior art.

Claims

1. A thread tightening quality judgment method based on curve similarity and clustering, The method is characterized in that: firstly, the displacement-tightening torque curve thread tightening quality data is preprocessed using curve similarity to obtain curve classification; then, the binary k-means clustering algorithm combined with the Spark distributed processing engine is used to extract feature points from each classification data after preprocessing; thirdly, for each classification, the extracted feature points are connected to obtain the thread tightening quality characteristic curve of each classification; Fourth, for the displacement-tightening torque curve to be tested, the category of the curve classification to which the tested curve belongs is determined based on the curve similarity; finally, based on the clear classification, real-time thread tightening quality data point judgment is performed.

2. The thread tightening quality judgment method based on curve similarity and clustering according to claim 1 is characterized in that It includes the following steps: Step 1: Use curve similarity to preprocess the displacement-tightening torque curve thread tightening quality data to obtain curve classification Q and two thread tightening curves O u With O v The similarity is expressed as the distance L between the curves in the vertical direction. uv Based on; Step 2: Use the binary k-means clustering algorithm combined with the Spark distributed processing engine to extract feature points from each classified data after preprocessing to obtain the feature data points of each curve classification; Step 3: For each category, connect the extracted feature points to obtain the thread tightening quality characteristic curve O of each category. t =(o t1 ,o t2 ,…,o tm ); Step 4: For the displacement-tightening torque curve to be tested, the category of the curve classification to which the curve to be tested belongs is determined based on the curve similarity, that is, by calculating the curve O r =(o r1 ,o r2 ,…,o rm ), o = (x, y) to any characteristic curve O t =(o t1 ,o t2 ,…,o tm ) rt , and select L rt The classification corresponding to the minimum value is the classification of the curve to be tested; Step 5: Based on the clear classification, make real-time judgment on the thread tightening quality data points; calculate the rate of change Δ(c) from the data points on the curve to be tested to the characteristic curve, and judge whether the curve to be tested is abnormal based on the rate of change.

3. The thread tightening quality judgment method based on curve similarity and clustering according to claim 2 is characterized in that: The two thread tightening curves O described in step 1 u With O v The vertical distance 4. The thread tightening quality judgment method based on curve similarity and clustering according to claim 2 is characterized in that: In step 4, the curve category to which the curve to be measured belongs is determined by the distance L between the two. rt For judgment basis, 5. The thread tightening quality judgment method based on curve similarity and clustering according to claim 2 is characterized in that: The rate of change of the data points on the curve to be measured and the characteristic curve Δ(c)=(y rc -y tc ) / y tc ×100%.

Images