A method for non-destructive evaluation of the quality of ultrasonic welding at the end of a wire

Abstract

The application provides a method for nondestructively evaluating the quality of ultrasonic welding of butt joints, which comprises the following steps: 1) collecting echo signals of each qualified welding butt joint and echo signals of each unqualified welding butt joint; 2) calculating waveform factors corresponding to the echo signals of each qualified welding butt joint and waveform factors corresponding to the echo signals of each unqualified welding butt joint; 3) taking the average value of the median values of the waveform factors corresponding to the qualified welding butt joints as CPOK1, and taking the average value of the median values of the waveform factors corresponding to the unqualified welding butt joints as CPNG1; 4) obtaining the median value of the waveform factor distribution diagram of the welding butt joint to be detected as CP1, and comparing CPOK1, CPNG1 and CP1 to determine whether the welding quality of the welding butt joint to be detected is qualified. The method has the advantages of nondestructiveness and can quickly identify whether the welding quality of the ultrasonic welding butt joint is qualified.

Description

Technical Field

[0001] This invention belongs to the field of welding technology and relates to welding quality evaluation technology, specifically a method for non-destructive evaluation of the quality of ultrasonic welding of wire ends. Background Technology

[0002] A wire connection refers to the joint where two or more wire harnesses are connected. Currently, the quality of ultrasonic welding at wire connections directly determines the safe operation and reliable power supply of electric vehicle power systems; therefore, evaluating the quality of ultrasonic welding is crucial. Commonly used methods for evaluating ultrasonic welding quality include pull-out force testing and resistance testing. Pull-out force testing uses the pull-out force of a tear-off force tester as a mechanical performance indicator of the wire connection joint. This testing technique is destructive and can only be performed through random sampling, making 100% coverage impossible. Resistance testing evaluates the quality of ultrasonic welding by measuring or calculating the resistance value. For example, patent application CN201010600658.8 discloses a method for checking the welding quality of stator winding joints in large generators, including the following steps: 1) A constant 50A DC current is output from a 50A transformer DC resistance tester and connected to both ends of the stator winding joint; 2) The current is measured using a 0.2A shunt and a DC voltmeter to monitor the DC current value; 3) The voltage across the joint resistance is measured using a high-precision digital multimeter and converted into a resistance value. Using this method, the measurement accuracy of the stator winding joint resistance can reach 0.2μΩ. This patent document uses the voltage value measured by a high-precision digital multimeter and the current value measured by a 0.2A shunt and a DC voltmeter to calculate the resistance of the stator winding joint in a large generator, and uses this resistance value to evaluate the welding quality of the stator winding joint. However, this method has a complex measurement process (requiring two measurements using a 0.2A shunt and a DC voltmeter, and a high-precision digital multimeter), and the detection speed is slow. Summary of the Invention

[0003] In view of the above-mentioned technical problems of existing methods for evaluating welding quality, such as high destructiveness, complex detection process and slow detection speed, this invention proposes a non-destructive evaluation method for ultrasonic welding quality of wire ends.

[0004] This invention utilizes an ultrasonic phased array probe to acquire the echo signal from the weld joint, calculates the signal characteristic quantity—waveform factor—of the echo signal, and evaluates the ultrasonic welding quality of the weld joint based on the median value of the waveform factor corresponding to the most influential signal characteristic quantity (waveform factor). The method of this invention is non-destructive, the evaluation process is simple, and it can quickly identify whether the weld joint is qualified.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A method for non-destructively evaluating ultrasonic welding quality of wire ends, comprising the following steps:

[0007] 1) Taking a plurality of qualified welded butt ends and a plurality of unqualified welded butt ends, and collecting echo signals of each qualified welded butt end and echo signals of each unqualified welded butt end;

[0008] 2) Calculating a waveform factor corresponding to the echo signal of each qualified welded butt end collected in step 1) and a waveform factor corresponding to the echo signal of each unqualified welded butt end;

[0009] 3) Drawing a waveform factor distribution diagram corresponding to each qualified welded butt end and a waveform factor distribution diagram corresponding to each unqualified welded butt end;

[0010] Determining a median waveform factor of each qualified welded butt end according to the waveform factor distribution diagram corresponding to each qualified welded butt end; taking an average value of the median waveform factors of the plurality of qualified welded butt ends, which is recorded as CPOK1;

[0011] Determining a median waveform factor of each unqualified welded butt end according to the waveform factor distribution diagram corresponding to each unqualified welded butt end; taking an average value of the median waveform factors of the plurality of unqualified welded butt ends, which is recorded as CPNG1;

[0012] 4) Repeating steps 1) to 3) to obtain a median value of the waveform factor distribution diagram of the welded butt end to be tested, which is recorded as CP1. If CP1≥CPOK1, the welding quality of the welded butt end to be tested is qualified; if CP1<CPNG1, the welding quality of the welded butt end to be tested is unqualified; if CPOK1>CP1≥CPNG1, the welded butt end to be tested is a piece with undetermined quality.

[0013] Further defined, said step 2) further comprises: calculating a clearance factor corresponding to the echo signal of each qualified welded butt end collected in step 1) and a clearance factor corresponding to the echo signal of each unqualified welded butt end, and calculating kurtosis corresponding to the echo signal of each qualified welded butt end collected in step 1) and kurtosis corresponding to the echo signal of each unqualified welded butt end.

[0014] Further defined, the steps of said method for non-destructively evaluating ultrasonic welding quality of wire ends further comprise:

[0015] 5) Perform range standardization on the waveform factor, the waveform factor, the margin factor, the margin factor, the kurtosis, and the kurtosis of the echo signals of each qualified and unqualified welded joints to obtain standardized qualified waveform factors, standardized unqualified waveform factors, standardized qualified margin factors, standardized unqualified margin factors, standardized qualified kurtosis, and standardized unqualified kurtosis.

[0016] Furthermore, the steps of the method for non-destructive evaluation of the quality of ultrasonic welding at the wire end also include:

[0017] 6) Using an ensemble learning algorithm, weights are assigned to the standardized qualified waveform factor, standardized qualified margin factor, and standardized qualified kurtosis, respectively. The evaluation factor CPOK for each qualified weld joint is obtained through weighted calculation. The qualified evaluation factor distribution map of each qualified weld joint is plotted using the evaluation factor CPOK of each qualified weld joint. The median of the qualified evaluation factor for each qualified weld joint is determined based on the qualified evaluation factor distribution map of each qualified weld joint.

[0018] An ensemble learning algorithm is used to assign weights to the standardized non-conforming waveform factor, standardized non-conforming margin factor, and standardized non-conforming kurtosis, respectively. The evaluation factor CPNG of each non-conforming weld joint is obtained through weighted calculation. The non-conforming evaluation factor distribution map of each non-conforming weld joint is plotted using the evaluation factor CPNG of each non-conforming weld joint. The median of the non-conforming evaluation factor of each non-conforming weld joint is determined based on the non-conforming evaluation factor distribution map of each non-conforming weld joint.

[0019] The average of the median values ​​of the qualified evaluation factors of multiple qualified welded joints is denoted as CPOK2; the average of the median values ​​of the unqualified evaluation factors of multiple unqualified welded joints is denoted as CPNG2; let CPHG = (CPOK2 + CPNG2) / 2, where CPHG is the judgment benchmark value;

[0020] 7) Repeat steps 5) and 6) to calculate the evaluation factor of the quality-undetermined part in step 4), denoted as CP2. If CP2 ≥ CPHG, the quality-undetermined part is determined as a qualified weld joint; otherwise, the quality-undetermined part is a substandard weld joint.

[0021] Further specifying, the formula for the weighted calculation in step 6) is:

[0022] Cp = a1*wf + a2*mf + a3*kf

[0023] In the formula, Cp is the evaluation factor, which is dimensionless; wf is the standardized waveform factor, which is dimensionless; mf is the standardized margin factor, which is dimensionless; kf is the standardized kurtosis, which is dimensionless; a1 is the weight of the waveform factor, which is dimensionless; a2 is the weight of the margin factor, which is dimensionless; and a3 is the weight of the kurtosis, which is dimensionless.

[0024] Further specifying, the method for determining the waveform factor, margin factor, and kurtosis in step 2) is as follows: calculate the signal feature quantities corresponding to the echo signals of each qualified welded joint and each unqualified welded joint collected in step 1), respectively, and use an ensemble learning algorithm to sort the signal feature quantities of each qualified welded joint and each unqualified welded joint, respectively, and obtain the top three signal feature quantities in terms of importance, which are the waveform factor, margin factor, and kurtosis.

[0025] Further specifying, step 1) specifically involves: using an ultrasonic phased array probe and a full matrix acquisition method to acquire the echo signal of each qualified weld joint and the echo signal of each unqualified weld joint.

[0026] Further specifying, the ultrasonic phased array probe is a 32-element, 5MHz ultrasonic phased array probe.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] 1. This invention provides a non-destructive evaluation method for ultrasonic welding quality of welded ends, comprising: acquiring echo signals from the welded ends using an ultrasonic phased array probe; selecting the most influential signal characteristic—waveform factor; plotting a waveform factor distribution map based on the waveform factor, then obtaining the median waveform factor; using the median waveform factor as a judgment benchmark; comparing this benchmark with the median waveform factor corresponding to the waveform factor distribution map of the welded end under test to determine whether the welding quality of the welded end under test is qualified. This invention evaluates the welding quality of the welded ends under test based on a highly influential signal characteristic; the method of this invention is simple, scientific, reasonable, low-cost, and non-destructive, and can quickly identify whether the welding quality of ultrasonically welded ends is qualified.

[0029] 2. This invention employs echo signal processing, ensemble learning algorithm, and ultrasonic phased array probe detection and acquisition technology to achieve rapid detection of welding quality at the wire end of ultrasonic welding. Compared with traditional machine learning methods, this invention is more intuitive and accurate.

[0030] 3. The present invention provides a method for non-destructive evaluation of the quality of ultrasonic welding of wire ends. It is not only applicable to the quality inspection of ultrasonic welding of wire ends in automobiles, but can also be applied to the welding quality inspection in other aspects, and has a wide range of applications.

[0031] 4. This invention collects the three most important signal features, namely waveform factor, margin factor and kurtosis, and assigns weights to waveform factor, margin factor and kurtosis respectively, and then obtains the evaluation factor Cp through weighted calculation; using the three most important signal features can improve the comprehensiveness of the influence of ultrasonic welding quality, and make the evaluation of ultrasonic welding quality more reasonable and sufficient. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the interface for the full matrix acquisition method;

[0033] Figure 2 A schematic diagram for filtering valid signals;

[0034] Figure 3 A schematic diagram illustrating the sorting of signal features using an ensemble learning algorithm;

[0035] Figure 4 This is the waveform factor distribution diagram in step 3);

[0036] Figure 5 This is a distribution chart of the margin factor in step 5);

[0037] Figure 6 This is the kurtosis distribution map in step 5);

[0038] Figure 7 This is a distribution diagram of the waveform factor after standardization. Detailed Implementation

[0039] The technical solution of the present invention will be further explained and described below with reference to the accompanying drawings and embodiments, but the present invention is not limited to the embodiments described below.

[0040] Example 1

[0041] This embodiment provides a method for non-destructive evaluation of the quality of ultrasonic welding at wire ends, comprising the following steps:

[0042] 1) Take multiple qualified welded joints and multiple unqualified welded joints, and use the multiple qualified welded joints and multiple unqualified welded joints as samples to collect the echo signal of each qualified welded joint and the echo signal of each unqualified welded joint.

[0043] 2) Calculate the waveform factor corresponding to the echo signal of each qualified welded joint and the waveform factor corresponding to the echo signal of each unqualified welded joint collected in step 1).

[0044] 3) drawing a waveform factor distribution diagram corresponding to each qualified welded butt end and a waveform factor distribution diagram corresponding to each unqualified welded butt end; wherein the waveform factor distribution diagram corresponding to each qualified welded butt end is drawn according to the waveform factor at each detection point corresponding to the echo signal of each qualified welded butt end, and the waveform factor distribution diagram corresponding to each unqualified welded butt end is drawn according to the waveform factor at each detection point corresponding to the echo signal of each unqualified welded butt end;

[0045] determining the median waveform factor corresponding to each qualified welded butt end according to the waveform factor distribution diagram corresponding to each qualified welded butt end, wherein the median waveform factor corresponding to each qualified welded butt end refers to the waveform factor corresponding to the median value on the X-axis in the waveform factor distribution diagram corresponding to each qualified welded butt end; taking the average value of the median waveform factors corresponding to a plurality of qualified welded butt ends, which is recorded as CPOK1;

[0046] determining the median waveform factor corresponding to each unqualified welded butt end according to the waveform factor distribution diagram corresponding to each unqualified welded butt end, wherein the median waveform factor corresponding to each unqualified welded butt end refers to the waveform factor corresponding to the median value on the X-axis in the waveform factor distribution diagram corresponding to each unqualified welded butt end; taking the average value of the median waveform factors corresponding to a plurality of unqualified welded butt ends, which is recorded as CPNG1;

[0047] 4) repeating steps 1) to 3) to obtain the median value of the waveform factor distribution diagram of the welded butt end to be tested, which is recorded as CP1. If CP1≥CPOK1, the welding quality of the welded butt end to be tested is qualified; if CP1<CPNG1, the welding quality of the welded butt end to be tested is unqualified; if CPOK1>CP1≥CPNG1, the welded butt end to be tested is a piece with undetermined quality.

[0048] Specifically, step 4) is: collecting the echo signal of the welded butt end to be tested; calculating the waveform factor corresponding to the echo signal of the welded butt end to be tested, and drawing the waveform factor distribution diagram of the welded butt end to be tested according to the waveform factor corresponding to the echo signal of the welded butt end to be tested; determining the median waveform factor of the welded butt end to be tested according to the waveform factor distribution diagram of the welded butt end to be tested, which is recorded as CP1. If CP1≥CPOK1, the welding quality of the welded butt end to be tested is qualified; if CP1<CPNG1, the welding quality of the welded butt end to be tested is unqualified; if CPOK1>CP1≥CPNG1, the welded butt end to be tested is a piece with undetermined quality.

[0049] with 75 mm 2 taking the ultrasonic evaluation on the welding quality of the butt end formed by a copper wire bundle and a copper terminal as an example, the specific steps of the non-destructive evaluation method for ultrasonic welding quality of a wire end in this embodiment are as follows:

[0050] 1) Take 50 qualified welded joints and 50 unqualified welded joints, and use a 32-element 5MHz ultrasonic phased array probe and full matrix acquisition method (see...). Figure 1 Collect the echo signal of each qualified weld joint and the echo signal of each unqualified weld joint; see [link / reference]. Figure 2 The excitation and reception signals of the first 6 and last 6 ultrasonic phased array probes corresponding to each qualified weld joint and each unqualified weld joint are deleted to filter out the valid signals.

[0051] Specifically, the number of qualified and unqualified welded joints can be arbitrarily chosen, but preferably, the ratio is 1:1; for example, 50 qualified welded joints and 50 unqualified welded joints, or 100 qualified welded joints and 100 unqualified welded joints. The waveform factor, ranked first in importance, is used as the initial judgment criterion. (See [reference needed]). Figure 4 ;exist Figure 4 In the diagram, the horizontal axis represents the number of the welded joints (including qualified and unqualified welded joints). OK1 represents the waveform factor characteristic value corresponding to welded joint number 1 among qualified welded joints; OK2 represents the waveform factor characteristic value corresponding to welded joint number 2 among qualified welded joints; OK3, OK4, OK5, and OK6 have the same meaning as OK1 and OK2; NG1 represents the waveform factor characteristic value corresponding to welded joint number 1 among unqualified welded joints; NG2 represents the waveform factor characteristic value corresponding to welded joint number 2 among unqualified welded joints; NG3, NG4, NG5, and NG6 have the same meaning as NG1 and NG2. The vertical axis represents the waveform factor corresponding to the 400 echo signals collected for each welded joint (including qualified and unqualified welded joints).

[0052] 2) Calculate the waveform factor corresponding to the echo signal of each qualified welded joint and the waveform factor corresponding to the echo signal of each unqualified welded joint collected in step 1).

[0053] 3) Based on the waveform factor corresponding to the echo signal of each qualified welded joint, draw the waveform factor distribution diagram corresponding to each qualified welded joint; based on the waveform factor distribution diagram corresponding to each qualified welded joint, draw the waveform factor distribution diagram corresponding to each unqualified welded joint.

[0054] Determine the median waveform factor for each qualified welded joint based on the waveform factor distribution diagram for each qualified welded joint; take the average of the median waveform factors for multiple qualified welded joints and denote it as CPOK1.

[0055] Determine the median waveform factor corresponding to each unqualified welded butt end according to the waveform factor distribution diagram corresponding to each unqualified welded butt end; take the average value of the median waveform factors corresponding to a plurality of unqualified welded butt ends, and record it as CPNG1;

[0056] 4) Repeat steps 1) to 3) to obtain the median value of the waveform factor distribution diagram of the welded butt end to be tested, which is recorded as CP1. If CP1≥CPOK1, the welding quality of the welded butt end to be tested is qualified; if CP1<CPNG1, the welding quality of the welded butt end to be tested is unqualified; if CPOK1>CP1≥CPNG1, the welded butt end to be tested is a piece with undetermined quality.

[0057] Example 2

[0058] The present embodiment provides a non-destructive evaluation method for ultrasonic welding quality of wire ends, based on Example 1, the method further comprises step 5), step 6) and step 7);

[0059] Step 2) in Example 1 further comprises calculating the margin factor corresponding to the echo signal of each qualified welded butt end and the margin factor corresponding to the echo signal of each unqualified welded butt end collected in step 1), and calculating the kurtosis corresponding to the echo signal of each qualified welded butt end and the kurtosis corresponding to the echo signal of each unqualified welded butt end collected in step 1).

[0060] Step 5) is: subjecting the waveform factor corresponding to the echo signal of each qualified welded butt end, the waveform factor corresponding to the echo signal of each unqualified welded butt end, the margin factor corresponding to the echo signal of each qualified welded butt end, the margin factor corresponding to the echo signal of each unqualified welded butt end, the kurtosis corresponding to the echo signal of each qualified welded butt end and the kurtosis corresponding to the echo signal of each unqualified welded butt end to range standardization processing, to obtain standardized qualified waveform factor, standardized unqualified waveform factor, standardized qualified margin factor, standardized unqualified margin factor, standardized qualified kurtosis and standardized unqualified kurtosis.

[0061] Step 6) is:

[0062] Assign weights to the standardized qualified waveform factor, the standardized qualified margin factor and the standardized qualified kurtosis respectively by using an ensemble learning algorithm, and obtain the evaluation factor CPOK of each qualified welded butt end through weighted operation; draw a qualified evaluation factor distribution diagram of each qualified welded butt end by using the evaluation factor CPOK of each qualified welded butt end, and determine the median qualified evaluation factor of each qualified welded butt end according to the qualified evaluation factor distribution diagram of each qualified welded butt end; wherein, the median qualified evaluation factor of each qualified welded butt end refers to the evaluation factor corresponding to the middle value of the X-axis in the qualified evaluation factor distribution diagram of each qualified welded butt end;

[0063] An ensemble learning algorithm is used to assign weights to the standardized non-conforming waveform factor, standardized non-conforming margin factor, and standardized non-conforming kurtosis, respectively. Weighted calculations are then performed to obtain the evaluation factor CPNG for each non-conforming weld joint. A distribution map of the non-conforming evaluation factors for each non-conforming weld joint is plotted using the CPNG. The median of the non-conforming evaluation factors for each non-conforming weld joint is determined based on this distribution map. The median of the non-conforming evaluation factors for each non-conforming weld joint refers to the distribution map of the non-conforming evaluation factors for each non-conforming weld joint within the distribution map of the non-conforming evaluation factors for each non-conforming weld joint.

[0064] The average of the median values ​​of the qualified evaluation factors of multiple qualified welded joints is denoted as CPOK2; the average of the median values ​​of the unqualified evaluation factors of multiple unqualified welded joints is denoted as CPNG2; let CPHG = (CPOK2 + CPNG2) / 2, where CPHG is the judgment benchmark value;

[0065] Step 7) is to repeat steps 5) and 6) and calculate the evaluation factor of the quality-undetermined part in step 4), denoted as CP2. If CP2 ≥ CPHG, the quality-undetermined part is determined as a qualified welded joint; otherwise, the quality-undetermined part is a substandard welded joint.

[0066] Specifically, step 7 is as follows: Collect the waveform factor, margin factor, and kurtosis corresponding to the echo signal of the component to be evaluated. Perform range standardization on the waveform factor, margin factor, and kurtosis corresponding to the echo signal of the component to be evaluated to obtain the standardized waveform factor, standardized margin factor, and standardized kurtosis corresponding to the component to be evaluated. Substitute the standardized waveform factor, standardized margin factor, and standardized kurtosis into the weighted calculation formula to obtain the evaluation factor of the component to be evaluated, denoted as CP2. If CP2 ≥ CPHG, the component to be evaluated is defined as a qualified weld joint; otherwise, the component to be evaluated is a substandard weld joint.

[0067] In this embodiment, the formula for weighted calculation in step 6) is:

[0068] Cp = a1*wf + a2*mf + a3*kf

[0069] In the formula, Cp is the evaluation factor, which is dimensionless; wf is the standardized waveform factor, which is dimensionless; mf is the standardized margin factor, which is dimensionless; kf is the standardized kurtosis, which is dimensionless; a1 is the weight of the waveform factor, which is dimensionless; a2 is the weight of the margin factor, which is dimensionless; and a3 is the weight of the kurtosis, which is dimensionless.

[0070] See Figure 3 In the above embodiments 1 and 2, the method for determining the waveform factor, margin factor and kurtosis in step 2) is as follows: calculate the signal feature quantity corresponding to the echo signal of each qualified welded joint and the echo signal of each unqualified welded joint collected in step 1), and use the ensemble learning algorithm to sort the signal feature quantity of each qualified welded joint and the signal feature quantity of each unqualified welded joint. This sorting refers to comprehensively sorting the signal feature quantity of each qualified welded joint and the signal feature quantity of each unqualified welded joint together, and obtaining the top three signal feature quantities in terms of importance, which are waveform factor, margin factor and kurtosis.

[0071] In Embodiments 1 and 2 above, step 1) specifically involves: acquiring the echo signal of each qualified weld joint and the echo signal of each unqualified weld joint using an ultrasonic phased array probe and a full matrix acquisition method. The ultrasonic phased array probe is a 32-element, 5MHz ultrasonic phased array probe.

[0072] Specifically, continuing from Example 1, with 75mm 2 Taking the ultrasonic evaluation of the welding quality between copper wire harnesses and copper terminals as an example, the steps corresponding to a non-destructive evaluation method for ultrasonic welding quality of wire terminals in this embodiment are explained:

[0073] In this embodiment, see Figure 5 and Figure 6 The waveform factor, margin factor, kurtosis, and kurtosis corresponding to the echo signals of each qualified and unqualified welded joint are all processed using the range normalization method to obtain standardized qualified waveform factors, standardized unqualified waveform factors, standardized qualified margin factors, standardized unqualified margin factors, standardized qualified kurtosis, and standardized unqualified kurtosis. Figure 5In the diagram, the horizontal axis represents the number of the welded joints (including qualified and unqualified welded joints). OK1 represents the margin factor characteristic value corresponding to welded joint number 1 among qualified welded joints; OK2 represents the margin factor characteristic value corresponding to welded joint number 2 among qualified welded joints; OK3, OK4, OK5, and OK6 have the same meaning as OK1 and OK2; NG1 represents the margin factor characteristic value corresponding to welded joint number 1 among unqualified welded joints; NG2 represents the margin factor characteristic value corresponding to welded joint number 2 among unqualified welded joints; NG3, NG4, NG5, and NG6 have the same meaning as NG1 and NG2. The vertical axis represents the margin factor corresponding to the 400 echo signals collected for each welded joint (including qualified and unqualified welded joints). Figure 6 In the diagram, the horizontal axis represents the number of the welded joints (including qualified and unqualified welded joints). OK1 represents the kurtosis characteristic value corresponding to welded joint number 1 among qualified welded joints; OK2 represents the kurtosis characteristic value corresponding to welded joint number 2 among qualified welded joints; OK3, OK4, OK5, and OK6 have the same meaning as OK1 and OK2; NG1 represents the kurtosis characteristic value corresponding to welded joint number 1 among unqualified welded joints; NG2 represents the kurtosis characteristic value corresponding to welded joint number 2 among unqualified welded joints; NG3, NG4, NG5, and NG6 have the same meaning as NG1 and NG2. The vertical axis represents the kurtosis corresponding to the 400 echo signals collected for each welded joint (including qualified and unqualified welded joints).

[0074] 6) Using an ensemble learning algorithm, weights are assigned to the standardized qualified waveform factor, standardized qualified margin factor, and standardized qualified kurtosis, respectively, to determine their respective weights, where a1 = 0.5, a2 = 0.3, and a3 = 0.2. Substituting these weights into Cp = 0.5*wf + 0.3*mf + 0.2*kf, the evaluation factor CPOK for each qualified weld joint is calculated. Using the evaluation factor CPOK for each qualified weld joint, a distribution map of the qualified evaluation factors for each qualified weld joint is plotted. Based on the distribution map of the qualified evaluation factors for each qualified weld joint, the median value of the qualified evaluation factor for each qualified weld joint is determined.

[0075] An ensemble learning algorithm is used to assign weights to the standardized non-conforming waveform factor, standardized non-conforming margin factor, and standardized non-conforming kurtosis, respectively, and to determine their respective weights, where a1 = 0.5, a2 = 0.3, and a3 = 0.2. Substituting these weights into Cp = 0.5*wf + 0.3*mf + 0.2*kf, the evaluation factor CPNG for each non-conforming weld joint is calculated. Using the evaluation factor CPNG for each non-conforming weld joint, a distribution map of the non-conforming evaluation factors for each non-conforming weld joint is plotted, and the median of the non-conforming evaluation factors for each non-conforming weld joint is determined based on the distribution map.

[0076] The average of the median values ​​of the qualification evaluation factors of multiple qualified welded joints is denoted as CPOK1; the average of the median values ​​of the non-qualification evaluation factors of multiple unqualified welded joints is denoted as CPNG1; the average of CPOK1 and CPNG1 is denoted as CPHG, where CPHG = 1.5;

[0077] 7) Repeat steps 5) and 6) to calculate the evaluation factor of the quality-undetermined part in step 4), denoted as CP2. If CP2 ≥ CPHG (1.5), the quality-undetermined part is determined as a qualified welded joint; otherwise, the quality-undetermined part is a substandard welded joint.

[0078] See Figure 7 The high degree of dispersion in the signal characteristic quantities of a qualified weld joint alters the magnitude of the original signal characteristic quantity data distribution, changes the judgment baseline, and compromises the rationality and sufficiency of ultrasonic welding quality evaluation. Figure 7 In the diagram, the horizontal axis represents the number of the welded joints (including qualified and unqualified welded joints). Echo signals were acquired from 6 qualified and 6 unqualified welded joints, yielding 400 ultrasonic signals from each joint. After signal feature extraction, 400 waveform factors, 400 margin factors, and 400 kurtosis values ​​were obtained. The waveform factors, margin factors, and kurtosis values ​​of each welded joint were then standardized according to the aforementioned range method, resulting in the following for each welded joint: 400 standardized waveform factors, 400 standardized margin factors, and 400 standardized kurtosis factors are assigned weights; the corresponding evaluation factors for each weld joint are calculated, and 400 Cp (evaluation factors) can be obtained for each sample; the average of the qualified evaluation factors of multiple qualified weld joints is taken as CPOK2; the average of the unqualified evaluation factors of multiple unqualified weld joints is taken as CPNG2; let CPHG = (CPOK2 + CPNG2) / 2, where CPHG is the judgment benchmark value.

[0079] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for non-destructively evaluating the quality of ultrasonic welding at wire ends, characterized in that, comprising the following steps: 1) taking a plurality of qualified welded butt ends and a plurality of unqualified welded butt ends, and collecting an echo signal of each qualified welded butt end and an echo signal of each unqualified welded butt end; 2) calculating a waveform factor corresponding to the echo signal of each qualified welded butt end and a waveform factor corresponding to the echo signal of each unqualified welded butt end collected in step 1); said step 2) further comprises: calculating a clearance factor corresponding to the echo signal of each qualified welded butt end and a clearance factor corresponding to the echo signal of each unqualified welded butt end collected in step 1), and calculating a kurtosis corresponding to the echo signal of each qualified welded butt end and a kurtosis corresponding to the echo signal of each unqualified welded butt end collected in step 1); 3) drawing a waveform factor distribution diagram corresponding to each qualified welded butt end and a waveform factor distribution diagram corresponding to each unqualified welded butt end; determining a median waveform factor of each qualified welded butt end according to the waveform factor distribution diagram corresponding to each qualified welded butt end; taking an average value of the median waveform factors corresponding to the plurality of qualified welded butt ends, which is recorded as CPOK1; determining a median waveform factor of each unqualified welded butt end according to the waveform factor distribution diagram corresponding to each unqualified welded butt end; taking an average value of the median waveform factors corresponding to the plurality of unqualified welded butt ends, which is recorded as CPNG1; 4) acquiring a median value of a waveform factor distribution diagram of a welded butt end to be tested, which is recorded as CP1, wherein if CP1≥CPOK1, the welding quality of the welded butt end to be tested is qualified; if CP1<CPNG1, the welding quality of the welded butt end to be tested is unqualified; if CPOK1>CP1≥CPNG1, the welded butt end to be tested is a piece with undetermined quality; the steps of said method for non-destructive evaluation of ultrasonic welding quality of wire ends further comprise: 5) performing range standardization processing on the waveform factor corresponding to the echo signal of each qualified welded butt end, the waveform factor corresponding to the echo signal of each unqualified welded butt end, the clearance factor corresponding to the echo signal of each qualified welded butt end, the clearance factor corresponding to the echo signal of each unqualified welded butt end, the kurtosis corresponding to the echo signal of each qualified welded butt end and the kurtosis corresponding to the echo signal of each unqualified welded butt end, to obtain a standardized qualified waveform factor, a standardized unqualified waveform factor, a standardized qualified clearance factor, a standardized unqualified clearance factor, a standardized qualified kurtosis and a standardized unqualified kurtosis; 6) assigning weights to the standardized qualified waveform factor, the standardized qualified clearance factor and the standardized qualified kurtosis respectively by using an ensemble learning algorithm, obtaining an evaluation factor CPOK of each qualified welded butt end through weighted calculation; drawing a qualified evaluation factor distribution diagram of each qualified welded butt end by using the evaluation factor CPOK of each qualified welded butt end, and determining a median qualified evaluation factor of each qualified welded butt end according to the qualified evaluation factor distribution diagram of each qualified welded butt end; An ensemble learning algorithm is used to assign weights to the standardized non-conforming waveform factor, standardized non-conforming margin factor, and standardized non-conforming kurtosis, respectively. The evaluation factor CPNG of each non-conforming weld joint is obtained through weighted calculation. The non-conforming evaluation factor distribution map of each non-conforming weld joint is plotted using the evaluation factor CPNG of each non-conforming weld joint. The median of the non-conforming evaluation factor of each non-conforming weld joint is determined based on the non-conforming evaluation factor distribution map of each non-conforming weld joint. The average of the median values ​​of the qualified evaluation factors of multiple qualified welded joints is denoted as CPOK2; the average of the median values ​​of the unqualified evaluation factors of multiple unqualified welded joints is denoted as CPNG2; let CPHG = (CPOK2 + CPNG2) / 2, where CPHG is the judgment benchmark value; 7) Calculate the evaluation factor of the quality-undetermined part in step 4), denoted as CP2. If CP2 ≥ CPHG, then the quality-undetermined part is determined as a qualified welded joint; otherwise, the quality-undetermined part is a substandard welded joint.

2. The method for non-destructive evaluation of ultrasonic welding quality at wire ends according to claim 1, characterized in that, The formula for the weighted calculation in step 6) is: Cp = a1*wf + a2*mf + a3*kf In the formula, Cp is the evaluation factor, which is dimensionless; wf is the standardized waveform factor, which is dimensionless; mf is the standardized margin factor, which is dimensionless; kf is the standardized kurtosis, which is dimensionless; a1 is the weight of the waveform factor, which is dimensionless; a2 is the weight of the margin factor, which is dimensionless; and a3 is the weight of the kurtosis, which is dimensionless.

3. The method for non-destructive evaluation of ultrasonic welding quality at wire ends according to claim 1, characterized in that, The method for determining the waveform factor, margin factor, and kurtosis in step 2) is as follows: calculate the signal feature quantities corresponding to the echo signals of each qualified welded joint and each unqualified welded joint collected in step 1), and sort the signal feature quantities of each qualified welded joint and each unqualified welded joint using an ensemble learning algorithm to obtain the top three signal feature quantities in terms of importance, which are waveform factor, margin factor, and kurtosis, respectively.

4. The method for non-destructive evaluation of ultrasonic welding quality at wire ends according to claim 1, characterized in that, Step 1) specifically involves: using an ultrasonic phased array probe and a full matrix acquisition method to acquire the echo signal of each qualified weld joint and the echo signal of each unqualified weld joint.

5. The method for non-destructive evaluation of ultrasonic welding quality at wire ends according to claim 4, characterized in that, The ultrasonic phased array probe is a 32-element, 5MHz ultrasonic phased array probe.

Images