CN116275674A - Resistance welding quality detection method and system - Google Patents
Resistance welding quality detection method and system Download PDFInfo
- Publication number
- CN116275674A CN116275674A CN202310364045.6A CN202310364045A CN116275674A CN 116275674 A CN116275674 A CN 116275674A CN 202310364045 A CN202310364045 A CN 202310364045A CN 116275674 A CN116275674 A CN 116275674A
- Authority
- CN
- China
- Prior art keywords
- curve
- welding
- voltage
- resistance
- power
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K31/00—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups
- B23K31/12—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups relating to investigating the properties, e.g. the weldability, of materials
- B23K31/125—Weld quality monitoring
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Mechanical Engineering (AREA)
- Resistance Welding (AREA)
Abstract
The invention discloses a method and a system for detecting the quality of resistance welding, wherein the method comprises the following steps: measuring welding current and voltage, and calculating the welding current and voltage to obtain dynamic resistance and power; fitting a welding current curve, a voltage curve, a dynamic resistance curve and a power curve based on welding current, voltage, dynamic resistance and power, and extracting characteristics of the curves to obtain training characteristic quantities; and constructing a resistance welding quality evaluation model based on a BA-Catoost algorithm according to the training feature quantity, and obtaining a test feature quantity for testing to obtain an evaluation result. The system comprises: the device comprises a measuring module, a dynamic resistance and power calculating module, a curve fitting module, a characteristic extracting module, a model constructing module and a testing module. By using the invention, 100% on-line detection of the resistance welding quality can be realized under the condition of not damaging the welding structure, and the invention has the characteristics of convenient use, good reliability and high detection precision. The invention can be widely applied to the technical field of measurement.
Description
Technical Field
The invention relates to the technical field of measurement, in particular to a method and a system for detecting the quality of resistance welding.
Background
The resistance welding is a welding forming method which is widely applied to various fields of automobile manufacturing, aerospace and other industries, and has the advantages of short welding time, low cost, high efficiency, simple operation, high mechanization and automation degree and the like. Because the resistance welding process is influenced by relevant factors of the field environment, the defects of insufficient welding, overburning, overlarge splashing and the like are easy to occur at the welding part, and the quality of the welding spot directly influences the service life of a welded part, the defect detection and quality evaluation of the welding spot are very important. The traditional resistance welding quality detection mainly comprises two methods of periodic sampling for destructive detection and post-welding nondestructive detection, but the periodic sampling for destructive detection has the defects of welding material waste, welding cost increase and the like, and the post-welding nondestructive detection cannot be comprehensively applied in the welding production process due to the self characteristics of the detection method, so that 100% detection of welding spots cannot be ensured.
The existing nondestructive detection methods have the defects that the ultrasonic detection also has the problems that the bottom echo interferes with the surface echo, the coupling effect in the detection process is poor, and the like, and the ultrasonic detection is used as a welding quality detection method after welding is finished, manual sampling is needed, and quality detection cannot be carried out on all welding parts; the infrared detection technology is easy to be interfered by the environment, the thermal state inside the weldment cannot be determined, and the price of a detection instrument is relatively high; the X-ray detection is insensitive to the detection of the micro defects, the detection cost is high, and the rays are harmful to human bodies; the magneto-optical imaging detection can image in real time, but has a small application range; the magnetic powder detection is only applicable to ferromagnetic materials, and strict requirements are set on the surface state of a workpiece; the eddy current detection can detect the surface and subsurface of the material, but is difficult to detect the defects in the welding part and difficult to judge the size, shape and type of the welding defects; penetration detection is limited to detection of surface opening defects; the acoustic emission signal detection is easily interfered by environmental factors such as site electromagnetic and noise, and the detection equipment is expensive and complex; the laser holographic nondestructive detection technology depends on whether the internal defect of the detected object can cause corresponding deformation of the object surface under the action of external force; at present, 100% on-line detection of the resistance welding quality is realized under the condition that a welding structure is not damaged by a method which is convenient to use, good in reliability and high in detection precision.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a method and a system for detecting the quality of resistance welding, which can realize 100% on-line detection of the quality of resistance welding under the condition of not damaging a welding structure and have the characteristics of convenient use, good reliability and high detection precision.
The first technical scheme adopted by the invention is as follows: a resistance welding quality detection method comprises the following steps:
measuring welding current and voltage to obtain welding current and voltage;
calculating welding current and voltage to obtain dynamic resistance and power;
performing curve fitting on welding current, welding voltage, dynamic resistance and power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve;
extracting characteristics of a welding current curve, a voltage curve, a dynamic resistance curve and a power curve to obtain training characteristic quantities;
constructing a resistance welding quality evaluation model based on a BA-Catoost algorithm according to the training feature quantity;
and obtaining a test characteristic quantity, inputting the test characteristic quantity into a resistance welding quality evaluation model based on a BA-Catboost algorithm, and obtaining an evaluation result.
Further, before the characteristics of the welding current curve, the voltage curve, the dynamic resistance curve and the power curve are extracted to obtain the training characteristic quantity, the method further comprises the steps of calculating and analyzing the stability of the resistance welding process to obtain the characteristic quantity of the detection stage, and the method specifically comprises the following steps:
calculating the variation coefficients of the effective value and the maximum value of the current, the voltage, the dynamic resistance and the power in the multiple welding processes;
and analyzing the variation coefficient, and selecting a dynamic resistance maximum value, a dynamic resistance minimum value, a dynamic resistance effective value, a voltage maximum value and a power maximum value as characteristic values of a detection stage.
Through the preferred step, the stability of the resistance welding process is explored, so that the extracted characteristic quantity with larger variation coefficient can reflect the variation of the resistance welding quality in the resistance welding process.
Further, the step of measuring the welding current and the voltage to obtain the welding current and the voltage specifically includes:
the method comprises the steps that a Rogowski coil is matched with an integral amplifying circuit to collect and measure a welding current signal at the secondary side of a transformer;
signal acquisition and measurement of welding voltage in secondary loop by embedding twisted pair at output of secondary rectifier diode of transformer
Through the preferred step, a mode of combining the rogowski coil with the integral amplifying circuit can obtain a larger measurement range and higher measurement accuracy, the secondary side of the transformer is used for measuring in order to ensure measurement accuracy, and the twisted pair can reduce induction noise generated by secondary current on a wire loop by reducing the surrounding area of the wire.
Further, the step of calculating the welding current and the welding voltage to obtain the dynamic resistance specifically includes:
calculating resistance through a voltage balance equation of the secondary loop to obtain dynamic resistance containing inductive components;
calculating a secondary inductance estimated value through a power balance method column according to an average value formula and a fixed integral formula;
constructing a system of equations based on the dynamic resistance and secondary inductance estimates comprising the inductive component;
and carrying out iterative solution on the equation set by adopting a Gaussian-Sedel iteration method to obtain the dynamic resistance.
Through the preferred step, the influence of the inductive component on the secondary voltage and the dynamic resistance is greatly reduced, so that the obtained dynamic resistance is more approximate to the actual resistance value compared with the traditional measurement calculation method.
Further, the step of curve fitting the welding current, the welding voltage, the dynamic resistance and the power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve specifically comprises the following steps:
dividing the welding current, welding voltage, dynamic resistance and power data into sections respectively to obtain a plurality of section data;
performing function construction on the interval data to obtain a plurality of interval construction functions;
establishing an equation set based on interpolation conditions and interval construction functions;
and solving the equation set to respectively obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve.
Through the preferred step, the curve fitting is performed by adopting a cubic spline interpolation approximation method, so that the characteristic quantity extracted subsequently can be completely reserved in the curve.
Further, the step of constructing a resistance welding quality evaluation model based on the BA-Catboost algorithm according to the training feature quantity specifically comprises the following steps:
reading training characteristic quantity and preprocessing to obtain a training set;
constructing a decision tree based on the training set through a Catboost algorithm to obtain a Catboost decision tree structure;
performing global optimization on the Catboost decision tree structure based on the BA algorithm to obtain optimization parameters;
and the Catboost decision tree structure acquires optimization parameters to obtain a resistance welding quality evaluation model based on the BA-Catboost algorithm.
Through the preferred step, the current error value can be reduced, and higher evaluation accuracy can be achieved.
The second technical scheme adopted by the invention is as follows: a resistance weld quality inspection system, comprising:
the measuring module is used for measuring the welding current and the voltage to obtain the welding current and the voltage;
the dynamic resistance and power calculation module is used for calculating welding current and voltage to obtain dynamic resistance and power;
the curve fitting module is used for performing curve fitting on welding current, welding voltage, dynamic resistance and power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve;
the characteristic extraction module is used for extracting characteristics of a welding current curve, a voltage curve, a dynamic resistance curve and a power curve to obtain training characteristic quantities;
the model construction module is used for constructing a resistance welding quality assessment model based on a BA-Catboost algorithm according to the training feature quantity;
the test module is used for obtaining the test characteristic quantity and inputting the test characteristic quantity into a resistance welding quality evaluation model based on the BA-Catboost algorithm to obtain an evaluation result.
The method and the system have the beneficial effects that: according to the invention, the welding current and the welding voltage are measured at the secondary side of the transformer in a mode of matching the Rogowski coil with the integral amplifying circuit, so that a larger measuring range and higher measuring accuracy are obtained; the resistance welding quality assessment model based on the BA-Catboost algorithm has high convergence rate in the training process, high iteration efficiency and high quality detection precision, and can realize 100% on-line detection of the resistance welding quality under the condition of not damaging a welding structure.
Drawings
FIG. 1 is a flow chart of the steps of a method for detecting the quality of a resistance weld according to the present invention;
FIG. 2 is a block diagram of a resistance weld quality inspection system according to the present invention;
FIG. 3 is a schematic view of the structure of a Rogowski coil in a resistance welding quality inspection method according to the present invention;
FIG. 4 is a schematic diagram of an integrating circuit of a method for detecting the quality of resistance welding according to the present invention;
FIG. 5 is a flowchart of the BA-Catboost algorithm of a method for detecting the quality of a resistance weld according to the present invention.
Detailed Description
The invention will now be described in further detail with reference to the drawings and to specific examples. The step numbers in the following embodiments are set for convenience of illustration only, and the order between the steps is not limited in any way, and the execution order of the steps in the embodiments may be adaptively adjusted according to the understanding of those skilled in the art.
The invention provides a resistance welding quality detection method, which comprises the following steps:
s1, measuring welding current and voltage to obtain welding current and voltage data;
in order to ensure the measurement accuracy of welding current, the scheme uses the Rogowski coil to be matched with an integral amplifying circuit for current detection. Because the welding current in the resistance welding process has the characteristics of short time, large amplitude and the like, ensuring the measurement accuracy of the welding current is a key and necessary premise for realizing the welding quality detection. The conventional medium-frequency resistance welding equipment has a complex structure, and the primary side and secondary side current signals of the transformer have larger difference and higher harmonic frequency, which is approximately in the range of 1 kHz-10 kHz. The common current measurement method comprises current transformer measurement, hall sensor measurement and Rogowski coil measurement, wherein the current transformer has smaller measuring range and is arranged on the primary side of the transformer, nonlinear errors such as loss of the transformer and a rectifying tube are easy to generate, the measurement range of a common Hall sensor is also smaller, the measurement requirement of the secondary current of resistance welding is difficult to meet, and the large-range Hall sensor has huge volume and cannot be arranged on resistance welding equipment. The Rogowski coil has the characteristics of wide frequency range, large measurement range, high measurement accuracy and the like.
The subsequent calculation of the dynamic resistance needs to measure welding current and voltage synchronously, and in order to ensure measurement accuracy, the signal acquisition of the welding current needs to measure from the secondary of the transformer. The voltage signal measurement is performed in the secondary circuit, and can be divided into an electrode voltage signal measured from the electrode end and a secondary voltage signal measured from the transformer secondary rectifying diode according to the position. If electrode voltage is measured at the end of an electrode, a larger space electromagnetic field is generated around by secondary current in the welding electrifying process, induced voltage is generated on a loop of a measuring wire, the measuring accuracy of the voltage is affected, the induced noise can be reduced by reducing the surrounding area of the wire by using a twisted pair, but the method also has certain requirements on the installation of the twisted pair, otherwise normal welding production is affected, the normal installation mode is that two measuring wires are arranged around the whole throat of a welding machine along a welding tongs horn, and the exposed wires are easy to break and damage or be damaged by welding spatter for a vehicle body resistance welding machine mainly comprising welding tongs. The solution thus uses a method of embedding wires at the output of the secondary rectifier diode of the transformer to measure the voltage of the secondary loop.
Referring to fig. 3, the main principle of the rogowski coil current detection method is based on faraday's law of electromagnetic induction and ampere loop law, and when the measured current passes through the center of the rogowski coil along the axis, the expression of the current available according to ampere loop law is as follows:
∮H·dl＝I(t) (1)
wherein H is the intensity of the magnetic field inside the coil, l is the perimeter of the Rogowski coil, and I (t) is the magnitude of the current passing through the center of the Rogowski coil.
Due to
B＝μH (2)
φ＝BS (3)
Wherein B is the magnetic induction intensity inside the coil, mu is the magnetic permeability in the air, H is the magnetic field intensity inside the coil, phi is the magnetic flux, and S is the cross-sectional area of the current passing through the Rogowski coil.
According to the law of electromagnetic induction, the calculation formula of the induced electromotive force is as follows:
wherein e (t) is induced electromotive force, N is the number of turns of the Rogowski coil, and phi is magnetic flux.
Substituting the formulas (1), (2) and (3) into the formula (4) can obtain:
in summary, the voltage signal e (t) output by the rogowski coil is actually proportional to the differential of the current signal, so the rogowski coil output signal must be processed by an integrating circuit to obtain the true value of the welding current.
Referring to FIG. 4, wherein R 1 ＝R 3 The homodromous input end of the integrated operational amplifier passes through a resistor R 2 The potential of the two points of the reverse input end and the same-direction input end is zero and is virtual ground. The current in the circuit through the capacitor C is equal to the current flowing through the resistor R 1 The expression of which is as follows:
the relation between the capacitance voltage and the output voltage is:
e 0 (t)＝-u C (7)
from the relationship between capacitance voltage and current, it is possible to:
substituting the formula (5) into the formula (8) can obtain the corresponding relation between the circuit output voltage signal and the tested current signal as follows:
therefore, the welding current in the welding process can be obtained through the Rogowski coil current detection method and the integration circuit.
S2, calculating welding current and voltage to obtain dynamic resistance and power;
the measured secondary voltage signal from the transformer secondary rectifier diode, while less affected by electromagnetic interference, loop inductance will affect the accuracy of conventional resistance calculation methods when calculating the dynamic resistance of the secondary loop. In order to eliminate the inductive component in the secondary voltage and the dynamic resistance, the scheme adopts an iterative calculation method to calculate the dynamic resistance.
And (3) equivalent the secondary loop to a RL circuit model to obtain a voltage balance equation of the secondary loop, wherein the equation is as follows:
wherein u is s For the secondary voltage measured at the rectifier diode, i s Secondary current when energizing welding, R s Resistance of secondary loop, L s Is the inductance of the secondary loop.
As can be seen from equation (10), the secondary voltage includes both a inductive component and a resistive component. Inductance L in the secondary loop s Using a secondary voltage u other than zero s Divided by secondary current i s The resulting value will also contain an perceptual component. In order to accurately find the most realistic secondary resistance value, it is necessary to eliminate the inductive components contained in the secondary voltage and the dynamic resistance by calculation. In an inversion period, calculating resistance through a voltage balance equation of a secondary loop to obtain dynamic resistance containing inductive components, wherein the formula is as follows:
wherein T is the inversion period of the power supply,
is the average value of the secondary voltage in one inversion period,/-, and>
is the average value of the secondary current in an inversion period, L s Is the inductance of the secondary loop.
From the average value formula and the fixed integral formula:
because in the process of one-time power-onIn the secondary inductance L s It can be considered that the method is almost unchanged, and the method adopts a power balance method to calculate L in a column mode s The power balance equation for the secondary loop is as follows:
wherein P is T For instantaneous total power, P E For the resistance to consume instantaneous active power, P L The inductance consumes instantaneous waste power.
The secondary inductance estimation values obtained by the various average values in the simultaneous (13) expression are as follows:
wherein the method comprises the steps of
Is the average value of the product of the secondary voltage and the secondary current, +.>
Is the average of the squares of the secondary currents.
The secondary inductance obtained by iterative calculation of each inversion period is caused by measurement errors and calculation errors
May be different, by taking a period of time +.>
Method for improving inductance L by average value s Accuracy of the measured values. The secondary resistance R can be obtained by combining the formula (11) and the formula (14) s And secondary inductance L s Adopts a Gaussian-Sedel iteration method to carry out iteration solution on the equation set, and R is after 3 iterations s And L is equal to s Is basically converged. Each inversion period can be calculated for 3 times by adopting the method, and the method is realized byThe method can greatly reduce the influence of the inductive component on the secondary voltage and the dynamic resistance, and further obtain the dynamic resistance value which is more similar to the real resistance.
The calculation of the power is relatively simple, and the calculation expression is as follows:
p＝u s ×i s (15)
where p represents power, u s For the secondary voltage measured at the rectifier diode, i s Secondary current when energized for welding.
S3, curve fitting is carried out on welding current, welding voltage, dynamic resistance and power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve;
s3.1, respectively dividing the welding current, the welding voltage, the dynamic resistance and the power data into sections to obtain a plurality of section data;
the welding voltage and welding current measured by high-frequency sampling and the dynamic resistance and power data obtained by calculation are expressed as discrete points, and the whole process interval of the collected n+1 sample points is divided into n intervals [ (x) 0 ,x 1 ),(x 1 ,x 2 ),…,(x n-1 ,x n )]N+1 points in total.
S3.2, performing function construction on the interval data to obtain a plurality of interval construction functions;
the invention adopts a cubic spline interpolation method, so that a cubic function is constructed for each interval, which is S respectively 0 (x)～S n-1 (x)。
S3.3, establishing an equation set based on interpolation conditions and interval construction functions;
the cubic spline interpolation method needs to satisfy the following interpolation conditions:
1. the n-piece cubic function must traverse all known nodes, namely: s is S i (x i )＝y i ,i＝0,1,2,…,n。
2. The 0 th order is continuous (data is guaranteed to be uninterrupted and no jump) at all other nodes except the first node and the last node, and the function value of the former section equation at the node and the function value of the latter section equation at the same nodeThe function values are equal, namely: s is S i (x i+1 )＝S i+1 (x i+1 )。
3. Except the first node and the last node, the 1 st order is continuous at all other nodes (ensuring that the nodes have the same slope and no intense jump exists on the original function curve), namely:
4. except for the first node and the last node, the 2 nd order succession at all the remaining nodes (ensuring the same curvature at the nodes) is:
the equation set expression is as follows:
and S3.4, solving the equation set to respectively obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve.
And S4, extracting characteristics of a welding current curve, a voltage curve, a dynamic resistance curve and a power curve to obtain training characteristic quantities.
After the design of the measuring device is completed and before the training characteristic quantity is extracted, three main welding process parameters of welding current, welding time and electrode pressure are selected, and an orthogonal experiment is designed, wherein the specific experimental steps comprise:
determining an optimal welding process parameter combination by taking the diameter of a welding core and the tensile shear strength as evaluation indexes;
specifically, a universal tensile testing machine is adopted to conduct a tensile test on the welding joint at normal temperature, the tensile strength of the welding spot is obtained from a load displacement curve in the tensile test, and the diameter of the welding core is obtained by measuring the root size of the welding core left on a base material after the welding spot is damaged by using a vernier caliper. Whether the quality of the welding spot is qualified or not is mainly judged by judging the welding strength of the welding spot, under the condition of ensuring good appearance, the nugget diameter and the tensile shear strength are important judging standards for the internal quality of the welding spot, and the qualified welding spot has rigid requirements on the nugget diameter and the tensile shear strength and is determined according to different workpieces made of different materials.
Carrying out continuous large-scale resistance spot welding on the basis of the optimal welding process parameter combination; measuring welding current and voltage in the welding process to obtain the welding current and voltage in the welding process; calculating variation coefficients and average values of a welding current curve, a secondary voltage curve and a dynamic resistance curve at the same moment in the welding process; analyzing the variation coefficient and the mean value to obtain an analysis result of the welding process;
the result shows that the resistance spot welding process is a process with larger early fluctuation and stable later period, and the fluctuation of the dynamic resistance curve is maximum in the early period, and the fluctuation of the secondary current curve and the fluctuation of the secondary voltage curve are relatively similar. The analysis is carried out according to the dynamic resistance curve, and the dynamic resistance in the resistance welding is also a process with larger early fluctuation, so the variation coefficient of the maximum value of the dynamic resistance is larger than the effective value of the dynamic resistance, and the difference analysis of the variation coefficient is required to be further explored.
To explore the difference in the magnitude of the coefficient of variation, the coefficients of variation of the effective and maximum values of current, voltage, dynamic resistance and power for the multiple welding processes are calculated. The coefficient of variation obtained by calculation and analysis is as follows in sequence from large to small: the dynamic resistance maximum value, the dynamic resistance effective value, the voltage maximum value, the voltage effective value, the power maximum value, the power effective value, the current maximum value and the current effective value. Aiming at the influence of electrode abrasion condition, weldment surface quality and small margin on resistance welding quality, a control variable method is adopted to analyze a welding current curve, a secondary voltage curve, a dynamic resistance curve and a power curve under the three influence factors respectively, data phase difference scale is large due to different units in the resistance spot welding process, and the fluctuation of different parameters in the welding process can be eliminated by using the variation coefficient to quantify the fluctuation of the different parameters in the welding process because the variation coefficient has no dimension, and meanwhile, the fluctuation of the parameters in the welding process can be objectively compared by omitting the dimension of the different parameters in the welding process. The larger the variation coefficient is, the more easily the surface parameter is affected by the interference of external factors, and the more compact the relation between the obvious variation of the parameter with larger variation coefficient and welding quality is, so after the curve condition is analyzed, the six characteristic quantities of the dynamic resistance maximum value, the dynamic resistance minimum value, the dynamic resistance effective value, the voltage maximum value and the power maximum value are finally extracted as the characteristic quantities for welding quality detection.
S5, constructing a resistance welding quality evaluation model based on a BA-Catboost algorithm according to the training feature quantity;
according to the invention, six characteristic quantities of the maximum value, the minimum value, the effective value, the maximum value and the maximum value of the voltage extracted in the step S3 are taken as inputs, the nugget diameter and the tensile and shearing strength of welding spots are taken as outputs, a decision tree is constructed by using a Catboost algorithm, global optimization is carried out on the Catboost decision tree structure based on a BA algorithm, and a resistance welding quality evaluation model based on the BA-Catboost algorithm is established. The Catboost is a novel machine learning algorithm based on a gradient enhancement decision tree algorithm, and is greatly improved when processing a large amount of data and features. Firstly, the Catboost algorithm avoids the problem of conditional displacement inherently existing in the iterative process, and enables the Catboost algorithm to train and learn by utilizing the whole data set; secondly, the Catboost algorithm converts the traditional gradient enhancement algorithm into an ordered enhancement algorithm, so that the unavoidable problem of gradient deviation in the iterative process is solved, the generalization capability is improved, the possibility of model overfitting is reduced, and the robustness of the model is enhanced; then, the Catboost algorithm constructs combinations of classification features through a greedy strategy, and the combinations are used as additional features, so that the model is facilitated to capture higher-order dependency relationships more easily, and prediction accuracy is further improved; in addition, the Catboost algorithm selects a forgetting decision tree as a basic prediction period, so that the possibility of over-fitting is reduced, and the execution speed of the model is increased. The BA algorithm (bat algorithm) performs global optimization of parameters in a manner of simulating echo positioning of bats to optimize the 3 most critical parameters of the CatBoost model, namely the number of decision trees, the learning rate and the maximum depth of the trees. The gradient enhancement function can be enhanced, and the prediction capability is obviously improved.
Referring to fig. 5, training feature values are read and preprocessed to obtain a training set; constructing a decision tree based on the training set through a Catboost algorithm to obtain a Catboost decision tree structure; and performing global optimization on the Catboost decision tree structure based on the BA algorithm to obtain optimization parameters.
Setting the nugget diameter and the tensile and shear strength of the welding spot as an objective function min f(x) The number of decision trees, the learning rate and the maximum depth of the tree are taken as target variables, i.e. x= (X) 1 ,x 2 ,…,x d ) T Is an optimization problem of (a). First, a set of initial solutions are set in a random manner in space, the maximum pulse volume is set as A 0 The maximum pulse rate is R 0 Search pulse frequency f i Is in the range f min To f max The attenuation coefficient of the pulse loudness is alpha, the enhancement coefficient of the search frequency is gamma, and the search precision epsilon or the maximum iteration number item is set max 。
Setting the bat position x of initial state i Finding the minimum solution X of the current error according to the approach degree of the target value * And updating the searching pulse frequency, speed and position of the bat:
f i ＝f min +(f max -f min )β (17)
wherein f i Is the search pulse frequency of bat i, beta is the pulse frequency of [0,1 ]]Is used to determine the random number of the uniform distribution,
respectively representing the speed of bat i at time t and time t-1, X * Is the best solution in all bats currently.
Then generating a uniformly distributed randomThe number rand e [0,1 ]]If rand>Pulse emissivity r i Randomly disturbing the current solution to generate a new solution and performing out-of-range processing; if rand<Pulse loudness A i And f (x) i )<f(X * ) Then a new solution is accepted and loudness and pulse rate are calculated according to the following equation:
wherein the method comprises the steps of
And->
Indicating the pulse loudness of bat i at time t+1 and t, respectively, +.>
Representing the pulse rate of bat i at time t + 1.
And then sequencing the update positions and the update speeds of the bats according to the fitness value, determining the minimum value of the current error, repeating the iteration steps until the set optimal solution condition is met or the maximum iteration times are reached, and finally outputting the global optimal value and the optimal solution. And finally, acquiring optimization parameters by the Catboost decision tree structure to obtain a resistance welding quality evaluation model based on the BA-Catboost algorithm.
The analysis of the bat algorithm calculation formula shows that the attenuation coefficient alpha of pulse loudness and the enhancement coefficient gamma of search frequency have great influence on the performance of the algorithm, and the algorithm performance is improved by continuously adjusting the values of alpha and gamma in the simulation process.
S6, obtaining test feature quantity and inputting the test feature quantity into a resistance welding quality evaluation model based on a BA-Catboost algorithm to obtain an evaluation result, wherein the evaluation result comprises a welding core diameter and a welding spot tensile shear strength and is used as an index for evaluating welding quality. And extracting characteristic quantities from the acquired data according to the method, establishing a quality evaluation model based on a BA-Catboost algorithm, randomly selecting 25% of the data as a test set, and using the rest data as an input training set. Experiments are carried out on low carbon steel with the thickness of 100mm multiplied by 40mm multiplied by 0.7mm, and the prediction capability of the BA-Catboost algorithm model is superior to that of the XGBboost algorithm, the Catboost algorithm and the random forest algorithm through comparative analysis, so that higher evaluation accuracy can be achieved.
As shown in fig. 2, a resistance welding quality detection system includes:
the measuring module is used for measuring the welding current and the voltage to obtain the welding current and the voltage;
the dynamic resistance and power calculation module is used for calculating welding current and voltage to obtain dynamic resistance and power;
the curve fitting module is used for performing curve fitting on welding current, welding voltage, dynamic resistance and power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve;
the characteristic extraction module is used for extracting characteristics of a welding current curve, a voltage curve, a dynamic resistance curve and a power curve to obtain training characteristic quantities;
the model construction module is used for constructing a resistance welding quality assessment model based on a BA-Catboost algorithm according to the training feature quantity;
the test module is used for obtaining the test characteristic quantity and inputting the test characteristic quantity into a resistance welding quality evaluation model based on the BA-Catboost algorithm to obtain an evaluation result.
The content in the method embodiment is applicable to the system embodiment, the functions specifically realized by the system embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
While the preferred embodiment of the present invention has been described in detail, the invention is not limited to the embodiment, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the invention, and these modifications and substitutions are intended to be included in the scope of the present invention as defined in the appended claims.
Claims (7)
1. The resistance welding quality detection method is characterized by comprising the following steps of:
measuring welding current and voltage to obtain welding current and voltage;
calculating welding current and voltage to obtain dynamic resistance and power;
performing curve fitting on welding current, welding voltage, dynamic resistance and power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve;
extracting characteristics of a welding current curve, a voltage curve, a dynamic resistance curve and a power curve to obtain training characteristic quantities;
constructing a resistance welding quality evaluation model based on a BA-Catoost algorithm according to the training feature quantity;
and obtaining a test characteristic quantity, inputting the test characteristic quantity into a resistance welding quality evaluation model based on a BA-Catboost algorithm, and obtaining an evaluation result.
2. The method for detecting the quality of the resistance welding according to claim 1, wherein before the characteristics of the welding current curve, the voltage curve, the dynamic resistance curve and the power curve are extracted to obtain the training characteristic quantity, the method further comprises the step of calculating and analyzing the stability of the resistance welding process to obtain the characteristic quantity of the detection stage, and the method specifically comprises the following steps:
calculating the variation coefficients of the effective value and the maximum value of the current, the voltage, the dynamic resistance and the power in the multiple welding processes;
and analyzing the variation coefficient, and selecting a dynamic resistance maximum value, a dynamic resistance minimum value, a dynamic resistance effective value, a voltage maximum value and a power maximum value as characteristic values of a detection stage.
3. The method of claim 1, wherein the step of measuring the welding current and voltage to obtain the welding current and voltage comprises:
the method comprises the steps that a Rogowski coil is matched with an integral amplifying circuit to collect and measure a welding current signal at the secondary side of a transformer;
and carrying out signal acquisition measurement of welding voltage in the secondary loop by embedding a twisted pair at the output end of the secondary rectifying diode of the transformer.
4. The method for detecting the welding quality of a resistor according to claim 1, wherein the step of calculating the welding current and the welding voltage to obtain the dynamic resistance and the power comprises the following steps:
calculating resistance through a voltage balance equation of the secondary loop to obtain dynamic resistance containing inductive components;
calculating a secondary inductance estimated value through a power balance method column according to an average value formula and a fixed integral formula;
constructing a system of equations based on the dynamic resistance and secondary inductance estimates comprising the inductive component;
carrying out iterative solution on the equation set by adopting a Gaussian-Sedel iteration method to obtain a dynamic resistance;
and carrying out product operation on the welding current and the voltage to obtain power.
5. The method for detecting the welding quality of a resistor according to claim 1, wherein the step of curve fitting the welding current, the welding voltage, the dynamic resistance and the power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve comprises the following steps:
dividing the welding current, welding voltage, dynamic resistance and power data into sections respectively to obtain a plurality of section data;
performing function construction on the interval data to obtain a plurality of interval construction functions;
establishing an equation set based on interpolation conditions and interval construction functions;
and solving the equation set to respectively obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve.
6. The method for detecting the quality of the resistance welding according to claim 1, wherein the step of constructing a resistance welding quality evaluation model based on a BA-Catboost algorithm according to the training feature quantity specifically comprises the steps of:
reading training characteristic quantity and preprocessing to obtain a training set;
constructing a decision tree based on the training set through a Catboost algorithm to obtain a Catboost decision tree structure;
performing global optimization on the Catboost decision tree structure based on the BA algorithm to obtain optimization parameters;
and the Catboost decision tree structure acquires optimization parameters to obtain a resistance welding quality evaluation model based on the BA-Catboost algorithm.
7. A resistance weld quality inspection system, comprising:
the measuring module is used for measuring the welding current and the voltage to obtain the welding current and the voltage;
the dynamic resistance and power calculation module is used for calculating welding current and voltage to obtain dynamic resistance and power;
the curve fitting module is used for performing curve fitting on welding current, welding voltage, dynamic resistance and power to obtain a welding current curve, a voltage curve, a dynamic resistance curve and a power curve;
the characteristic extraction module is used for extracting characteristics of a welding current curve, a voltage curve, a dynamic resistance curve and a power curve to obtain training characteristic quantities;
the model construction module is used for constructing a resistance welding quality assessment model based on a BA-Catboost algorithm according to the training feature quantity;
the test module is used for obtaining the test characteristic quantity and inputting the test characteristic quantity into a resistance welding quality evaluation model based on the BA-Catboost algorithm to obtain an evaluation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310364045.6A CN116275674A (en) | 2023-04-07 | 2023-04-07 | Resistance welding quality detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310364045.6A CN116275674A (en) | 2023-04-07 | 2023-04-07 | Resistance welding quality detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116275674A true CN116275674A (en) | 2023-06-23 |
Family
ID=86779894
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310364045.6A Pending CN116275674A (en) | 2023-04-07 | 2023-04-07 | Resistance welding quality detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116275674A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117993345A (en) * | 2024-04-07 | 2024-05-07 | 广东电网有限责任公司汕尾供电局 | Pole tower ground resistance value data periodic monitoring and analysis system |
-
2023
- 2023-04-07 CN CN202310364045.6A patent/CN116275674A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117993345A (en) * | 2024-04-07 | 2024-05-07 | 广东电网有限责任公司汕尾供电局 | Pole tower ground resistance value data periodic monitoring and analysis system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | An application of back propagation neural network for the steel stress detection based on Barkhausen noise theory | |
| JP4885068B2 (en) | Eddy current flaw detector and eddy current flaw detection method | |
| CN116275674A (en) | Resistance welding quality detection method and system | |
| WO2024036858A1 (en) | Eddy-current testing circuit, method and system, storage medium, and terminal | |
| Pereira et al. | Modeling and design optimization of an eddy current sensor for superficial and subsuperficial crack detection in inconel claddings | |
| CN109521087A (en) | A kind of pulse eddy current detection method with the corrosion of clad ferromagnetic conduit wall thickness | |
| Bao et al. | Thickness measurement of titanium alloy sheet based on eddy current method with a novel simplified model | |
| Wu et al. | Magnetic flux leakage course of inner defects and its detectable depth | |
| Ge et al. | Investigation of the approximate decomposition of alternating current field measurement signals from crack colonies | |
| Zhao et al. | Coiled tubing wall thickness evaluation system using pulsed alternating current field measurement technique | |
| Faurschou et al. | Pulsed eddy current probe optimization for steel pipe wall thickness measurement | |
| Knopp et al. | Considerations in the validation and application of models for eddy current inspection of cracks around fastener holes | |
| Li et al. | Estimation of stress distribution in ferromagnetic tensile specimens using low cost eddy current stress measurement system and BP neural network | |
| CN112945427A (en) | Method for measuring two-dimensional stress at welding seam by utilizing Barkhausen effect and detection instrument | |
| He et al. | Evaluation of mechanical properties in medium carbon steel with a point mode electromagnetic sensor | |
| Zhao et al. | A surface crack assessment method unaffected by lift-off based on ACFM | |
| Li et al. | Thickness measurement of thermal barrier coating based on mutual inductance of eddy current system | |
| Goldfine et al. | Introduction to the Meandering Winding Magnetometer (MWM) and the grid measurement approach | |
| Huang et al. | Magnetic Flux Leakage: Theories and Imaging Technologies | |
| Bao et al. | A deep belief network and least squares support vector machine method for quantitative evaluation of defects in titanium sheet using eddy current scan image | |
| Bernieri et al. | A biaxial probe for nondestructive testing on conductive materials | |
| Zhao et al. | A defect visualization method based on ACFM signals obtained by a uniaxial TMR sensor | |
| CN117607243B (en) | Magnetic memory detection system and method for cross pipeline welding seams | |
| CN205941245U (en) | A portable hardness testing device for bearing ring | |
| CN218412363U (en) | Eddy current detection probe and detection circuit based on combination of differential bridge and transformer conditioning circuit |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |