CN106624266B - A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding - Google Patents
A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding Download PDFInfo
- Publication number
- CN106624266B CN106624266B CN201611267754.9A CN201611267754A CN106624266B CN 106624266 B CN106624266 B CN 106624266B CN 201611267754 A CN201611267754 A CN 201611267754A CN 106624266 B CN106624266 B CN 106624266B
- Authority
- CN
- China
- Prior art keywords
- weld seam
- welding
- penetration
- deviation
- penetration signal
- 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.)
- Active
Links
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
- B23K9/00—Arc welding or cutting
- B23K9/095—Monitoring or automatic control of welding parameters
-
- 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
- B23K37/00—Auxiliary devices or processes, not specially adapted to a procedure covered by only one of the preceding main groups
-
- 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
- B23K9/00—Arc welding or cutting
- B23K9/02—Seam welding; Backing means; Inserts
-
- 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
- B23K9/00—Arc welding or cutting
- B23K9/095—Monitoring or automatic control of welding parameters
- B23K9/0956—Monitoring or automatic control of welding parameters using sensing means, e.g. optical
-
- 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
- B23K2101/00—Articles made by soldering, welding or cutting
- B23K2101/006—Vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Plasma & Fusion (AREA)
- Optics & Photonics (AREA)
- Butt Welding And Welding Of Specific Article (AREA)
Abstract
The invention belongs to Automobile Welding equipment automatization technical field, specially a kind of weld seam deviation and penetration signal monitoring method for Automobile Welding includes the following steps：S1：Crater image is obtained, and intercepts pending region；S2：Information is extracted from pending region；S3：Using molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and the welding current I in information as welding process characteristic parameters, it is input to weld seam deviation and penetration signal one neural network model, weld seam deviation e and penetration signal p is calculated；S4：Welding process is predicted by weld seam deviation e and penetration signal p.Purpose of the present invention is to be guiding with the information of the crater image of acquisition, construct the weld seam deviation and penetration signal one neural network model for integrating welding deviation and penetration signal prediction, ensure to realize that electric arc is directed at weld seam in welding process, and realize the effect of joint penetration, entire welding process is predicted and monitored, realizes Automobile Welding process automation.
Description
Technical field
The invention belongs to Automobile Welding equipment automatization technical fields, specially a kind of weld seam deviation for Automobile Welding
And penetration signal monitoring method.
Background technology
Automobile has become current internal people and lives a kind of indispensable vehicles, vehicle body, chassis, rear axle,
The critical components such as subframe are the stress parts by being welded together.These parts are mainly punch welding part, plate
Thickness is the thin plate of 1.5mm~4mm, and weld seam splicing is close.Welding structure is to overlap, based on corner joint.The quality of welding quality,
The quality or even traffic safety of direct relation vehicle.
Welding is (such as to weld external medium by the means such as additional heat source (such as laser, plasma, electric arc)
Silk, welding rod or weldment itself) melt, to realize the connection of workpiece.For Automobile Welding part, good effect is obtained
Fruit, the control to welding process are crucial.Control Welding Process includes weld joint tracking and final appearance of weld control two
Broad aspect.
Weld joint tracking, i.e. weld seam deviation track, and are mainly ensured that in the welding process, heat source, that is, electric arc can be right always
Quasi- weld seam.And weld joint tracking lead-through teaching and sensor tracing before being weldering using more at present, for lead-through teaching before weldering, by
Weldment before welding clamps that there are the weldments in error and welding process there are thermal deformation, if heat source is still according to advance fixed
Path movement, just often deviate weld seam, this method can not overcome tracking error caused by welding thermal deformation.And for passing
Sensor tracing, mainly by additional sensor (such as mechanical probes, ultrasonic wave, arc sensor etc.) realize weld seam with
Track, in mechanical probes formula tracing, probe is easy to wear；And ultrasound involves arc sensor method and is affected by welding procedure.
For some emerging welding procedures, such as laser welding, heat source is very small (minimum up to 160m), laser beam and weld seam it
Between deviation workpiece can be caused to scrap when being more than 0.2mm.And the effect of traditional tracing is more helpless.Wherein vision with
Track method is used widely because of the advantages that its acquisition contains much information, non-contact in weld joint tracking field, but its generally require by
The algorithm operation of a series of complex, real-time are to be improved.
Appearance of weld mainly reflects the penetration signal situation of weld seam, is directly related to weld seam and periphery welding hot shadow
The metal structure and performance in area are rung, and then determines the mechanical characteristic of final weldment, it is huge to the reliability effect of final products
Greatly.For appearance of weld at present still by the experience of welder, the x-ray image by observing welding region, or to weldment
It is sampled, splits welding region and controlled.Control method is inconvenient.
Invention content
In order to overcome shortcoming and defect existing in the prior art, the purpose of the present invention is to provide one kind to weld for automobile
The weld seam deviation and penetration signal monitoring method connect is guiding with the information of the crater image of acquisition, forms Multi-information acquisition
For the method for welding process monitoring.By visual sensor and the collected information of current sensor, it is inclined that extraction influences weld seam
The welding process characteristic parameters of difference and penetration signal, and the welded condition measurement amount of entire welding process is extracted, construct collection
Welding deviation and penetration signal prediction ensure welding process in the weld seam deviation and penetration signal one neural network model of one
Middle realization electric arc is directed at weld seam, and realizes the effect of joint penetration, and entire welding process is predicted and monitored, and realizes vapour
Vehicle welding process automation.
Technical scheme is as follows：A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding, packet
Include following steps：
S1：Crater image is obtained, and intercepts pending region；
S2：Information is extracted from pending region；
S3：Using molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and the welding current I in information as welding
Process characteristic parameter is input to weld seam deviation and penetration signal one neural network model, and weld seam deviation e is calculated and melts
Saturating state p；
S4：Welding process is predicted by weld seam deviation e and penetration signal p.
Further, the acquisition methods of weld seam deviation and penetration signal one neural network model are specially：
S3.1：Crater image information is obtained, and intercepts pending region；
S3.2：Information is extracted from pending region, for information as welding process characteristic parameters, described information includes molten
Pond grey scale centre of gravity C, outer pool width W, interior pool width N and welding current I；
S3.3：Weld seam deviation e is measured from pending region0And penetration signal p0；
S3.4：For molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and welding current I and corresponding weld seam
Deviation e0And penetration signal p0Establish several groups sample；
S3.5：Establish weld seam deviation and penetration signal one neural network model：Wherein, the weld seam deviation and penetration shape
The input layer of state one neural network model includes molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and welding current
Totally 4 neurons, output layer include weld seam deviation e and penetration signal p totally 2 neurons to I, and hidden layer includes 20 nerves
Member, hidden layer using tansig () as transmission function, output layer with purelin () for transmission function, and to weld seam deviation and molten
Saturating state one neural network model is trained.
Further, values of the grey scale centre of gravity C in molten bath on the directions i is：Wherein, i, j are institute
The both direction in pending region is stated, K, L are respectively i, and the quantity of the directions j pixel, g (i, j) is the ash at pixel (i, j)
Angle value.
Further, the processing method for obtaining crater image information is to add neutral dim light to handle using narrow-band-filter.
Further, during obtaining crater image, judge electric arc whether be aligned weld seam the specific steps are：
S1.1：Neutral dim light processing is added to obtain bath image information by CCD camera units and narrow-band-filter；
S1.2：The intercepting process area information in the bath image information, by measure and calculation processing area information,
Obtain gray uniformization, current conditions and weld seam ambient heat state；
S1.3：If entire weldment is in a kind of state of balance, specially：Molten bath and arc shape rule, gray scale are equal
Even, electric current is steady and weld seam ambient heat beinthebalancestate, then may determine that and be directed at weld seam for electric arc；
S1.4：If entire weldment is in a kind of unbalanced state, specially：Molten bath and arc shape are irregular, grey
It spends that uneven, electric current is unstable and weld seam ambient heat is uneven, then may determine that as electric arc misalignment weld seam.
Beneficial effects of the present invention：The purpose of the present invention is to provide a kind of weld seam deviations and penetration for Automobile Welding
State monitoring method is guiding with the information of the crater image of acquisition, and form Multi-information acquisition is directed to what welding process monitored
Method.By visual sensor and the collected information of current sensor, extraction influences the weldering of weld seam deviation and penetration signal
Termination process characteristic parameters, and the welded condition measurement amount of entire welding process is extracted, construct collection welding deviation and penetration signal
It predicts the weld seam deviation and penetration signal one neural network model in one, ensures to realize electric arc alignment weldering in welding process
Seam, and realize the effect of joint penetration, entire welding process is predicted and monitored, realizes that Automobile Welding process is automatic
Change.
Description of the drawings
Fig. 1 is the flow chart of the present invention.
Fig. 2 is the weld seam deviation and penetration signal one neural network model schematic diagram of the present invention.
Fig. 3 is the weld seam deviation verification situation comparison diagram of the present invention.
Fig. 4 is the penetration signal verification situation comparison diagram of the present invention.
Fig. 5 is the schematic diagram of image obtained by the common process of the present invention.
Fig. 6 is the schematic diagram that the narrow-band-filter of the present invention adds neutral dim light processing gained image.
Fig. 7 is the schematic diagram in the pending region of interception of the present invention.
Schematic diagram when Fig. 8 is the electric arc alignment weld seam of the present invention.
Schematic diagram when Fig. 9 is the electric arc deviation weld seam of the present invention.
Specific implementation mode
For the ease of the understanding of those skilled in the art, the present invention is made with reference to specific embodiment and attached drawing further
Explanation, the content that embodiment refers to not is limitation of the invention.
In the present embodiment, several welding currents I is set, the reality per group welding is carried out for each welding current I of setting
It tests, wherein the welding current I is measured by current sensor, and the Welding experiment needs to set specific welding procedure ginseng
Number, specially：
a：Experiment workpiece is set as 45# steel, and size is：200mm×150mm×2mm；b：Experiment uses argon gas, argon gas stream
Amount is：9L/min；c：The value range of welding current I is：70A-80A；d：The speed of welding of Welding experiment is：0.6m/min-
1.2m/min；e：The period that CCD camera units use for：40ms.
During progress is tested in every group welding, by CCD camera units, i.e. visual sensor, and narrowband is used
It filters plus neutral dim light processing obtains crater image.Because when Welding experiment carries out, the arc light sent out strongly, according to
Common process, then as shown in figure 5, since the welding process of automobile is a typical non-linear process, often be accompanied by arc light,
The interference such as flue dust, splashing, the information inside crater image are often covered by strong arc light and can not be obtained.And use narrowband
It filters plus neutral dim light is handled, then as shown in Figure 6, it will be able to reduce the information being capped in the crater image of acquisition, increase obtains
Information in the crater image obtained.
In the present embodiment, the crater image of acquisition needs to include following situations：a：The case where 1 face weld seam 4 of electric arc；b：Electricity
Arc 1 deviates the case where weld seam 4；c：The case where 4 penetration of weld seam；d：The case where 4 non-penetration of weld seam；e：The excessively saturating situation of weld seam 4.This
Sample can obtain the information of a variety of situations encountered in welding process so that prediction is more accurate.
In the present embodiment, as shown in fig. 7, pending region 5 is intercepted on the bath image, from pending region
5 middle extraction information obtains welding process characteristic parameters, corresponding molten bath grey scale centre of gravity C is specially tested per group welding, outside
Pool width W, interior pool width N and welding current I.
The information in the pending region 5 specifically needs to be truncated to：The image information of electric arc 1, outer molten bath 3 image letter
Breath, the image informations such as the image information in interior molten bath 2 and the weld seam 4 of fore-end.
Values of the molten bath grey scale centre of gravity C on the directions i be：
Wherein, i, j are the both direction in the pending region 5, and the K in formula (1), L is respectively i, the directions j pixel
Quantity, g (i, j) are the gray value at pixel (i, j), and gray value can be directly in the information in pending region 5 directly
It reads.Correction phenomenon is not generated since the directions j are a directions of feed for plane-welding, it is little to deviation action, thus it is molten
Pond grey scale centre of gravity C omits the effect in the directions j.
The acquisition methods of the outer pool width W are specially：It is directly measured in the information in the pending region 5 outer
The width in molten bath 3, as outer pool width W.
The acquisition methods of the interior pool width N are specially：In directly being measured in the information in the pending region 5
The width in molten bath 2, as interior pool width N.
In the present embodiment, weld seam deviation e is measured from pending region 50And penetration signal p0.Specially：Institute
In experiment workpiece after stating pending region 5 and being tested per group welding, welded condition measurement amount is obtained, specially weld seam is inclined
Poor e0And penetration signal p0, and weld seam deviation e0And penetration signal p0As desired output signal.
The weld seam deviation e0Acquisition methods be specially：Electric arc letter is obtained from the information in the pending region 5
Breath and weld information, and obtain the bias of arc information and weld information, bias is weld seam deviation e0。
The penetration signal p0Acquisition methods be specially：It is measured from the experiment back of work after every group welding experiment molten
Saturating state p0。
In the present embodiment, for molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and welding current I and correspondence
Weld seam deviation e0And penetration signal p0Establish several groups sample.
In the present embodiment, weld seam deviation and penetration signal one neural network model are established.As shown in Fig. 2, weld seam deviation
And penetration signal one neural network model is configured to three layers, including input layer, hidden layer and output layer, with welding process spy
Property parameter, i.e., per group welding experiment obtain molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and welding current I
4 neurons as input layer；Hidden layer includes 30 neurons；With welded condition premeasuring, i.e. weld seam deviation e and molten
Saturating 2 neurons of the state p as output layer, and weld seam deviation e and penetration signal p are as real output signal.It is implicit
Layer is using tansig () as transmission function, as shown in formula (2)；Output layer with purelin () be transmission function, such as formula (3) institute
Show：
F (x)=x (3)
Weld seam deviation and penetration signal one neural network model to foundation are trained, and pass through training and iteration meter
It calculates, optimum weight coefficient is obtained, to obtain optimal weld seam deviation and penetration signal one neural network model.
In the present embodiment, assume initially that input layer shares M input signal, any one input signal indicates (m=with m
1,2,3,4)；Hidden layer shares I neuron, any one neuron indicated with i (i=1,2 ..., 30)；Output layer is shared J
A neuron, wherein any one neuron are indicated (j=1,2) with j.The connection weight W of input layer and hidden layermiIt indicates,
The connection weight W of hidden layer and output layerijIt indicates.
In addition, the input of setting neuron is indicated with u, excitation output indicates that the subscript expression layer of u, v, subscript are indicated with v
Neuron in layer, such asIndicate the input of i-th of neuron of hidden layer；F (x) is set as the transmission function of neuron；If
It is X=[x to determine input signal1,x2,x3,x4]T, output signal is set as Y=[y1,y2]T；Set desired output signal as：K=
[k1,k2]T, n is the number that iteration carries out.
The whole process of the training is：
a：The positive of input information is transmitted.Specially：Input signal is inputted from input layer, by hidden layer, in output layer
Real output signal is generated, connection weight of each layer network remains unchanged during this, such as formula (4), formula (5), formula (6) and formula
(7) shown in：
b：Error back propagation.In the present embodiment, error signal, the i.e. difference of real output signal and desired output signal,
Successively to input layer backpropagation since output layer, in communication process, the weight coefficient of neural network is automatically fed back
It corrects so that real output signal increasingly approaches desired output signal.
Specially：Define j-th of neuron of output layer error signal be：
The real output signal of j-th of neuron is：
Y=[y1,y2]T (9)
The desired output signal of j-th of neuron is：
The error energy function for defining neuron j is：
The error total energy function of all neurons of output layer is E, then：
Output layer derives as follows to hidden layer modified weight:
Wherein：
It obtains：
In formula (13), η is e-learning rate, Δ wij(n) it is the renewal amount of weights.Formula (14) is updated to formula (13), is obtained
Go out：
Δwij(n)=- η × ej(n)×vi H(n) (15)
The network weight coefficient update that output layer to the hidden layer at the (n+1)th moment can then be obtained is：
wij(n+1)=wij(n)+Δwij(n) (16)
The weights modification process of hidden layer to input layer derives as follows：
Wherein：
And because：
It obtains：
Formula (20), formula (21) are substituted into formula (19), can be obtained：
Formula (21) is substituted into formula (17), can be obtained：
Δwmi(n)=- η × f'(ui H(n))×ej(n)×f'(uj O(n))×wij×xm (23)
Weight coefficient to obtain the (n+1)th moment hidden layer to input layer is updated to：
wmi(n+1)=wmi(n)+Δwmi(n) (24)
The initial value value of each layer network weight coefficient is 1, by inputting training sample data, is transported by a series of iteration
It calculates, i.e., such as optimal weight coefficient is obtained from formula (4) to shown in formula (24), it is possible thereby to build optimal weld seam deviation and penetration
State one neural network model：
In formula (25), XT=[C, W, N, I] is that input vector is (i.e. molten wide W outside molten bath grey scale centre of gravity C, molten bath, molten in molten bath
Wide N and welding current I)；uHFor hidden layer input quantity；wHFor hidden layer weight coefficient；bHRepresent hidden layer threshold value；vHFor hidden layer
Output vector；uoFor output layer input quantity；woFor output layer weight coefficient；boFor output layer threshold value；Y=[e, p] represent output to
Amount, i.e. weld seam deviation e and molten bath penetration signal p；δ is compensation vector, is compensated to the modeling error of prediction model.Thus
The welding process of automobile can accurately predict and monitor by formula (25), realizes Automobile Welding automation.
In the present embodiment, optimal weld seam deviation and penetration signal one neural network model are verified.Specific step
Suddenly it is：
a：Other 400 pairs of input vectors are chosen, optimal weld seam deviation and penetration signal one neural network mould are input to
In type, the predicted value of corresponding weld seam deviation and the predicted value of penetration signal are obtained；Weld seam deviation is obtained by practical measurement
Actual value and penetration signal actual value.
b：The actual value of predicted value and weld seam deviation to weld seam deviation carries out figure point distributional analysis, and it is inclined to calculate weld seam
Average value, absolute error and the relative error of difference.
As shown in figure 3, the dot in figure represents, dot represents the actual value of weld seam deviation, and it is inclined that bold portion represents weld seam
The predicted value of difference.As seen from the figure, dot is substantially distributed in the both sides of bold portion, and fluctuating range is little, illustrates model
Weld seam deviation prediction case it is ideal, the predicted value of part weld seam deviation and the measured value of weld seam deviation are as shown in table 1.Weld seam is inclined
The average value of difference is 0.011mm, and the wherein average value of weld seam deviation is defined as under formula (26)：
In formula (26), N is sampled point number, and y is the actual value of weld seam deviation, and y' is the predicted value of weld seam deviation, relatively
Error be absolute error divided by weld seam deviation actual value after take the value that percentage obtains, the part that step a and step b are obtained
Data are as shown in the table：
c：The actual value of predicted value and penetration signal to penetration signal carries out figure point distributional analysis, and calculates accurately
Rate.
As shown in figure 4, figure orbicular spot represents the actual value of penetration signal, asterisk part represents the verification situation of penetration signal
The predicted value of penetration signal.As seen from Figure 4, in most data the actual value of penetration signal and penetration signal prediction
Value is identical, and a small number of data predictions are inaccurate, and final accuracy rate is 95%, illustrates that model has certain accuracy.Its
Middle model accuracy rate is defined as follows：
In formula (27), N is test point number, and N=400 in the present embodiment, n are prediction correctly point number, the present embodiment
Middle n=380.
In the present embodiment, during obtaining crater image, judge electric arc 1 whether be aligned weld seam 4 the specific steps are：
S1.1：Neutral dim light processing is added to obtain bath image information by CCD camera units and narrow-band-filter；
S1.2：The intercepting process area information in the bath image information, by measure and calculation processing area information,
Obtain 4 ambient heat state of gray uniformization, current conditions and weld seam；
S1.3：If entire weldment is in a kind of state of balance, as shown in figure 8, being specially：The shape of molten bath and electric arc 1
State rule, uniform gray level, electric current be steady and 4 ambient heat beinthebalancestate of weld seam, then may determine that and be directed at weld seam for electric arc 1
4, the molten bath intensity profile almost symmetry of 4 left and right sides of weld seam, such welded condition disclosure satisfy that technological requirement at this time；
S1.4：If entire weldment is in a kind of unbalanced state, as shown in figure 9, being specially：Molten bath and electric arc 1
Form is irregular, gray scale is uneven, electric current is unstable and weld seam 4 ambient heat is uneven, then may determine that for electric arc 1 it is not right
Quasi- weld seam 4, needs in time to be adjusted electric arc 1 the phenomenon that will appear burn-through 6 at this time.
Above step can be monitored to whether electric arc 1 is directed at weld seam 4 during the welding process, ensure that electric arc 1 can
It is directed at weld seam 4, the case where generating burn-through is reduced and occurs.
In conclusion the purpose of the present invention is to provide a kind of weld seam deviations for Automobile Welding and penetration signal to monitor
Method is guiding with the information of the crater image of acquisition, forms the method for welding process monitoring of Multi-information acquisition.Pass through
Visual sensor and the collected information of current sensor, extraction influence the welding process characteristic of weld seam deviation and penetration signal
Parameter, and the welded condition measurement amount of entire welding process is extracted, it constructs and integrates welding deviation and penetration signal prediction
Weld seam deviation and penetration signal one neural network model, ensure welding process in realize electric arc be aligned weld seam, and realize
The effect of joint penetration is predicted and is monitored to entire welding process, realizes Automobile Welding process automation.
Finally it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than the present invention is protected
The limitation of range is protected, although being explained in detail to the present invention with reference to preferred embodiment, those skilled in the art answer
Work as understanding, the technical solution of invention can be modified or replaced equivalently, without departing from the essence of technical solution of the present invention
And range.
Claims (4)
1. a kind of weld seam deviation and penetration signal monitoring method for Automobile Welding, it is characterised in that：Include the following steps：
S1：Crater image is obtained, and intercepts pending region；
S2：Information is extracted from pending region；
S3：Using molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and the welding current I in information as welding process
Characteristic parameters are input to weld seam deviation and penetration signal one neural network model, weld seam deviation e and penetration shape are calculated
State p；
S4：Welding process is predicted by weld seam deviation e and penetration signal p；
The acquisition methods of weld seam deviation and penetration signal one neural network model are specially：
S3.1：Crater image information is obtained, and intercepts pending region；
S3.2：Information is extracted from pending region, for information as welding process characteristic parameters, described information includes molten bath ash
Spend center of gravity C, outer pool width W, interior pool width N and welding current I；
S3.3：Weld seam deviation e is measured from pending region0And penetration signal p0；
S3.4：For molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and welding current I and corresponding weld seam deviation
e0And penetration signal p0Establish several groups sample；
S3.5：Establish weld seam deviation and penetration signal one neural network model：Wherein, the weld seam deviation and penetration signal one
The input layer of somatic nerves network model includes molten bath grey scale centre of gravity C, outer pool width W, interior pool width N and welding current I totally 4
A neuron, output layer include weld seam deviation e and penetration signal p totally 2 neurons, and hidden layer includes 20 neurons, hidden
Containing layer using tansig () as transmission function, output layer with purelin () for transmission function, and to weld seam deviation and penetration shape
State one neural network model is trained.
2. a kind of weld seam deviation and penetration signal monitoring method for Automobile Welding according to claim 1, feature
It is：Values of the molten bath grey scale centre of gravity C on the directions i be：Wherein, i, j are the pending area
The both direction in domain, K, L are respectively i, and the quantity of the directions j pixel, g (i, j) is the gray value at pixel (i, j).
3. a kind of weld seam deviation and penetration signal monitoring method for Automobile Welding according to claim 1, feature
It is：The processing method for obtaining crater image information is to add neutral dim light to handle using narrow-band-filter.
4. a kind of weld seam deviation and penetration signal monitoring method for Automobile Welding according to claim 1, feature
It is：During obtaining crater image, judge electric arc whether be aligned weld seam the specific steps are：
S1.1：Neutral dim light processing is added to obtain bath image information by CCD camera units and narrow-band-filter；
S1.2：The intercepting process area information in the bath image information is obtained by measurement and calculation processing area information
Gray uniformization, current conditions and weld seam ambient heat state；
S1.3：If entire weldment is in a kind of state of balance, specially：Molten bath and arc shape rule, uniform gray level, electricity
Levelling is steady and weld seam ambient heat beinthebalancestate, then may determine that and be directed at weld seam for electric arc；
S1.4：If entire weldment is in a kind of unbalanced state, specially：Molten bath and arc shape be irregular, gray scale not
Uniformly, electric current is unstable and weld seam ambient heat is uneven, then may determine that as electric arc misalignment weld seam.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611267754.9A CN106624266B (en) | 2016-12-31 | 2016-12-31 | A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611267754.9A CN106624266B (en) | 2016-12-31 | 2016-12-31 | A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106624266A CN106624266A (en) | 2017-05-10 |
| CN106624266B true CN106624266B (en) | 2018-08-07 |
Family
ID=58839013
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611267754.9A Active CN106624266B (en) | 2016-12-31 | 2016-12-31 | A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106624266B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108500498B (en) * | 2018-03-26 | 2020-05-19 | 华中科技大学 | Weld joint forming quality monitoring method |
| CN110935983B (en) * | 2018-09-21 | 2021-07-20 | 天津大学 | Method for controlling welding penetration by utilizing reflected laser stripe image |
| CN111061231B (en) * | 2019-11-29 | 2021-11-30 | 上海交通大学 | Weld assembly gap and misalignment feed-forward molten pool monitoring system and penetration monitoring method |
| CN112381095B (en) * | 2021-01-15 | 2021-05-28 | 南京理工大学 | Electric arc additive manufacturing layer width active disturbance rejection control method based on deep learning |
| CN113828947B (en) * | 2021-11-23 | 2022-03-08 | 昆山宝锦激光拼焊有限公司 | BP neural network laser welding seam forming prediction method based on double optimization |
| CN114799600B (en) * | 2022-05-13 | 2024-03-15 | 中车工业研究院有限公司 | Method, equipment, system, medium and product for controlling melting width |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1546268A (en) * | 2003-12-17 | 2004-11-17 | 南昌大学 | Intelligent control system for weld seam tracking and fusion penetration in spiral pipes |
| CN103264216A (en) * | 2013-05-15 | 2013-08-28 | 山东大学 | Controlled perforation plasma-arc welding system and process on basis of back small-hole visual inspection |
| CN103521890A (en) * | 2013-10-12 | 2014-01-22 | 王晓宇 | Device and method for double-faced double-arc vertical welding near-infrared vision sensing and penetration control |
| CN103567606A (en) * | 2013-10-18 | 2014-02-12 | 湘潭大学 | Automatic welding control method and system based on dual-mode real-time welding seam tracking |
| DE102013215362A1 (en) * | 2013-08-05 | 2015-02-05 | Trumpf Werkzeugmaschinen Gmbh + Co. Kg | Method and device for determining a welding depth during laser welding |
| CN104551347A (en) * | 2014-12-30 | 2015-04-29 | 江苏科技大学 | Infrared vision sensing detection method and device for narrow-gap weld seam deviation |
| CN105478976A (en) * | 2016-01-26 | 2016-04-13 | 清华大学 | Edge micro-plasma arc welding forming control method based on identification of dynamical system |
| CN105499772A (en) * | 2016-01-26 | 2016-04-20 | 清华大学 | Control system for micro-beam plasma welding formation of thin-wall slit circular longitudinal seam |
-
2016
- 2016-12-31 CN CN201611267754.9A patent/CN106624266B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1546268A (en) * | 2003-12-17 | 2004-11-17 | 南昌大学 | Intelligent control system for weld seam tracking and fusion penetration in spiral pipes |
| CN103264216A (en) * | 2013-05-15 | 2013-08-28 | 山东大学 | Controlled perforation plasma-arc welding system and process on basis of back small-hole visual inspection |
| DE102013215362A1 (en) * | 2013-08-05 | 2015-02-05 | Trumpf Werkzeugmaschinen Gmbh + Co. Kg | Method and device for determining a welding depth during laser welding |
| CN103521890A (en) * | 2013-10-12 | 2014-01-22 | 王晓宇 | Device and method for double-faced double-arc vertical welding near-infrared vision sensing and penetration control |
| CN103567606A (en) * | 2013-10-18 | 2014-02-12 | 湘潭大学 | Automatic welding control method and system based on dual-mode real-time welding seam tracking |
| CN104551347A (en) * | 2014-12-30 | 2015-04-29 | 江苏科技大学 | Infrared vision sensing detection method and device for narrow-gap weld seam deviation |
| CN105478976A (en) * | 2016-01-26 | 2016-04-13 | 清华大学 | Edge micro-plasma arc welding forming control method based on identification of dynamical system |
| CN105499772A (en) * | 2016-01-26 | 2016-04-20 | 清华大学 | Control system for micro-beam plasma welding formation of thin-wall slit circular longitudinal seam |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106624266A (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106624266B (en) | A kind of weld seam deviation and penetration signal monitoring method for Automobile Welding | |
| Naik et al. | Optimization of tensile strength in TIG welding using the Taguchi method and analysis of variance (ANOVA) | |
| Pal et al. | Artificial neural network modeling of weld joint strength prediction of a pulsed metal inert gas welding process using arc signals | |
| Wu et al. | Online monitoring and model-free adaptive control of weld penetration in VPPAW based on extreme learning machine | |
| Nomura et al. | Burn-through prediction and weld depth estimation by deep learning model monitoring the molten pool in gas metal arc welding with gap fluctuation | |
| Liu et al. | Control of human arm movement in machine-human cooperative welding process | |
| CN107081503A (en) | The infrared nondestructive detection device and its Infrared Non-destructive Testing method of a kind of arc-welding quality | |
| CN107855687A (en) | A kind of increasing material manufacturing fusion penetration on-line checking and control method and system | |
| Sarkar et al. | Machine learning method to predict and analyse transient temperature in submerged arc welding | |
| Gyasi et al. | Modeling of an artificial intelligence system to predict structural integrity in robotic GMAW of UHSS fillet welded joints | |
| Yu et al. | Identification of butt welded joint penetration based on infrared thermal imaging | |
| Liu et al. | Prediction of welding residual stress and deformation in electro-gas welding using artificial neural network | |
| Li et al. | Penetration depth monitoring and control in submerged arc welding | |
| Barot et al. | Process monitoring and internet of things feasibility for submerged arc welding: State of art | |
| Zhang et al. | Adaptive control for laser welding with filler wire of marine high strength steel with tight butt joints for large structures | |
| Dhas et al. | Modeling and prediction of HAZ using finite element and neural network modeling | |
| Gyasi et al. | Digitalized automated welding systems for weld quality predictions and reliability | |
| Zhang et al. | Intelligent control of pulsed GTAW with filler metal | |
| Wang | Three-dimensional vision applications in GTAW process modeling and control | |
| CN113478082B (en) | Flexible laser welding method and device for skin-stringer | |
| Chen et al. | Computer vision sensing and intelligent control of welding pool dynamics | |
| Sterjovski et al. | Artificial neural networks for predicting diffusible hydrogen content and cracking susceptibility in rutile flux-cored arc welds | |
| Węglowski | Investigation on the arc light spectrum in GTA welding | |
| CN206764093U (en) | A kind of infrared nondestructive detection device of arc-welding quality | |
| Einerson et al. | Development of an intelligent system for cooling rate and fill control in GMAW.[Gas Metal Arc Welding (GMAW)] |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |