A process method for large-angle forming of sheet metal based on multiple scanning

Abstract

The present invention discloses a process method for large-angle forming of sheet materials based on multiple scanning, comprising the following steps: Step 1: pre-set a set of process parameters for bending and forming processing, record the bending angle of the processed sheet material as the target angle, and perform rough processing with the target angle minus the safety amount as the first target amount; Step 2: subtract the current measured processing amount from the target angle as the second target amount, and re-determine a set of process parameters based on the second target amount, and accurately compensate the second target amount through a single scan. The present invention incorporates the steps of real-time measurement and feedback. By timely adjusting the processing parameters of the margin, the accuracy of the bending processing can be greatly improved, and the requirements of high-end application scenarios can be better met; the entire process of the present invention has relatively small requirements on the scale of the experimental database. When the scheme is applied to new materials and new sizes of sheets, only a small amount of experimental accumulation is required to complete the required neural network training.

Description

Technical Field

[0001] The invention relates to the technical field of laser bending and forming, and in particular to a process method for large-angle forming of sheet materials based on multiple scanning processes. Background Art

[0002] As the main load-bearing components of aircraft, missiles and other aircraft, the wing rudder surface requires the skin to be fitted on the skeleton / honeycomb core with high precision. At present, the bending line of the wing rudder skin is formed by cold bending with a bending machine. This forming method has problems such as low material utilization, low production efficiency and low forming accuracy.

[0003] (1) Titanium alloy materials have a high yield strength ratio (yield strength / tensile strength), a small plastic deformation range, poor plastic deformation during cold forming, and easy cracking in the bending area during cold bending. Therefore, it is difficult to obtain structural parts that meet the surface and size requirements when titanium alloy sheets are cold stamped;

[0004] (2) The titanium alloy sheet has a large springback value (the ratio of yield limit to elastic modulus) after forming, which causes a large springback after forming. At present, at the end of cold bending, springback is easily generated due to the recovery of elastic deformation, which directly affects the dimensional accuracy and the fit between the subsequent skin and the frame;

[0005] (3) Since the bending line of the wing and rudder skin is not parallel to the working edge line, the laser cutting process currently cuts out a margin area parallel to the bending line for each bending line to ensure that the bending machine has a positioning function when bending the skin, which greatly reduces the material utilization rate;

[0006] (4) The current process flow for the production of rudder skins has a low production efficiency. Firstly, the traditional process of cold bending the skins increases the grinding time and processing time. Secondly, the current process flow requires dispatching from the welding group to the sheet metal group and then to the fitter group. The processes are relatively scattered and the dispatching time is increased. These two reasons lead to the low production efficiency of the rudder skins and the process flow needs to be further optimized.

[0007] The bending angle of wing and rudder skin is generally above 5°, and multiple laser scans are required. Therefore, the processing parameters for each scan must be planned. At the same time, in order to achieve real-time optimization of process parameters between multiple scans, it is necessary to quickly measure the geometric features of the workpiece after each scan in real time. Therefore, it is necessary to design a set of planning methods that can universally determine the number and sequence of processing and measurement processes for plates with different processing requirements, as well as determine the processing parameters of each processing step.

[0008] At present, there are some basic achievements in the research field of laser bending processing methods. For example, in patent CN201710439402.5, neural network technology and robotic devices are used to design a basic single laser scanning bending forming processing method. However, this solution requires further design planning when facing the problem of large-angle sheet bending that requires multiple laser scanning processes. There are also some method results for the precision control of laser bending forming. For example, patent CN202111583791.1 adopts a method of laser bending forming by alternating upper and lower layers to calibrate the elastic recovery deformation of the bending area in laser multi-pass scanning bending forming. However, in the bending processing problem of multiple scans on the same bending line of the sheet, there is currently a lack of method results focusing on precision control. Summary of the invention

[0009] The purpose of the present invention is to provide a process method for large-angle forming of sheet metal based on multiple scanning processes. The bending forming process is first performed using relatively rough process parameters, and the bending angle of the sheet metal to be processed is processed to an angle slightly smaller than the target bending angle. The excess is converted into an equivalent small angle through the designed angle conversion algorithm and small-angle process planning algorithm, and then a new set of process parameters is determined to accurately compensate for the excess through a single scan.

[0010] The technical problems solved by the present invention are:

[0011] (1) During the laser bending process of large-angle sheet metal forming, the sheet metal needs to be laser scanned multiple times. For different target bending angles, the optimal number of processing times and the processing parameters for each processing need to be planned;

[0012] (2) Large-angle laser bending of sheet metal requires real-time measurement to control processing accuracy, so a feedback method needs to be planned to effectively reflect the real-time measured bending angle results in the changes in processing parameters;

[0013] (3) Laser bending process planning is difficult to establish based on limited experimental data. Therefore, it is necessary to design the neural network training content and the composition of the relevant experimental database to provide high-precision support for the planning method with the smallest possible experimental database.

[0014] The purpose of the present invention can be achieved by the following technical solutions:

[0015] A process method for large-angle forming of sheet metal based on multiple scanning processes, comprising the following steps:

[0016] Step 1: Pre-set a set of process parameters for bending and forming processing, record the bending angle of the processed plate as the target angle, and perform rough processing with the target angle minus the safety amount as the first target amount;

[0017] Step 2: The target angle minus the currently measured processing amount is used as the second target amount. A set of process parameters is re-determined based on the second target amount, and the second target amount is accurately compensated through a single scan.

[0018] As a further solution of the present invention: Step 1 comprises the following steps:

[0019] W1: Subtract a safety value from the target angle α, and then divide it by an appropriate number n to obtain a reasonable single-scan bending angle α0;

[0020] W2: Through the single-scan process planning algorithm (P, v) = f(α0), a set of process parameters is obtained and n bending processes are performed.

[0021] As a further solution of the present invention: In W2, a set of process parameters are obtained by a single scan process planning algorithm (P, v) = f(α0):

[0022] W21: Get output: Substitute the target bending angle α0, take a series of equidistant values ​​for P and v, input them into the corresponding network, obtain the corresponding output parameter values, and sort and merge all the corresponding P and v data;

[0023] W22: Based on the minimum linear energy density P / v, select a set of optimal P and v values ​​as the output result.

[0024] As a further solution of the present invention: the process planning algorithm in W2 includes:

[0025] S1: Database establishment: Through experimental methods, a series of P and v values ​​are taken, and single scanning processing is performed on the same model and the same material plate to obtain the corresponding α and establish the first database;

[0026] S2: Network training: Based on the first database, two neural networks are trained, one training network takes α and P as input and v as output, and the other training network takes α and v as input and P as output.

[0027] As a further solution of the present invention: Step 2 comprises the following steps:

[0028] By real-time measurement of α 实 , obtain the actual bending angle and find the margin Δα; by training the angle conversion neural network α1=g(α 实 , Δα, n), calculate and obtain the output, and obtain the remaining machining process parameters through (P1, v1) = f(α1).

[0029] As a further solution of the present invention: the conversion neural network includes:

[0030] By using multiple sets of P and v parameter values, each set of process parameters is repeatedly processed for more than 10 times on the same material and the same model of plate, and the bending angle after each processing is recorded to establish a second database;

[0031] According to the second database, the angle conversion neural network α1=g(α 实 , Δα, n).

[0032] Beneficial effects of the present invention:

[0033] (1) The present invention plans a multi-scanning process planning method with strong versatility, which can be applied to the processing of different bending angles of plates of various material specifications;

[0034] (2) The present invention incorporates the steps of real-time measurement and feedback. By timely adjusting the processing parameters of the allowance, the accuracy of the bending process can be greatly improved, which better meets the requirements of high-end application scenarios;

[0035] (3) The entire process of the present invention has relatively small requirements for the scale of the experimental database. When the scheme is applied to new materials and new sizes of plates, only a small amount of experimental accumulation is needed to complete the training of the required neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention will be further described below in conjunction with the accompanying drawings.

[0037] Figure 1 It is a schematic diagram of a method for planning a multi-scan process for laser bending processing according to the present invention;

[0038] Figure 2 It is a single scan process parameter planning flow chart of the present invention;

[0039] Figure 3 It is a flow chart of multiple scanning process planning of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] See also Figure 1-Figure 3 As shown, the present invention is a process method for large-angle forming of sheet metal based on multiple scanning, comprising the following steps:

[0042] Step 1: Pre-set a set of process parameters for bending and forming, and process the bending angle of the plate to be processed to an angle slightly smaller than the target bending angle;

[0043] Step 2: Re-determine a set of process parameters and accurately compensate the margin through a single scan;

[0044] The specific steps include:

[0045] W1: Subtract a safety value from the target angle α, and then divide it by an appropriate number n to obtain a reasonable single-scan bending angle α0;

[0046] W2: Obtain a set of process parameters through a single scan process planning algorithm (P, v) = f(α0); use the process parameters to process n times,

[0047] W3: By measuring α in real time 实 , obtain the actual bending angle and find the margin Δα;

[0048] W4: By using multiple sets of P and v parameter values, each set of process parameters is repeatedly processed for more than 10 times on the same material and the same model of plate, and the bending angle after each processing is recorded to establish a database;

[0049] According to the database, the angle conversion neural network α1=g(α 实 , Δα, n), the output is obtained through calculation, and the allowance machining process parameters are obtained through (P1, v1) = f(α1).

[0050] The single scanning process inputs the target bending angle α and outputs the corresponding optimal process parameters, including laser power P and scanning speed v;

[0051] The specific steps of the single scan process are as follows:

[0052] S1: Database establishment: Through experimental methods, a series of P and v values ​​are taken, and single scanning processing is performed on the same model and the same material plate to obtain the corresponding α and establish the database;

[0053] S2: Network training: Based on the database, two neural networks are trained. One training network takes α and P as input and v as output, and the other training network takes α and v as input and P as output.

[0054] S3: Get output: Substitute the target bending angle α0, take a series of equidistant values ​​for P and v, input them into the corresponding network, obtain the corresponding output parameter values, and sort and merge all the corresponding P and v data;

[0055] S4: Based on the minimum linear energy density P / v, select a set of optimal P and v values ​​as the output result.

[0056] In one specific embodiment:

[0057] (1) Determine the target bending angle of the sheet metal to be 7.68°, the safety margin to be 0.1°, and the number of scans to be 7, then:

[0058] (2) Based on the existing experimental data in Table 1, a single-scan process parameter planning algorithm is established.

[0059] Table 1 Experimental data of laser bending single scanning processing

[0060]

[0061] Using this database, we train networks with α and V as input and P as output, and α and P as input and V as output. Substitute α0 = 1.08°, v = (10, 20, ..., 100) and P = (20, 40, ..., 200) into the corresponding neural network, and obtain the results shown in Table 2:

[0062] Table 2 Output results of single scan process planning for α0;

[0063]

[0064] The optimal output is selected according to the minimum linear energy density, and P = 202.96W, v = 100mm / s are taken as the first 7 processing parameters;

[0065] (3) Perform processing seven times according to the process parameters determined in the previous step, and measure the actual bending angle after processing. For example, if the actual bending angle is 7.00° after seven scans, the margin Δα is 0.68°.

[0066] (4) Training angle conversion network α1 = g(α 实 , Δα, n), since the experimental database cannot be obtained, the custom relationship Δα=α1*(1-0.025α 实 );

[0067] In this way, a database is established and the network is trained. 实 , Δα, n), substitute α 实 =7°, Δα=0.68°, n=7, and the output α1=0.7916° is obtained;

[0068] The single scan process planning algorithm trained in step (2) is used again, and =0.7916° is substituted into v=(10,20,…,100) and P=(20,40,…,200) in the corresponding neural network. The results are shown in Table 3:

[0069] Table 3 Output results of single scan process planning for α1

[0070]

[0071] According to the minimum linear energy density, the optimal process parameters are selected as P=169.91W, v=100mm / s.

[0072] To summarize the entire demonstration process, the process of bending at an angle of α = 7.68° based on the allowance compensation method is as follows:

[0073] (1) Scan 7 times with parameters of P = 202.96W, v = 100mm / s;

[0074] (2) The bending angle was measured to be 7.00°;

[0075] (3) Perform a compensation process with P = 169.91W, v = 100mm / s.

[0076] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A process method for large-angle forming of sheet metal based on multiple scanning, characterized in that: The following steps are involved: Step 1: Pre-set a set of process parameters for bending and forming processing, record the bending angle of the processed plate as the target angle, and perform rough processing with the target angle minus the safety amount as the first target amount: The steps include: W1: Set the target angle Subtract a safety amount and divide by an appropriate number n to obtain a reasonable single scan bending angle. ; W2: Process planning algorithm via single scan , obtain a set of process parameters and perform n bending processes; Step 2: The target angle minus the currently measured processing amount is used as the second target amount, and a set of process parameters is re-determined based on the second target amount, and the second target amount is accurately compensated through a single scan; The steps include: Through real-time measurement , get the actual bending angle and find the allowance ; Transforming Neural Networks by Training Perspective , calculate the output, and pass Obtain the remaining machining process parameters; The conversion neural network consists of: By using multiple sets of P and v parameter values, each set of process parameters is repeatedly processed for more than 10 times on the same material and the same model of plate, and the bending angle after each processing is recorded to establish a second database; According to the second database, the angle conversion neural network is trained based on the BP neural network principle =g( , , n), that is, , and n is the input layer, is the output layer.

2. According to claim 1, a process method for large-angle forming of sheet metal based on multiple scanning, characterized in that: In W2, a single scan process planning algorithm is used , the process of obtaining a set of process parameters is: W21: Get output: Substitute the target bending angle , take a series of equidistant values ​​for P and v, input them into the corresponding networks respectively, obtain the corresponding output parameter values, and sort and merge all the corresponding P and v data; W22: Based on the minimum linear energy density P / v, select a set of optimal P and v values ​​as the output result.

3. According to the process method of claim 1 for large-angle forming of sheet metal based on multiple scanning, it is characterized in that: The process planning algorithms in W2 include: S1: Database establishment: Through experimental methods, a series of P and v values ​​are taken, and the same model and the same material plate are scanned and processed once to obtain the corresponding , establish the first database; S2: Network training: Based on the first database, two BP neural networks are trained. One training network is , P is the input layer, v is the output layer, and another training network is , v is the input layer, and P is the output layer.

Images