A pipe fitting processing die and its design optimization method

Abstract

The present invention belongs to the technical field of injection molds, and specifically relates to a pipe fitting processing mold and its design optimization method. This elbow injection mold includes a base, on which a lower mold base is installed. Multiple lower cavities are arranged on the lower mold base. An elevating plate is arranged above the base, and an upper mold base is installed at the bottom of the elevating plate. Multiple upper cavities are formed on the upper mold base. Core component assemblies are installed at both ends of the lower cavity. A core pulling cooperation assembly for core pulling of the core component assemblies is installed on the elevating plate. A ejector component is installed on the base, avoiding the need for a complex gear transmission structure for core pulling, with simple core pulling. Moreover, when the mold is opened, the core component assemblies can automatically extend into the interior of the cavity, and when the mold is opened, the core component assemblies can automatically disengage from the cavity. In addition, this elbow injection mold realizes that when the mold is opened, several ejector pins automatically eject the injection molded elbow pipe fittings, and when the mold is closed, the several ejector pins can automatically move downward to reset.

Description

Technical Field

[0001] The invention relates to the technical field of injection molds, in particular to a pipe processing mold and a design optimization method thereof. Background Art

[0002] Plastic pipes are developing steadily, and various pipe fittings are also constantly developing and innovating. Plastic products of various shapes and shapes continue to appear, and the mold structure continues to become more complex, and the requirements for mold design are also constantly increasing.

[0003] In the design of plastic injection molding molds for elbows and pipe fittings, we often encounter some plastic parts that require lateral arc core pulling. For the molding of such plastic parts, two straight tubes are generally connected through a small arc transition to form angles of different angles. The 90° semicircular elbow cannot be demolded using the traditional core pulling method, and complex mechanisms such as gear rack arc core pulling are often required. If a gear rack arc core pulling mechanism is used, the core pulling mechanism is very complicated, and even several levels of transmission gear sets must be designed. In order to change the transmission direction, in addition to using cylindrical spur gears, it is sometimes necessary to design a bevel gear transmission. The mold structure is complex and the production cost is high. In order to facilitate core pulling during the injection molding of the elbow, we propose a pipe fitting processing mold. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a pipe processing mold and a design optimization method thereof. The pipe bending injection mold avoids the need for a complex gear transmission structure for core pulling, and the core pulling is simple. When the mold is opened, the core body assembly can automatically extend into the interior of the cavity, and when the mold is opened, the core body assembly can automatically detach from the cavity. In addition, the pipe bending injection mold can realize that when the mold is opened, a number of ejectors automatically eject the injection-molded bent pipe parts, and when the mold is closed, a number of ejectors can automatically move downward and reset, thereby solving the background problem.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] A pipe processing mold comprises a base, a lower mold base is installed on the base, a plurality of lower cavities are arranged on the lower mold base, a lifting plate is arranged above the base, an upper mold base is installed at the bottom of the lifting plate, a plurality of upper cavities are provided on the upper mold base, core body components are installed at both ends of the lower cavity, a core pulling matching component for core pulling of the core body component is installed on the lifting plate, and a ejecting component is installed on the base.

[0007] As a further aspect of the present application, the core component includes a core column. An inclined surface is provided on the core column, and a plurality of groups of sliding grooves are provided on the inclined surface. Sliders are slidably connected to the sliding grooves, and a bending member is commonly connected to the plurality of groups of sliders. A sliding rod is fixedly connected to the slider, and the sliding rod slidably penetrates through the core column. An installation block is fixedly connected to the end of the sliding rod. A first spring is sleeved on the outer surface of the sliding rod. One end of the first spring is fixedly connected to the core column, and the other end of the first spring is fixedly connected to the installation block.

[0008] As a further aspect of the present application, the core-pulling and mating component includes a guide rail. The guide rail is fixedly installed on the lower die base, and a moving seat is slidably installed on the guide rail. A plurality of groups of inclined holes are provided on the moving seat, and mating rods slidably penetrate through the inclined holes. The top of the mating rod is fixedly connected to the lifting plate, and the moving seat is fixedly connected to the core column through a plurality of connecting rods.

[0009] As a further aspect of the present application, a first cooling flow channel is provided on the core column, and a second cooling flow channel is provided on the bending member. Four corresponding two groups of first cooling flow channels and two groups of second cooling flow channels form an internal coolant flow channel.

[0010] As a further aspect of the present application, a plurality of groups of external cooling flow channels are provided on both the upper die base and the lower die base.

[0011] As a further aspect of the present application, semi-circular grooves are provided at both ends of the lower cavity and both ends of the upper cavity, and a sealing ring block is fixedly sleeved on the outer surface of the core column.

[0012] As a further aspect of the present application, the ejector component includes moving plates on the left and right sides. Fixed blocks are fixedly connected to the left and right end faces of the lower die base. A plurality of moving grooves are provided on the fixed blocks, and moving blocks are slidably connected to the moving grooves. The moving blocks are fixedly connected to the moving plates. A connecting plate is fixedly connected between the two groups of moving plates. A first inclined guide block is fixedly connected to the connecting plate. A plurality of ejector pins slidably penetrate through the lower die base. The bottoms of the plurality of ejector pins are fixedly connected to an installation plate. A second spring is sleeved on the outer surface of the ejector pin. One end of the second spring is fixedly connected to the bottom of the lower die base, and the other end of the second spring is fixedly connected to the installation plate. A second inclined guide block is fixedly connected to the installation plate. A vertical sliding opening and an inclined sliding opening are provided on the moving plate. The vertical sliding opening is communicated with the inclined sliding opening. Fixed columns are fixedly connected to the left and right end faces of the lifting plate, and the fixed columns slidably penetrate through the vertical sliding opening.

[0013] As a further aspect of the present application, a plurality of positioning columns are fixedly connected to the bottom of the lifting plate, and a plurality of positioning holes are provided on the base.

[0014] A design optimization method for a pipe fitting processing mold of the present invention models and optimizes the mold processing process through a deep learning algorithm to achieve automatic adjustment of processing parameters. The specific process is as follows: Step 1: Data collection; Step 2: Data preprocessing; Step 3: Model establishment and training; Step 4: Model deployment.

[0015] The present invention provides a pipe fitting processing mold. Compared with the prior art, it has the following beneficial effects:

[0016] 1. For this pipe fitting processing mold, the inner core of the bent pipe is composed of two sets of bending parts and two sets of core columns. When the mold is opened, under the action of the first spring, the bending parts will remain inside the cavity, and the core columns will first disengage from the cavity. During the process of the core columns disengaging, under the action of the first spring, the bending parts will slide along the inclined surfaces on the core columns, and the bending parts will move away from the bent part of the cavity. The core columns continue to move outwards, and the bending parts will disengage from the cavity.

[0017] 2. For this pipe fitting processing mold, through the cooperation rod, when the mold is closed, the core body assembly enters the inside of the cavity. When the mold is opened, the core body assembly disengages from the cavity, avoiding the need to add external driving electrical components. Moreover, when the core body assembly is assembled or disassembled, the structure is simple, avoiding the need to design a complex gear transmission structure.

[0018] 3. For this pipe fitting processing mold, through the cooperation of the moving plate, when the mold is opened, the ejector pins will automatically eject. When the mold is closed, the fixed column will slide along the vertical sliding opening. At this time, under the action of the second spring, several ejector pins will move downward to reset, realizing that when the mold is closed, several ejector pins automatically reset.

[0019] 4. This pipe fitting processing mold introduces an adaptive learning system, which uses a deep learning algorithm to optimize processing parameters in real time, improving processing accuracy and efficiency. Specifically, based on real-time acquisition of sensor data, the model can automatically predict and adjust the best positions and forces of the moving plate and the ejector pins to ensure high-quality processing effects under different processing conditions. In addition, the system can self-update and optimize according to historical data and real-time data, thereby continuously improving the processing ability of the mold, reducing debugging time and costs, and improving production stability and consistency. Through this intelligent control method, the dependence on manual intervention is greatly reduced, and the automation level of the entire processing process is improved.

[0020] This pipe fitting processing mold achieves an efficient processing process through a variety of innovative designs. First, an inner core structure composed of two sets of bending parts and a core column, combined with the action of the first spring, ensures that the bending parts and the core column are sequentially separated from the cavity when the mold is opened, improving the demoulding efficiency of the mold. Second, through the use of a mating rod, the core body assembly enters the cavity when the mold is closed and separates from the cavity when the mold is opened, simplifying the structural design and avoiding complex gear transmissions and external drive electrical components. Third, the cooperation of the moving plate and the fixed column enables the ejector pin to automatically eject when the mold is opened and reset under the action of the second spring when the mold is closed, ensuring the automated operation of the mold. On this basis, combined with an adaptive learning system, deep learning algorithms are used to optimize the processing parameters in real time, improving the processing accuracy and efficiency. This system can obtain sensor data in real time, automatically predict and adjust the optimal positions and forces of the moving plate and the ejector pin, and self-update and optimize according to historical data and real-time data, significantly reducing the debugging time and cost, and enhancing the production stability and consistency. Description of the Drawings

[0021] Figure 1 It is a front view structural schematic diagram of the main body of the present invention;

[0022] Figure 2 It is a rear view structural schematic diagram of the main body of the present invention;

[0023] Figure 3 It is a side view structural schematic diagram of the main body of the present invention;

[0024] Figure 4 It is a schematic diagram of the upper mold base of the present invention;

[0025] Figure 5 It is a schematic diagram of the external cooling flow channel structure inside the upper mold base of the present invention;

[0026] Figure 6 It is a schematic diagram of the ejector assembly structure of the present invention;

[0027] Figure 7 It is a schematic diagram of the core body assembly structure when the mold of the present invention is closed;

[0028] Figure 8 It is a schematic diagram of the internal coolant flow channel structure of the present invention;

[0029] Figure 9 It is a schematic diagram of the core body assembly structure of the present invention;

[0030] Figure 10 It is a smart adaptive pipe fitting processing control flow chart of the present invention.

[0031] In the figure: 1, base; 2, fixing block; 3, moving plate; 4, guide rail; 5, moving seat; 6, mating rod; 7, lifting plate; 8, upper die holder; 9, positioning post; 10, positioning hole; 11, lower die holder; 12, external cooling channel; 13, lower cavity; 14, bending part; 15, core post; 16, upper cavity; 17, second spring; 18, connecting plate; 19, second inclined guide block; 20, first inclined guide block; 21, ejector pin; 22, fixing post; 23, semi-circular groove; 24, mounting plate; 25, vertical sliding opening; 26, inclined sliding opening; 27, sealing ring block; 28, connecting rod; 29, mounting block; 30, internal coolant channel; 31, first spring; 32, sliding rod; 33, slider; 34, first cooling channel; 35, second cooling channel. Detailed implementation manners

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0033] Please refer to Figure 1-9 , the present invention provides a technical solution: a pipe fitting processing mold, including a base 1, a lower die holder 11 is installed on the base 1, a plurality of groups of lower cavities 13 are arranged on the lower die holder 11, a lifting plate 7 is arranged above the base 1, an upper die holder 8 is installed at the bottom of the lifting plate 7, a plurality of groups of upper cavities 16 are opened on the upper die holder 8, core body components are installed at both ends of the lower cavity 13, a core pulling and mating component for core pulling of the core body components is installed on the lifting plate 7, a blanking component is installed on the base 1, the core body components include a core post 15, an inclined surface is opened on the core post 15, a plurality of groups of sliding grooves are opened on the inclined surface, sliders 33 are slidably connected to the sliding grooves, a bending part 14 is jointly connected to the plurality of groups of sliders 33, a sliding rod 32 is fixedly connected to the slider 33, the sliding rod 32 slidably penetrates through the core post 15, an end of the sliding rod 32 is fixedly connected to a mounting block 29, a first spring 31 is sleeved on the outer surface of the sliding rod 32, one end of the first spring 31 is fixedly connected to the core post 15, and the other end of the first spring 31 is fixedly connected to the mounting block 29.

[0034] During mold clamping, the bending part 14 will first extend into the interior of the cavity. When the bending part 14 contacts the inner wall of the cavity, the bending part 14 will slide inward along the inclined surface of the core column 15, and the bending part 14 will slide into the interior of the core column 15. The inner core of the bent pipe is composed of two groups of bending parts 14 and two groups of core columns 15. During mold opening, under the action of the first spring 31, the bending part 14 will remain in the interior of the cavity, and the core column 15 will first disengage from the cavity. During the disengagement process of the core column 15, under the action of the first spring 31, the bending part 14 will slide along the inclined surface on the core column 15, and the bending part 14 will move away from the bent part of the cavity. The core column 15 continues to move outward, and the bending part 14 will disengage from the cavity. The mold clamping state of the core body assembly during mold clamping is shown in Figure 7 .

[0035] The core pulling and mating assembly includes a guide rail 4, the guide rail 4 is fixedly installed on the lower mold base 11, a moving seat 5 is slidably installed on the guide rail 4, multiple groups of inclined holes are formed in the moving seat 5, a mating rod 6 is slidably inserted through the inclined holes, the top of the mating rod 6 is fixedly connected to the lifting plate 7, and the moving seat 5 is fixedly connected to the core column 15 through multiple groups of connecting rods 28.

[0036] During mold clamping, the lifting plate 7 will move downward. When the lifting plate 7 moves downward, it will drive the moving seat 5 to move towards the cavity through the mating rod 6. Furthermore, the core body assembly will enter the interior of the cavity. During mold opening, under the action of the mating rod 6, the core body assembly will disengage from the cavity, realizing that during mold clamping, the core body assembly enters the interior of the cavity, and during mold opening, the core body assembly disengages from the cavity, avoiding the need to add external driving electrical components. Moreover, when the core body assembly is assembled or disassembled, the structure is simple, avoiding the need to design a complex gear transmission structure.

[0037] A first cooling flow channel 34 is formed in the core column 15, and a second cooling flow channel 35 is formed in the bending part 14. The corresponding two groups of first cooling flow channels 34 and the corresponding two groups of second cooling flow channels 35 form an internal coolant flow channel 30.

[0038] During mold clamping, the two groups of first cooling flow channels 34 and the two groups of second cooling flows 35 form an internal coolant flow 30, thereby enabling internal cooling during the injection molding of the plastic bent pipe.

[0039] Multiple groups of external cooling flow channels 12 are formed on both the upper mold base 8 and the lower mold base 11. External cooling can be performed during the injection molding of the plastic bent pipe through the multiple groups of external cooling flow channels 12.

[0040] Semicircular grooves 23 are formed at both ends of the lower cavity 13 and both ends of the upper cavity 16, and a sealing ring block 27 is fixedly sleeved on the outer surface of the core column 15.

[0041] After mold clamping, the sealing ring block 27 will be pressed tightly between the two groups of semicircular grooves 23, thereby sealing the two ends of the cavity.

[0042] The ejector assembly includes moving plates 3 on the left and right sides. Fixed blocks 2 are fixedly connected to the left and right end faces of the lower die base 11. A number of moving slots are provided on the fixed blocks 2, and moving blocks are slidably connected to the moving slots. The moving blocks are fixedly connected to the moving plates 3. A connecting plate 18 is fixedly connected between the two groups of moving plates 3. A first inclined guide block 20 is fixedly connected to the connecting plate 18. A number of ejector pins 21 are slidably inserted through the lower die base 11. The bottoms of multiple groups of ejector pins 21 are fixedly connected to a mounting plate 24. A second spring 17 is sleeved on the outer surface of the ejector pin 21. One end of the second spring 17 is fixedly connected to the bottom of the lower die base 11, and the other end of the second spring 17 is fixedly connected to the mounting plate 24. A second inclined guide block 19 is fixedly connected to the mounting plate 24. A set of vertical sliding openings 25 and a set of inclined sliding openings 26 are provided on the moving plate 3. The vertical sliding opening 25 communicates with the inclined sliding opening 26. Fixed columns 22 are fixedly connected to the left and right end faces of the lifting plate 7. The fixed columns 22 are slidably inserted through the vertical sliding openings 25.

[0043] During mold opening, the lifting plate 7 will drive the fixed column 22 to move upward. The fixed column 22 will slide along the vertical sliding opening 25. As the lifting plate 7 continues to move upward, the fixed column 22 will slide along the inclined sliding opening 26. At this time, the fixed column 22 will drive the moving plate 3 to move forward. When the moving plate 3 moves forward, it will drive the first inclined guide block 20 to move forward. The first inclined guide block 20 is in inclined surface cooperation with the second inclined guide block 19. The second inclined guide block 20 will drive a number of ejector pins 21 to move upward, realizing the automatic ejection of the ejector pins 21 during mold opening. During mold closing, the fixed column 22 will slide along the vertical sliding opening 25. At this time, under the action of the second spring 17, a number of ejector pins 21 will move downward to reset, realizing the automatic reset of a number of ejector pins 21 during mold closing.

[0044] A number of positioning columns 9 are fixedly connected to the bottom of the lifting plate 7. A number of positioning holes 10 are provided on the base 1. During mold closing, a number of positioning columns 9 will be inserted into the corresponding positioning holes 10 to ensure the accuracy of mold closing.

[0045] Working principle: During mold closing, the bending part 14 will first extend into the interior of the cavity. When the bending part 14 contacts the inner wall of the cavity, the bending part 14 will slide inward along the inclined surface of the core column 15. The bending part 14 will slide into the interior of the core column 15. The inner core of the bent pipe is composed of two groups of bending parts 14 and two groups of core columns 15. During mold opening, under the action of the first spring 31, the bending part 14 will remain in the interior of the cavity. The core column 15 will first disengage from the cavity. During the process of the core column 15 disengaging, under the action of the first spring 31, the bending part 14 will slide along the inclined surface on the core column 15. The bending part 14 will move out of the bent part of the cavity. As the core column 15 continues to move outward, the bending part 14 will disengage from the cavity. The mold closing state of the core body assembly during mold closing is shown in Figure 7 ;

[0046] During mold clamping, the lifting plate 7 will move downward. When the lifting plate 7 moves downward, it will drive the moving seat 5 to move towards the cavity through the cooperation rod 6. Further, the core component will enter the interior of the cavity. During mold opening, under the action of the cooperation rod 6, the core component will disengage from the cavity, realizing that during mold clamping, the core component enters the interior of the cavity, and during mold opening, the core component disengages from the cavity, avoiding the need to add external driving electrical components. Moreover, when the core component is assembled or disassembled, the structure is simple, avoiding the need to design a complex gear transmission structure;

[0047] During mold opening, the lifting plate 7 will drive the fixed column 22 to move upward. The fixed column 22 will slide along the vertical sliding port 25. As the lifting plate 7 continues to move upward, the fixed column 22 will slide along the inclined sliding port 26. At this time, the fixed column 22 will drive the moving plate 3 to move forward. When the moving plate 3 moves forward, it will drive the first inclined guide block 20 to move forward. The first inclined guide block 20 is in inclined surface cooperation with the second inclined guide block 19, and the second inclined guide block 20 will drive a plurality of ejector pins 21 to move upward, realizing that during mold opening, the ejector pins 21 are automatically ejected. During mold clamping, the fixed column 22 will slide along the vertical sliding port 25. At this time, under the action of the second spring 17, a plurality of ejector pins 21 will move downward to reset, realizing that during mold clamping, a plurality of ejector pins 21 are automatically reset;

[0048] This elbow injection mold avoids the need for a complex gear transmission structure for core pulling. The core pulling is simple. Moreover, during mold opening, the core component can automatically extend into the interior of the cavity, and during mold opening, the core component can automatically disengage from the cavity. In addition, this elbow injection mold realizes that during mold opening, a plurality of ejector pins 21 automatically eject the injection molded elbow pipe fitting, and during mold clamping, a plurality of ejector pins 21 can automatically move downward to reset.

[0049] The design method of the present invention models and optimizes the mold processing process through a deep learning algorithm to realize the automatic adjustment of processing parameters. The specific process is as follows:

[0050] Step 1: Data collection

[0051] 1. Sensor data collection formula

[0052] The data collected by the sensor at time t is: X(t) = [P(t), T(t), L(t)]

[0053] P(t) is the pressure sensor data at time t;

[0054] T(t) is the temperature sensor data at time t;

[0055] L(t) is the position sensor data at time t;

[0056] 2. Processing time data

[0057] Δt = t - t0

[0058] Δt is the processing time, t is the current time, and t0 is the start time of processing;

[0059] 3. Processing result data

[0060] Y(t) = [Q(t), A(t)]

[0061] Y(t) is the processing result data vector at time t, Q(t) is the product precision data at time t, and A(t) is the product quality data at time t;

[0062] Comprehensive data collection formula

[0063] Combining sensor data, processing time, and processing results to form a complete data collection formula:

[0064] Z(t) = [P(t), T(t), L(t), Δt, Q(t), A(t)]

[0065] where Z(t) is the comprehensive data vector collected at time t, P(t) is the pressure data at time t, T(t) is the temperature data at time t, L(t) is the position data at time t, Δt is the processing time, Q(t) is the product precision data at time t, and A(t) is the product quality data at time t;

[0066] Step 2: Data preprocessing

[0067] 1. Data cleaning

[0068] Remove noise and use the moving average method to smooth the data;

[0069]

[0070] where x′ i is the smoothed data point, k is half of the window size, 2k + 1 is the smoothing window, j is the index variable, and x i+j is the original data point at the i + j-th position, represents the sum of all x i+j data points from j = -k to j = k;

[0071] 2. Data standardization

[0072] Standardize the data to transform it into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardized data zi is expressed as:

[0073]

[0074] z i is the standardized data point, xi is the original data point, μ is the mean of the data, and σ is the standard deviation of the data. For the standardization process;

[0075] 3. Feature extraction

[0076] Extract the mean and standard deviation from the processed data as features, and the feature vector is expressed as:

[0077] f(t) = [μ P (t), σ P (t), μT(t), σ T (t), μ L (t), σ L (t)]

[0078] f(t) is the feature vector, μ P (t) is the mean of the pressure sensor data P(t) at time t, σ P (t) is the standard deviation of the pressure sensor data P(t) at time t, μ T (t) is the mean of the temperature sensor data T(t) at time t, σ T (t) is the standard deviation of the temperature sensor data T(t) at time t, μ L (t) is the mean of the position sensor data L(t) at time t, σ L (t) is the standard deviation of the position sensor data L(t) at time t;

[0079] 4. Dataset division

[0080] Divide the processed data into a training set, a validation set, and a test set; the division ratios are p train , p val , p test , satisfying:

[0081] p train + p val + p test = 1

[0082] p train is the ratio of the training set, p val is the ratio of the validation set, p test is the ratio of the test set;

[0083] The division formula is:

[0084] D train = D × P train

[0085] D val = D × P val

[0086] D test = D × P test

[0087] D train is the training set, D val is the validation set, D test is the test set, D is the total data set, P train is the proportion of the training set, P val is the proportion of the validation set, P test is the proportion of the test set;

[0088] Step 3 Model establishment and training

[0089] 1. Model input

[0090] The input layer contains sensor data, expressed as a time series X(t):

[0091] X(t) = [P(t), T(t), L(t), Δt]

[0092] X(t) is a time series, P(t) is the pressure sensor data at time t, T(t) is the temperature sensor data at time t, L(t) is the position sensor data at time t, and Δt is the processing time;

[0093] 2. LSTM cell formula

[0094] Each LSTM cell contains an input gate, a forget gate, and an output gate to control the information flow. The specific formulas are as follows:

[0095] Input gate:

[0096] i t = σ(W i · [h t-1 , X(t)] + b i )

[0097] i t is the input gate vector, σ is the Sigmoid activation function, W i is the weight matrix of the input gate, h t-1 is the hidden state vector at the previous moment (time step t - 1), X(t) is the input data vector at the current moment (time step t), and b i is the bias vector of the input gate;

[0098] Forget gate:

[0099] f t = σ(W f · [h t-1 , X(t)] + b f )

[0100] f tis the output of the forget gate at time step t, σ is the Sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state vector at the previous time step, X(t) is the input data vector at the current time step, b f is the bias vector of the forget gate;

[0101] Output gate:

[0102] o t = σ(W o · [h t-1 , X(t)] + b o )

[0103] o t is the activation value of the output gate, σ is the Sigmoid activation function, W o is the weight matrix of the output gate, h t-1 is the hidden state vector at the previous time step, X(t) is the input data vector at the current time step, b o is the bias vector of the output gate;

[0104] Candidate memory cell state:

[0105]

[0106] is the candidate memory cell state at time t, tanh is the hyperbolic tangent function, W C is the weight matrix, h t-1 is the hidden state vector at the previous time step, X(t) is the input data vector at the current time step, b C is the bias vector;

[0107] Memory cell state update:

[0108]

[0109] C t is the memory cell state at time t, f t is the forget gate at time t, C t-1 is the memory cell state at time t-1, i t is the input gate at time t, is the candidate memory cell state at time t;

[0110] Hidden state update:

[0111] h t = o t · tanh(C t )

[0112] h tis the hidden state at time step t, o t is the output gate at time step t, tanh(C t ) is the memory cell state at time step t;

[0113] 3. Output layer formula

[0114] The output of the LSTM layer is passed to the fully connected layer to predict the processing parameters;

[0115] Y(t) = [M(t), F(t)]

[0116] Y(t) is the processing parameter prediction vector at time t, M(t) is the predicted value of the moving plate position at time t, and F(t) is the predicted value of the thimble force at time t;

[0117] Output of the fully connected layer:

[0118] Y(t) = W y ·h t + b y

[0119] Y(t) is the processing parameter vector at time t, W y is the weight matrix of the fully connected layer, h t is the hidden state vector at time t, b y is the bias vector of the fully connected layer;

[0120] 4. Loss function and optimization

[0121] Use the mean squared error MSE as the loss function to optimize the parameters of the LSTM model. Loss function formula:

[0122]

[0123] L is the value of the loss function, N is the number of samples, is the summation symbol, Yi is the true processing parameter vector of the i-th sample, is the predicted processing parameter vector of the i-th sample, is the square of the Euclidean distance between the true value and the predicted value of the i-th sample;

[0124] Calculate the gradient of the loss function with respect to the model parameters through the backpropagation algorithm, and use the gradient descent algorithm to update the model parameters:

[0125]

[0126] θ is the model parameter, η is the learning rate, is the gradient of the loss function with respect to the model parameters;

[0127] 5. Model tuning

[0128] For each set of hyperparameter combinations λ:

[0129] Train the model to obtain the model parameters θλ

[0130] Calculate the loss function on the validation set:

[0131]

[0132] L val (λ) is the validation loss function value, N val is the number of samples in the validation set, is the coefficient for taking the average, is the summation symbol, Y i is the true machining parameter vector of the i-th validation set sample, is the predicted machining parameter vector of the i-th validation set sample, is the square of the Euclidean distance between the true value and the predicted value of the i-th validation set sample;

[0133] Select the hyperparameter combination λ with the minimum validation set loss * :

[0134] λ * = argmin λ L val (λ)

[0135] λ * is the optimal hyperparameter combination, argmin λ is to find the parameter λ that minimizes the objective function L val minimum, L val (λ) is the loss function value on the validation set, and λ is the hyperparameter combination;

[0136] Step 4 Model deployment

[0137] 1. Real-time data acquisition

[0138] Real-time acquisition of sensor data, expressed as a time series X(t):

[0139] X(t) = [P(t), T(t), L(t), Δt]

[0140] X(t) is a time series, P(t) is the pressure sensor data at time t, T(t) is the temperature sensor data at time t, L(t) is the position sensor data at time t, and Δt is the processing time;

[0141] 2. Model prediction

[0142] Use the trained deep learning model f(X(t); θ) to predict the optimal machining parameters where θ is the model parameter:

[0143]

[0144] The machining parameter vector predicted for time t, f is the trained deep learning model, X(t) is the sensor data vector at time t, and θ is the model parameter;

[0145] The predicted machining parameters include the moving plate position and the ejector pin force:

[0146]

[0147] The machining parameter vector predicted for time t, The moving plate position predicted for time t, The ejector pin force predicted for time t;

[0148] 3. Automatic adjustment

[0149] Transfer the predicted machining parameters to the industrial control system to automatically adjust the position and force of the moving plate (3) and the ejector pin (21) of the mold:

[0150] Moving plate position adjustment

[0151] According to the predicted value Adjust the moving plate position:

[0152]

[0153] M current M(t + 1) is the current position of the moving plate at time t + 1, is the moving plate position predicted for time t;

[0154] Ejector pin force adjustment

[0155] According to the predicted value Adjust the ejector pin force:

[0156]

[0157] F current F(t + 1) is the current force of the ejector pin at time t + 1, is the ejector pin force predicted by the trained model at time t;

[0158] Step Five Real-time learning and optimization

[0159] 1. Real-time data acquisition and update

[0160] Acquire sensor data in real time and add the new data to the dataset for online learning and model update;

[0161] X new (t) = [P new (t), T new (t), L new (t), Δt new

[0162] X new (t) is the new sensor data vector at time t, P new (t) is the new pressure sensor data at time t, T new (t) is the new temperature sensor data at time t, L new (t) is the new position sensor data at time t, Δt new is the new processing time;

[0163] The corresponding new processing parameter is Y new (t):

[0164] Y new (t) = [M new (t), F new

[0165] Y new (t) is the new true processing parameter vector at time t, M new (t) is the new moving plate position at time t, F new (t) is the new ejector pin force at time t;

[0166] 2. Online learning and model update

[0167] Use the new data to perform online learning and update on the model:

[0168] Calculate the new loss function:

[0169]

[0170] L new is the value of the new loss function, Y new (t) is the new true processing parameter vector at time t, is the processing parameter vector predicted by the model at time t, is the new true processing parameter vector Y new (t) and the processing parameter vector predicted by the model the square of the Euclidean distance between them;

[0171] Update the model parameters by gradient descent:

[0172]

[0173] θ is the model parameter, η is the learning rate, ​​is the gradient of the loss function Lnew with respect to the model parameters θ;

[0174] 3. Regularly evaluate the model performance

[0175] Regularly use the validation set data to evaluate the model performance to ensure the continuous optimization and improvement of the model;

[0176] Evaluate the loss function:

[0177] At each evaluation period T eval , calculate the average loss function on the validation set:

[0178]

[0179] L val is the average loss function value on the validation set, N val is the number of samples in the validation set, is the coefficient for calculating the average value, is the summation symbol, Y val，i is the true processing parameter vector of the i-th sample in the validation set, is the predicted processing parameter vector of the i-th sample in the validation set, is the square of the Euclidean distance between the true value and the predicted value of the i-th sample;

[0180] Adjust the model according to the evaluation results:

[0181] If the validation set loss L val exceeds the preset threshold, adjust the model hyperparameters.

[0182] Numerical case analysis:

[0183] Sensor data:

[0184] Pressure: P(t) = 100 Pa, 105 Pa

[0185] Temperature: T(t) = 25 °C, 27 °C

[0186] Position: L(t) = 5 mm, 5.2 mm

[0187] Processing time: Δt = 10 s, 12 s

[0188] Processing parameters:

[0189] Moving plate position: M(t) = 5.0 mm, 5.1 mm

[0190] Ejector pin force: F(t) = 100 N, 105 N

[0191] Train the model using the above data with initial parameters θ and learning rate η = 0.01.

[0192] New data:

[0193] Pressure: 110 Pa

[0194] Temperature: 30 °C

[0195] Position: 5.5 mm

[0196] Processing time: 15 s

[0197] Model prediction:

[0198] Predict the position of the moving plate and the force of the ejector pin:

[0199]

[0200] Automatic adjustment

[0201] Adjust the moving plate and the ejector pin:

[0202] M current (t + 1) = 5.3

[0203] F current (t + 1) = 108

[0204] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A pipe fitting processing die, comprising a base (1), characterized in that: A lower die base (11) is installed on the base (1). Multiple groups of lower cavities (13) are arranged on the lower die base (11). An elevating plate (7) is arranged above the base (1). An upper die base (8) is installed at the bottom of the elevating plate (7). Multiple groups of upper cavities (16) are formed on the upper die base (8). Core component assemblies are installed at both ends of the lower cavity (13). A core-pulling cooperation component for core-pulling of the core component assemblies is installed on the elevating plate (7). A knockout component is installed on the base (1). The core component assembly includes a core column (15). An inclined surface is formed on the core column (15). Multiple groups of chutes are formed on the inclined surface. Sliders (33) are slidably connected to the chutes. A bending part (14) is connected to the multiple groups of sliders (33) together. A slide bar (32) is fixedly connected to the slider (33). The slide bar (32) slidably penetrates through the core column (15). An installation block (29) is fixedly connected to the end of the slide bar (32). A first spring (31) is sleeved on the outer surface of the slide bar (32). One end of the first spring (31) is fixedly connected to the core column (15), and the other end of the first spring (31) is fixedly connected to the installation block (29). The core-pulling cooperation component includes a guide rail (4). The guide rail (4) is fixedly installed on the lower die base (11). A moving seat (5) is slidably installed on the guide rail (4). Multiple groups of inclined holes are formed on the moving seat (5). A cooperation rod (6) slidably penetrates through the inclined holes. The top of the cooperation rod (6) is fixedly connected to the elevating plate (7). The moving seat (5) is fixedly connected to the core column (15) through multiple groups of connecting rods (28).

2. The pipe fitting processing die according to claim 1, characterized in that: A first cooling flow channel (34) is formed on the core column (15). A second cooling flow channel (35) is formed on the bending part (14). The corresponding two first cooling flow channels (34) and the corresponding two second cooling flow channels (35) form an internal coolant flow channel (30).

3. A pipe fitting processing mold according to claim 2, characterized in that: Multiple groups of external cooling flow channels (12) are formed on both the upper die base (8) and the lower die base (11).

4. A pipe fitting processing die according to claim 3, characterized in that: Semicircular grooves (23) are formed at both ends of the lower cavity (13) and both ends of the upper cavity (16). A sealing ring block (27) is fixedly sleeved on the outer surface of the core column (15).

5. A pipe fitting processing die according to claim 4, characterized in that: The ejector component includes moving plates (3) on the left and right sides. Fixed blocks (2) are fixedly connected to the left and right end faces of the lower die base (11). A plurality of moving grooves are formed in the fixed blocks (2), and moving blocks are slidably connected to the moving grooves. The moving blocks are fixedly connected to the moving plates (3). A connecting plate (18) is fixedly connected between the two groups of moving plates (3). A first inclined guide block (20) is fixedly connected to the connecting plate (18). A plurality of ejector pins (21) are slidably inserted through the lower die base (11). The bottoms of multiple groups of ejector pins (21) are fixedly connected to a mounting plate (24). A second spring (17) is sleeved on the outer surface of the ejector pin (21). One end of the second spring (17) is fixedly connected to the bottom of the lower die base (11), and the other end of the second spring (17) is fixedly connected to the mounting plate (24). A second inclined guide block (19) is fixedly connected to the mounting plate (24). A set of vertical sliding openings (25) and a set of inclined sliding openings (26) are formed in the moving plate (3). The vertical sliding opening (25) is communicated with the inclined sliding opening (26). Fixed columns (22) are fixedly connected to the left and right end faces of the lifting plate (7). The fixed columns (22) are slidably inserted through the vertical sliding openings (25).

6. The pipe fitting processing die according to claim 5, wherein: A plurality of positioning columns (9) are fixedly connected to the bottom of the lifting plate (7). A plurality of positioning holes (10) are formed in the base (1).

7. The design optimization method of a pipe fitting processing die according to claim 5, characterized in that: Modeling and optimizing the mold processing process through deep learning algorithms to achieve automatic adjustment of processing parameters. The specific process is as follows: Step 1: Data collection 1. Sensor data collection formula The data collected by the sensor at time t is: X(t)=[P(t),T(t),L(t)] P(t) is the pressure sensor data at time t; T(t) is the temperature sensor data at time t; L(t) is the position sensor data at time t; 2. Processing time data ； is the processing time, t is the current time, is the start time of processing; 3. Processing result data ; is the processing result data vector at time t, is the product precision data at time t, is the product quality data at time t; Comprehensive data collection formula Combining sensor data, processing time, and processing results to form a complete data collection formula: ； Among them, is the comprehensive data vector collected at time t, is the pressure data at time t, is the temperature data at time t, is the position data at time t, is the processing time, is the product precision data at time t, is the product quality data at time t; Step 2: Data preprocessing 1. Data cleaning Removing noise uses the moving average method to smooth the data; ； Among them, is the data point after smoothing, is half of the window size, is the smoothing window, is the index variable, is the th original data point at the position, represents the sum of all to data points;​ 2. Data standardization Standardize the data and transform it into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardized data zi is expressed as: ; is the standardized data point, is the original data point, is the mean of the data, is the standard deviation of the data, is the standardization process; 3. Feature extraction Extract the mean and standard deviation as features from the processed data. The feature vector is expressed as: ； is the eigenvector, is the mean value of the pressure sensor data P(t) at time t, is the standard deviation of the pressure sensor data P(t) at time t, is the mean value of the temperature sensor data T(t) at time t, is the standard deviation of the temperature sensor data T(t) at time t, is the mean value of the position sensor data L(t) at time t, is the standard deviation of the position sensor data L(t) at time t; 4. Dataset division Divide the processed data into a training set, a validation set, and a test set; the division ratios are respectively , , , satisfying: ； is the proportion of the training set, is the proportion of the validation set, is the proportion of the test set; The division formula is: ； is the training set, is the validation set, is the test set, is the total dataset, is the proportion of the training set, is the proportion of the validation set, is the proportion of the test set; Step 3: Model establishment and training 1. Model input The input layer contains sensor data, represented as a time series : ； is a time series, is the pressure sensor data at time t, is the temperature sensor data at time t, is the position sensor data at time t, is the processing time; 2. LSTM unit formula Each LSTM unit contains an input gate, a forget gate, and an output gate to control the information flow. The specific formulas are as follows: Input gate: ； is the input gate vector, is the Sigmoid activation function, is the weight matrix of the input gate, is the hidden state vector at the previous time step (time step t - 1), is the input data vector at the current time step (time step t), is the bias vector of the input gate; Forget gate: ； is the output of the forget gate at time step t, is the Sigmoid activation function, is the weight matrix of the forget gate, is the hidden state vector at the previous time step, is the input data vector at the current time step, is the bias vector of the forget gate; Output gate: ； is the activation value of the output gate, is the Sigmoid activation function, is the weight matrix of the output gate, is the hidden state vector at the previous moment, is the input data vector at the current moment, is the bias vector of the output gate; Candidate memory cell state: ； is the candidate memory cell state at time t, is the hyperbolic tangent function, is the weight matrix, is the hidden state vector at the previous moment, is the input data vector at the current moment, is the bias vector; Memory cell state update: ； is the state of the memory cell at time t, is the forget gate at time t, is the state of the memory cell at time t - 1, is the input gate at time t, is the candidate memory cell state at time t; Hidden state update: ； is the hidden state at time step t, is the output gate at time step t, is the state of the memory cell at time step t; 3. Output layer formula The output of the LSTM layer is passed to the fully connected layer to predict the processing parameters; ； is the predicted vector of the processing parameters at time t, is the predicted value of the position of the moving plate at time t, is the predicted value of the thimble force at time t; Fully connected layer output: ； is the processing parameter vector at time t, is the weight matrix of the fully connected layer, is the hidden state vector at time t, is the bias vector of the fully connected layer; 4. Loss function and optimization Use the mean squared error MSE as the loss function to optimize the parameters of the LSTM model. The loss function formula: ； is the value of the loss function, the number of samples, is the summation symbol, is the true machining parameter vector of the i-th sample, is the predicted machining parameter vector of the i-th sample, is the square of the Euclidean distance between the true value and the predicted value of the i-th sample; Calculate the gradient of the loss function with respect to the model parameters through the backpropagation algorithm and update the model parameters using the gradient descent algorithm: ； is a model parameter, is the learning rate, is the gradient of the loss function with respect to the model parameter; 5. Model Tuning For each set of hyperparameter combinations : Train the model to obtain model parameters ; Calculate the loss function on the validation set: ； To verify the loss function value, is the number of samples in the validation set, is the coefficient for calculating the average value, is the summation symbol, is the true machining parameter vector of the i-th validation set sample, is the predicted machining parameter vector of the i-th validation set sample, is the square of the Euclidean distance between the true value and the predicted value of the i-th validation set sample; Select the hyperparameter combination with the smallest validation set loss : ； is the best hyperparameter combination, To find the parameters that minimize the objective function ; , is the loss function value on the validation set, is the hyperparameter combination; Step 4: Model Deployment 1. Real-time Data Acquisition Obtain sensor data in real time, represented as a time series : ； is a time series, is the pressure sensor data at time t, is the temperature sensor data at time t, is the position sensor data at time t, is the processing time; 2. Model Prediction Using a trained deep learning model to predict the optimal processing parameters , where are model parameters: ； The process parameter vector predicted for time t, The trained deep learning model, The sensor data vector at time t, The model parameters; The predicted processing parameters include the moving plate position and the ejector pin force: ； The process parameter vector predicted for time t, The moving plate position predicted for time t, The ejector pin force predicted for time t; 3. Automatic Adjustment Transfer the predicted processing parameters to the industrial control system to automatically adjust the position and force of the moving plate (3) and the ejector pin (21) of the mold: Moving Plate Position Adjustment According to the predicted value Adjust the position of the moving plate: ； is the current position of the moving plate at time t+1, is the predicted position of the moving plate at time t; Ejector Pin Force Adjustment According to the predicted value Adjust the thimble force: ； is the current force of the ejector pin at time t+1, is the ejector pin force predicted by the trained model at time t; Step 5: Real-time Learning and Optimization 1. Real-time Data Acquisition and Update Real-time acquire sensor data and add the new data to the dataset for online learning and model update; ； is the new sensor data vector at time t, is the new pressure sensor data at time t, is the new temperature sensor data at time t, is the new position sensor data at time t, is the new processing time; The corresponding new processing parameters are : ； is the new true processing parameter vector at time t, is the new moving plate position at time t, is the new ejector pin force at time t; 2. Online Learning and Model Update Use the new data for online learning and updating of the model: Calculate the new loss function: ； is the value of the new loss function, is the new true machining parameter vector at time t, is the machining parameter vector predicted by the model at time t, is the new true machining parameter vector and the machining parameter vector predicted by the model the square of the Euclidean distance between them; Update the model parameters by gradient descent: ； is a model parameter, is the learning rate, is the loss function The gradient of the model parameter ; 3. Regularly Evaluate Model Performance Regularly evaluate the model performance using the validation set data to ensure continuous optimization and improvement of the model; Evaluate the loss function: At each evaluation period T eval , calculate the average loss function on the validation set: ； The average loss function value on the validation set, The number of samples in the validation set, The coefficient for average value calculation, The summation symbol, The true machining parameter vector of the i-th sample in the validation set, The predicted machining parameter vector of the i-th sample in the validation set, The square of the Euclidean distance between the true value and the predicted value of the i-th sample; Adjust the model according to the evaluation results: If the validation set loss L val exceeds the preset threshold, adjust the model hyperparameters.

Images