BP neural network, 1394 cable wrapping method, system and sintering method

Abstract

A BP neural network, a 1394 cable wrapping method, device and sintering method. The wrapping method comprises: using a slitting knife pad with precisely controlled size to cut a low-density PTFE insulating tape into widths x and y; using low-density PTFE insulating tapes of widths x and y to wrap two layers respectively; wherein x, y and the main machine speed, tension, pitch, and outer diameter parameters in the wrapping process are input into a BP neural network model, and the optimal process parameters are determined based on the output values ​​of the BP neural network model to execute the 1394 cable wrapping method. The method of the present invention can ensure that the dielectric constant of the low-density PTFE tape in the wrapped insulation 1 and the wrapped insulation 2 after sintering is controllable and meets the design requirements, and can ensure stable and reliable output of products with constant quality in industrial production.

Description

Technical Field

[0001] The present invention relates to the field of communication cable manufacturing, and more specifically to a BP neural network, a 1394 cable wrapping method using the same, a wrapping system and a wrapping sintering method. Background Art

[0002] Computer interface IEEE1394, commonly known as FireWire interface, is a high-speed transmission interface developed by a development alliance led by Apple in 1995. The data transmission rate is generally 800MB / s to 1.6GB / s, which is very suitable for video image transmission. IEEE1394 is mainly used for video acquisition. In addition to Apple, it is also widely used on INTEL high-end motherboards and digital video cameras (DV). Most of its technical standards come from companies such as Texas Instruments, Sony, DEC, IBM, and STMicroelectronics.

[0003] 1394 cable connectors are divided into four-pin, six-pin, nine-pin and other forms. The cable part of the 1394 cable is matched with several signal lines, which are wrapped and sintered for insulation packaging, and then bundled into cables together with the buffer layer. In order to increase the adhesion of the insulation layer, solve the insulation shrinkage, improve the core wire strength, and facilitate subsequent processing, 1394 cables are usually sintered using a hot-melt PTFE process (wrapping and sintering process). In the sintering process, the core indicator dielectric constant is very easy to change. Foreign countries may have advanced control technology, but it has been kept secret; and domestic companies cannot guarantee process stability and produce products with stable quality due to inconsistent process parameters and raw materials of each manufacturer. Furthermore, due to the large number of cable process steps, the process parameters involved show an exponential growth. Since there is a control range for the consistency of production materials and the accuracy of equipment, the process control is even more complex and demanding. Therefore, it is urgent to develop a wrapping / sintering method that can accurately control various influencing factors and produce high-quality 1394 cables with quality and quantity. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to provide a wrapping method and a sintering method for a 1394 cable, in order to at least partially solve the above technical problems.

[0005] In order to achieve the above object, as a first aspect of the present invention, a method for constructing a BP neural network model of a wrapping sintering process is proposed, comprising the following steps:

[0006] Using one or more of the raw material parameter performance and parameter data of the wrapping and sintering process as input and the error between the product performance test result and the standard value as output, constructing the wrapping and sintering process BP neural network model;

[0007] The BP neural network model of the wrapping and sintering process is trained through actual production data to solve the weight and threshold of each node.

[0008] As a second aspect of the present invention, a method for establishing a wrapping sintering process simulation database is also proposed, comprising the following steps:

[0009] Constructing a wrapping and sintering process simulation database, inputting part or all of the raw material parameter performance, wrapping and sintering process parameters, and product performance test result data in the wrapping production process into the wrapping and sintering process simulation database;

[0010] The BP neural network model for wrapping and sintering process constructed by the method for constructing the BP neural network model for wrapping and sintering process as described above is used to expand the data of the wrapping and sintering process simulation database; wherein the input of the BP neural network model for wrapping and sintering process includes new parameters not stored in the simulation database, and the output of the BP neural network model for wrapping and sintering process is the predicted value of the BP neural network model for wrapping and sintering process based on the input including the new parameters.

[0011] As a third aspect of the present invention, a 1394 cable wrapping method is also proposed, comprising the following steps:

[0012] A slitting knife pad with precisely controlled size is used to slit the low-density PTFE insulating tape, and the slitting widths are x and y, respectively, where x and y are both positive integers;

[0013] Use low-density PTFE insulation tape with a width of x and a thickness of 3 mil. The main machine of the equipment is reversed to wrap the first layer, with 2 overlapping layers.

[0014] Use low-density PTFE insulation tape with a width of y and a thickness of 3 mil. The main machine of the equipment rotates forward to wrap the second layer, with 2 overlapping layers.

[0015] Among them, the four parameters of main machine speed, tension, pitch and outer diameter in the first layer wrapping process and the second layer wrapping process are input into the wrapping and sintering process BP neural network model constructed by the construction method of the wrapping and sintering process BP neural network model as described above, and the optimal process parameters are determined based on the output value of the wrapping and sintering process BP neural network model, and the slitting widths x and y are selected based on years of production experience to execute the wrapping method of the 1394 cable.

[0016] As a fourth aspect of the present invention, a 1394 cable wrapping control device is also proposed, comprising:

[0017] A signal acquisition unit, used to collect the main machine speed, tension, overlap rate and outer diameter parameter values ​​in the first layer wrapping process and the second layer wrapping process respectively;

[0018] The actuator is used to control the specific values ​​of the main machine speed, tension, overlap rate and outer diameter in the first layer wrapping process and the second layer wrapping process respectively; wherein the main machines of the first layer wrapping process and the second layer wrapping process rotate in opposite directions and have the same corresponding process parameters;

[0019] A control unit receives the signal input of the signal acquisition unit, receives the input of the width x and y of the low-density PTFE insulating tape in the first layer wrapping process and the second layer wrapping process, and determines the optimal process parameters of the wrapping process based on the output value predicted by the wrapping and sintering process BP neural network model constructed based on the construction method of the wrapping and sintering process BP neural network model as described above, or based on the search result of the wrapping and sintering process simulation database established by the construction method of the wrapping and sintering process simulation database as described above, and drives the actuator to execute the wrapping method of the 1394 cable as described above.

[0020] As a fifth aspect of the present invention, a wrapping and sintering method for a 1394 cable is also proposed, comprising the following steps:

[0021] Wrapping the conductive metal cable using the 1394 cable wrapping method described above;

[0022] High-temperature welding of the conductive metal cable wrapped with the sinterable PTFE film at 350° C. to 450° C.;

[0023] Among them, the temperature control adopts the RKC temperature controller self-feedback control system to carry out precise temperature control, and the sintering time control uses the equipment transmission coefficient ratio and the main machine speed for precise control.

[0024] Based on the above technical solution, it can be seen that the wrapping method and sintering method of the 1394 cable of the present invention have at least one of the following beneficial effects compared with the prior art:

[0025] The method of the present invention can ensure that the dielectric constant of the low-density PTFE tape in the first wrapping insulation process and the second wrapping insulation process after sintering is controllable and meets the design requirements, and can ensure stable and reliable output of products with constant quality in industrial production;

[0026] The method of the invention can accurately control and monitor the tension of the wrapping process in real time, reduce the fluctuation range of the insulating medium low-density PTFE insulating tape, and stabilize the electrical performance of the finished cable product. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1It is a framework diagram of the BP neural network model of the present invention;

[0028] Figure 2 It is a flow chart of the steps of the BP neural network model of the present invention for performing convergence calculation;

[0029] Figure 3 It is a framework diagram of a BP neural network model in a preferred embodiment of the present invention;

[0030] Figure 4 It is a calculation formula diagram of hidden layer nodes in the BP neural network model of the present invention;

[0031] Figure 5 It is a structural schematic diagram of a 1394 cable wrapping process execution device using the BP neural network model of the present invention;

[0032] Figure 6 It is a point-line graph of the prediction results using the BP neural network model of the present invention;

[0033] Figure 7 It is a schematic diagram of the cross-sectional structure of the 1394 cable finally prepared by the present invention. DETAILED DESCRIPTION

[0034] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0035] The meanings of some technical terms in this invention are as follows:

[0036] Cable is a general term for electrical cables and optical cables. In the present invention, the cable includes bare wires without connectors, and complete electrical cables or optical cables with connectors.

[0037] The 1394 cable was first defined by Apple in 1986 as "FireWire" and registered as a trademark. Based on this technology, the Institute of Electrical and Electronics Engineers (IEEE) of the United States formulated the IEEE 1394-1995 standard in 1995. The 1394 cable is a cable that adopts this standard. The interface of the 1394 cable is a serial interface, which has several types such as 4-pin, 6-pin, and 9-pin. It is characterized by fast transmission speed and is suitable for transmitting digital image signals.

[0038] Low-density polytetrafluoroethylene, commonly known as "expanded polytetrafluoroethylene", refers to polytetrafluoroethylene with a density range of 0.7 to 0.9 kg / m 3 Polytetrafluoroethylene (PTFE) is commonly used in insulating and sealing materials, such as raw tape.

[0039] Dielectric constant, also known as permittivity or relative permittivity, is an important data that characterizes the electrical properties of dielectrics or insulating materials, and is usually represented by ε. It refers to the ratio of the capacitance when the same material is used as the dielectric and when it is in a vacuum in the same capacitor, indicating the relative ability of the dielectric to store electrostatic energy in an electric field.

[0040] 1394 cables have a series of international standards, national standards and enterprise standards, and have unified regulations and requirements for the main structural performance parameters of each part of the cable. For example, for a cable with several signal lines, there are clear requirements for the characteristic impedance range of the signal lines, but for many details of the specific implementation, only the numerical range is given, or there is no limit at all. For example, the insulation thickness can be in a wider numerical range, and there is no requirement for the number and method of winding the insulation material, as long as the final product meets the standard limit. The performance of the cable is determined by its structure and material. Through the traditional electromagnetic field principle, a standard design scheme can be obtained if the structure is certain. However, in the actual production process, since each manufacturer producing 1394 cables has inconsistent processes and raw materials, it is necessary to modify the design according to the specific process capabilities and material characteristics. However, due to the large number of cable process steps, the process parameters involved show an exponential growth, and since there is a control range for the consistency of production materials and the accuracy of equipment, the process becomes a complex process of dynamic adjustment. Many domestic manufacturers are also unable to produce stable, reliable and constant quality products because they cannot grasp such complex scenarios.

[0041] Dielectric constant is an inherent property of insulating materials. As one of the commonly used insulating materials, low-density PTFE is a unidirectional shaped material with poor lateral strength. However, in high-end cables, such as stable amplitude and phase RF coaxial cables, low-temperature superconducting coaxial cables, and aerospace high-speed data bus cables, insulation is crucial. Low-density polytetrafluoroethylene (ePTFE) with low dielectric constant has the characteristics of corrosion resistance and non-combustion, and is the best choice for high-frequency communication insulation medium. Therefore, high-end cables can only be prepared by wrapping with ePTFE wrapping film. However, due to the gap between domestic materials and imported materials, if wrapping is carried out, PTFE is easily affected by wrapping pressure and changes in specific gravity during the process, and the corresponding dielectric constant will also change. At the same time, the insulation of high-frequency and high-speed cables has 5 to 6 ePTFE wrapping processes, and undergoes multiple processing links at the same time. Affected by tension and tension, PTFE needs to withstand the tension of the supply tape, and the soft PTFE will be stretched and deformed. Although this part does not change the specific gravity of the PTFE tape, it will affect the calculated dielectric constant, resulting in a huge deviation between the finished product indicators and the design indicators. Taking 1394B products as an example, the dielectric constant of domestic insulation materials is 1.273, the dielectric constant of finished products is 1.52, and the thickness of domestic film materials is 0.050±0.006mm. To achieve the purpose of insulation, it is necessary to wrap it multiple times. Therefore, if the insulation thickness is in the range of 0.264 to 0.336mm, the characteristic impedance of the produced cable may differ by more than 10Ω, which cannot meet the task requirements. Therefore, it is necessary to develop a production process that can accurately predict the dielectric constant of the product based on various process parameters, and achieve the unity of product design performance and actual product performance by accurately controlling the product process parameters, and accurately control the product process parameters.

[0042] In view of the above technical problems, the inventors have proposed the following solutions for the wrapping sintering process after a lot of repeated experiments:

[0043] 1. A complete database has been established through simulation and process practice, achieving what you see is what you get in design, and the design scheme is consistent with the performance of the finished product.

[0044] 2. Dynamic real-time monitoring during the production process. After the previous process is completed, the algorithm automatically controls the process parameters of the next process to achieve dynamic adaptive production and ensure the consistency of product performance.

[0045] 3. The tension of the wrapping process is precisely controlled and monitored in real time. The process parameters are dynamically controlled by the algorithm mentioned in point 2 to reduce the fluctuation range of the insulating medium, the low-density PTFE insulating tape, achieve precise control, and stabilize electrical performance.

[0046] In the above solution, the present invention uses BP (back propagation) neural network to assist in the realization of simulation and dynamic control. BP neural network is a concept proposed by scientists led by Rumelhart and McClelland in 1986. It is a multi-layer feedforward neural network trained according to the error back propagation algorithm. The neural network model does not need to determine the mathematical equation of the mapping relationship between input and output in advance. It only learns certain rules through its own training and obtains the result closest to the expected output value when the input value is given.

[0047] The structure of the BP neural network model usually includes an input layer, a hidden layer (hidden layer) and an output layer. Figure 1 As shown, the input layer of the BP neural network model contains m nodes, and the output layer contains n nodes, which can be regarded as a mapping from an m-dimensional vector to an n-dimensional vector. The selection of hidden layer nodes is based on the following empirical formula:

[0048]

[0049] Among them, h is the number of hidden layer nodes, m is the number of input layer nodes, n is the number of output layer nodes, and a is a regulation constant between 1 and 10.

[0050] BP neural network is a multi-layer feedforward network trained by error back propagation (abbreviated as error back propagation). Its algorithm is called BP algorithm. Its basic idea is gradient descent method, which uses gradient search technology to minimize the mean square error (MSE) of the error between the actual output value and the expected output value of the network. Figure 2 As shown: If the MSE does not meet the predetermined requirement e, the back propagation process is entered, and the output error signal MSE is transmitted back layer by layer in the form of a gradient along the original forward propagation path, and the error signal MSE is distributed to all neurons in each layer, thereby obtaining the error signal MSE of each layer of neurons j (j=1, 2, 3), the error signal MSE j As the basis for correcting the connection weights and thresholds, and modifying them, the two processes of forward propagation of information and reverse propagation of error are repeatedly run until the error signal MSE converges to e.

[0051] Although BP neural network is the most widely used algorithm in artificial neural network, it also has some defects, such as slow learning convergence speed, inability to guarantee convergence to the global minimum point, and difficulty in determining the network structure. In addition, the selection of network structure, initial connection weights and thresholds has a great influence on network training, but it cannot be accurately obtained.

[0052] In the present invention, targeted improvements are proposed for the above-mentioned shortcomings of the BP neural network and the specific application scenarios of the present invention:

[0053] 1. In the present invention, the BP neural network is mainly used for the establishment of a data simulation library, so that the BP neural network model can be trained and used on a large scale during informal production, avoiding the defects of slow convergence speed and the like that affect formal production;

[0054] 2. The present invention simplifies the BP neural network model based on actual production experience to avoid exponential growth of the entire operation due to too many parameters;

[0055] 3. The present invention combines the simulation of BP neural network with risk prevention and control to design the optimal emergency avoidance route.

[0056] Specifically, the present invention proposes a method for constructing a BP neural network model of a wrapping sintering process, comprising the following steps:

[0057] Using one or more of the raw material parameter performance and parameter data of the wrapping and sintering process as input and the error between the product performance test result and the standard value as output, constructing the wrapping and sintering process BP neural network model;

[0058] The BP neural network model of the wrapping and sintering process is trained through actual production data to solve the weight and threshold of each node.

[0059] The present invention also proposes a method for establishing a wrapping sintering process simulation database, comprising the following steps:

[0060] Input some or all of the raw material parameters and properties actually used in the wrapping production process, the parameters of the wrapping sintering process, and the product performance test results into the simulation database;

[0061] Taking one or more of the above typical raw material parameter performance, wrapping and sintering process parameters, and product performance test results as input to establish a BP neural network model, establishing the error with the standard value based on the product performance test results as output, and using the above simulation database to train the BP neural network model in large quantities;

[0062] The BP neural network model obtained through the above training is used to predict new raw materials or process parameters to guide subsequent production. At regular intervals, such as a quarter or a year, the actual production conditions and product performance parameters of these new raw materials or process parameters encountered in formal production are collected and collated, and the predicted values ​​and BP neural network model are corrected based on the actual test performance parameter results, and the actual test performance parameter results are added to the above simulation database.

[0063] Among them, the raw material parameter performance includes, for example, the density, thickness, width of the low-density PTFE insulating tape, the size of the wrapped conductor (silver-plated copper stranded wire), the coating thickness, conductivity, etc.; the parameters of the wrapping and sintering process include, for example, the number of wrapping layers, the speed of the wrapping main machine, the rotation direction of the main machine, the tension, the speed of the take-up motor, the pitch, the sintering temperature, the sintering rate (baking time), etc.; the product performance test results include, for example, the dielectric sintering thickness, dielectric constant, etc.

[0064] Among them, the input layer can select, for example, the density and width of the PTFE tape, the main engine speed, direction, tension, sintering temperature, sintering rate, etc., so that the number of input layer nodes m is between 2 and 10, preferably 2 to 4, to simplify the structure of the model and improve the efficiency of its convergence solution.

[0065] Among them, the output layer can, for example, select interlayer sealing, dielectric constant, differential impedance, insertion loss, transmission delay, etc., so that the number of output layer nodes n is between 1 and 8, preferably 1 to 3, to simplify the structure of the model and improve the efficiency of its convergence solution.

[0066] The applicant's historical production data is imported into the above BP neural network model to solve the branch weight ω of each node. ij ,ω jk etc., and threshold θ (bias value); wherein i, j, k, etc. represent the numbers of nodes at different layers. In order to enable the model to be solved quickly and stably, the inventors streamlined and optimized the above input and output vectors based on historical experience and physical and electrical calculation formulas. By continuously using the above BP neural network model in actual production, as well as the data obtained from the trial production of new materials, a stable and reliable prediction model is finally obtained, and the prediction model and physical and electrical calculation formulas are used, based on the possible parameters of the purchased raw materials and specific processes, as well as the raw materials and process parameters that may be obtained in the future, the simulation data under possible parameter conditions is output and imported into the above simulation database.

[0067] In actual application, the above BP neural network model can be directly used for prediction, or the simulation database can be directly searched to directly search for matching results, so as to reduce the calculation time of model prediction. In addition, by performing predictive simulation in advance, various situations can be predicted. At the same time, this solution can also be used for risk avoidance in unexpected situations. For example, in unexpected situations, if the performance detection unit at the back end finds that the product does not meet the production specifications, it can immediately find which parameter in the simulation database has the smallest adjustment range and the output value is closer to the target value, and immediately adjust the front-end parameters. For example, first simulate and calculate to adjust any of the main engine speed, overlap rate (coverage, pitch), tension, and outer diameter to see whether the output dielectric constant performance is within the specified standard and the output value is closest to the target value, and then quickly issue the corresponding adjustment control instructions.

[0068] In actual production, the inventors further discovered that it is not necessary to consider all variables, but to focus on several key parameters, and the remaining variables are controlled by improving the process or selected based on empirical data. For example, the density and size of the purchased domestic low-density PTFE insulating tape may have certain numerical fluctuations. In order to ensure the precise controllability of the subsequent process, the tape slitting blade is carefully ground in the present invention. The key point of slitting is the width accuracy, and the accuracy of the tape width is related to the width accuracy of the slitting knife pad. Therefore, the knife pad in the present invention is processed by a precision grinder, and secondary screening is used to ensure the consistency of the slitting film. Therefore, after purchasing the raw materials of the low-density PTFE insulating tape, you only need to pay attention to its material specifications, and its dimensional parameters are all guaranteed to be consistent in value in the slitting process.

[0069] In addition, based on years of production experience, the inventors found that the best effect is achieved when the first layer of wrapping uses a low-density PTFE insulating tape with a width of 4.1mm±0.2mm and a thickness of 3mil, and the second layer of wrapping uses a low-density PTFE insulating tape with a width of 5.5mm±0.2mm and a thickness of 3mil.

[0070] In addition, in order to solve the model quickly and stably, the inventors simplified and optimized the input and output vectors based on historical experience and physical electrical calculation formulas, and adjusted the weights. The most critical improvement is that the material specifications of the PTFE belt are not used as input, but as an influence coefficient to correct the branch weight ω ij , that is, multiplying the branch weight by an empirical correction constant γ. The empirical correction constant γ takes different empirical values ​​based on the material of the PTFE tape. Usually, the PTFE tape of the most commonly used specification of large manufacturers is taken as 1, and other specifications are given an empirical value relative to 1 based on the thickness change after sintering. Therefore, a correction relative to the material specification is added during training to simplify the complexity of the model.

[0071] Based on the above improvements, the present invention proposes a simplified BP neural network model, and trains the BP neural network model through the following steps:

[0072] 1. Data collection and preprocessing

[0073] By collecting parameters such as processing equipment speed, tension, temperature, insulation material density and thickness, and dielectric constant data corresponding to the final product, the data is cleaned and standardized, and finally exported to a CSV file for further processing.

[0074] 2. Construct BP neural network structure

[0075] The present invention adopts a three-layer BP neural network model, which can fit and approximate any function. According to the Kolmogorov theorem (Formula 1) combined with (Formula 2), the final BP neural network structure is 4-9-1, that is, the input layer has 4 nodes, the hidden layer (also called hidden layer or implicit layer) has 9 nodes, and the output layer has 1 node, for a total of three layers.

[0076] S=2n+1(Formula 1)

[0077]

[0078] Where S is the number of hidden layer nodes, m is the number of input layer nodes, n is the number of output layer nodes, a is a constant between 1 and 10, and all values ​​are natural numbers.

[0079] It can be seen from this that Figure 3 , 4 As shown, the BP neural network model structure of the present invention is as follows:

[0080] Input layer: The number of neurons is equal to the number of input parameters, namely, rotation speed, tension, overlap rate, and outer diameter.

[0081] Hidden layer: one, containing 9 neurons, using Sigmoid function; any neuron in the hidden layer satisfies:

[0082]

[0083] Where y is the output value; γ is a constant that is empirically corrected for the material of relatively low-density PTFE insulating tape; ω ij is the weight between the input layer and the hidden layer; x i is the input of the i-th neuron; θ is the threshold;

[0084] Output layer: One neuron that outputs the predicted dielectric constant.

[0085] 3. Model training:

[0086] Training parameter settings: The learning rate determines the amount of weight change produced in each training cycle. A large learning rate may lead to inaccuracy and instability; however, a small learning rate leads to a longer training time, so 0.02 is selected; there are many fuzzy factors influencing processing capacity, so the error should not be too small, so 0.001 is selected; the maximum number of training times is 1000. To avoid local minimum values ​​in gradient descent, the Adam algorithm is used for optimization.

[0087] When the training cycle ends, the global mean square error MSE (Mean Squared Error) = 0.000099778, the BP neural network converges to the set target error, and the training ends. Use 10 sets of verification sample test data to test and generate a comparison chart between the predicted value and the actual value, such as Figure 6 As shown in Figure 2, the fitting correlation coefficient R = 0.99935.

[0088] The BP neural network model algorithm can be programmed, for example, through MATLAB, PYTHON, C language, etc., and can also be generated into an executable program that can be recognized and executed by a single-chip microcomputer, a single-board computer, a programmable logic controller (PLC), a programmable gate array (FPGA), a server, etc., so that it can be used for pre-production simulation, control and error correction in actual production, risk management, post-production evaluation, etc.

[0089] Specifically in the processing process, the following steps can be achieved:

[0090] 1. Collect the expected cable electrical performance parameters provided by the customer;

[0091] 2. Input the task into the trained neural network prediction model and output the corresponding processing parameter prediction value;

[0092] 3. Implement the wrapping machine processing control based on the predicted values ​​of the processing parameters output by the prediction model.

[0093] Based on the above model, the present invention proposes a 1394 cable wrapping method, comprising the following steps:

[0094] A slitting knife pad with precisely controlled size is used to slit the low-density PTFE insulating tape, and the slitting widths are x and y, respectively, where x and y are both positive integers;

[0095] Use low-density PTFE insulation tape with a width of x and a thickness of 3 mil. The main machine of the equipment is reversed to wrap the first layer, with 2 overlapping layers.

[0096] Use low-density PTFE insulation tape with a width of y and a thickness of 3 mil. The main machine of the equipment rotates forward to wrap the second layer, with 2 overlapping layers.

[0097] Among them, the four parameters of x, y, main engine speed, tension, overlap rate and outer diameter in the first layer wrapping process and the second layer wrapping process are input into the wrapping and sintering process BP neural network model constructed by the construction method of the wrapping and sintering process BP neural network model as described above, and the optimal process parameters are determined based on the output value of the wrapping and sintering process BP neural network model to execute the wrapping method of the 1394 cable.

[0098] Furthermore, based on the simplified parameters, the present invention also proposes a 1394 cable wrapping method, comprising the following steps:

[0099] Use low-density PTFE insulation tape with a width of 4.1mm±0.2mm and a thickness of 3mil. The main machine of the equipment is reversed to wrap the first layer, with 2 overlapping layers.

[0100] Use low-density PTFE insulation tape with a width of 5.5mm±0.2mm and a thickness of 3mil. The main machine of the equipment rotates forward to wrap the second layer, with 2 overlapping layers.

[0101] The first layer of wrapping and the second layer of wrapping are reversed and forwarded, and the other process parameters except the width are as similar as possible, so as to ensure that the shrinkage stress of the low-density PTFE insulating tape offsets each other during sintering, and finally shrinks evenly to form a stable dielectric constant;

[0102] Among them, the four parameters of rotation speed, tension, overlap rate and outer diameter in the first layer of wrapping and the second layer of wrapping are optimized by searching the simulation database or the BP neural network model trained by the above method to obtain an optimal solution.

[0103] After the first layer of wrapping, appearance inspection, outer diameter inspection and / or wrapping overlap rate inspection are also carried out; if deformation, breakage, color difference, wrapping, scratches, or foreign particles are found, the appearance is deemed unqualified; if the outer diameter exceeds 0.88±0.03mm, it is considered that the outer diameter does not meet the quality requirements; a high-magnification imager is used to confirm the wrapping overlap rate. If the wrapping overlap rate is not 50%, it is considered that it does not meet the quality requirements.

[0104] After the second layer of wrapping, appearance inspection and / or outer diameter inspection are also carried out; if deformation, breakage, color difference, blistering, scratches, or foreign particles are found, the appearance is unqualified; if the outer diameter exceeds 1.14±0.03mm, it is considered that the outer diameter does not meet the quality requirements.

[0105] The cables wound by the first and second layer wrapping processes are 19 concentric compressed copper stranded conductors of 24AWG and an outer diameter of 0.127mm, plated with a 6μm silver layer.

[0106] Among them, in the first layer wrapping process, the specific process parameters include: a / b is 0.1704 / 0.1858, the main machine speed is 1500rpm, the wrapping direction is reverse, the pay-off / take-up tension is 25%, and the take-up motor speed is 30rpm.

[0107] Among them, in the second layer wrapping process, the specific process parameters include: a / b is 0.20 / 0.23, the main machine speed is 1500rpm, the wrapping direction is forward, the pay-off / take-up tension is 25%, and the take-up motor speed is 30rpm.

[0108] Among them, the low-density PTFE insulating tape with a width of 4.1mm±0.2mm and a thickness of 3mil and the low-density PTFE insulating tape with a width of 5.5mm±0.2mm and a thickness of 3mil come from the PTFE tape slitting process. In the PTFE tape slitting process, the key point of slitting control is width accuracy, and the width accuracy is related to the width accuracy of the slitting knife pad, so the knife pad is processed by a precision grinder, and secondary screening is used to ensure the consistency of the slitting film.

[0109] The present invention also proposes a sintering method for a 1394 cable, comprising the following steps:

[0110] High temperature welding of the 1394 cable wrapped with the sinterable PTFE film is performed at 350° C. to 450° C., preferably 400° C.;

[0111] Among them, the temperature control adopts the RKC temperature controller self-feedback control system to carry out precise temperature control, and the sintering time control uses the equipment transmission coefficient ratio and the main machine speed for precise control.

[0112] The imbalance in communication cables will seriously affect the transmission quality of signals, and the dielectric constant of the insulation layer is directly related to the sintering temperature and the time the wire core is baked in the sintering furnace. Therefore, precise control is required. After the first piece of wrapping and sintering is completed, confirm whether the outer diameter meets 1.24mm±0.03mm. The same inspection process as the previous wrapping process is used, and additional items are added: whether the copper wire is oxidized, yellowed, or blackened; check the core wire color code by comparing the color card, and test items such as water immersion withstand voltage 2000vrms.

[0113] The key point of the sintering process is the matching of sintering temperature and speed. The basic principle of heat sealing is high temperature and high speed.

[0114] Temperature stability control of the heat-melting sealing process of the wrapping layer: The PID power control + heating power temperature curve correction is adopted for the infrared heater of the sintering furnace to ensure high-precision temperature stability during the heat-melting sealing process.

[0115] Speed ​​stability control: adopt crawler traction equipment controlled by stepper motor to achieve stable and consistent control of cable traction speed;

[0116] Among them, the double-sided foam track contact pressure is further used to pull and hold the cable to prevent pressure deformation of the cable.

[0117] The insulation structure changes significantly before and after sintering, and the structural consistency and stability are improved.

[0118] like Figure 5 As shown, the present invention also provides a 1394 cable wrapping control device, comprising:

[0119] Signal acquisition units (only four 1-4 are listed in the figure), which are used to collect the main machine speed, tension, overlap rate and outer diameter parameter values ​​in the first layer wrapping process and the second layer wrapping process respectively;

[0120] Actuators (only four 5-8 are listed in the figure) are used to control the specific values ​​of the main machine speed, tension, overlap rate and outer diameter in the first layer wrapping process and the second layer wrapping process respectively; wherein, the main machine rotation directions of the first layer wrapping process and the second layer wrapping process are opposite, and the corresponding process parameters are the same;

[0121] A control unit 9 receives the signal input of the signal acquisition unit, receives the input of the width x and y of the low-density PTFE insulating tape in the first layer wrapping process and the second layer wrapping process, and determines the optimal process parameters of the wrapping process based on the output value predicted by the wrapping and sintering process BP neural network model 10 constructed based on the construction method of the wrapping and sintering process BP neural network model as described above, or based on the search result of the wrapping and sintering process simulation database 11 established based on the construction method of the wrapping and sintering process simulation database as described above, and drives the actuator to execute the wrapping method of the 1394 cable.

[0122] The wrapping control device controls the wrapping of the wrapping machine by an actuator to wrap the workpiece and perform specific wrapping operations.

[0123] The signal acquisition unit is a signal acquisition device known in the art. For example, for the main engine speed, a speed meter, a speed sensor, a rotary encoder, etc. can be used; for the tension, a belt tension tester, a tension sensor, etc. can be used; for the overlap rate, an optical camera, a CCD width tester, etc. can be used; for the outer diameter, an online diameter tester, a single-axis or double-axis diameter gauge, etc. can be used.

[0124] Preferably, the width x can be selected as 4.1±0.2 mm; the width y can be selected as 5.5±0.2 mm.

[0125] Preferably, the control unit 9 can be, for example, a single chip microcomputer, a single board computer, a programmable logic controller (PLC), a programmable gate array (FPGA), a server, etc. The control unit 9 is not limited thereto, as long as it can store an executable program and execute the wrapping method described above.

[0126] The present invention will be further described below through specific examples. It should be noted that the following examples are only for illustration and are not intended to limit the present invention.

[0127] Example 1

[0128] Embodiment 1 includes a slitting process, a first layer wrapping process, a second layer wrapping process and a wrapping sintering process. The specific parameters of each process are as follows:

[0129] 1. Cutting process

[0130] Table 1 Slitting PTFE tape process table

[0131]

[0132] The PTFE tape is cut and the process design uses two insulation wrappings with tape widths of 4.1 mm and 5.5 mm respectively. A 50% overlap rate is the best process to complete the design.

[0133] 2. The first layer wrapping process (wrapping insulation 1)

[0134] Table 2 Process table of the first layer wrapping process

[0135]

[0136] After the tape is cut, the first layer of wrapping process begins. First, the equipment is powered on for preheating and installation. After production is ready, a low-density PTFE insulating tape with a width of 4.1mm is used for the first layer of wrapping. To prevent the wrapping torque from damaging the copper wire structure, wrapping is done in the opposite direction of the copper wire twist. The main machine of the equipment is reverse.

[0137] The first piece of the first layer of wrapping process needs to be inspected for appearance. There must be no deformation, damage, color difference, package opening, scratches, impurity particles and other appearance defects. A micrometer is used to check whether the outer diameter meets 0.88±0.03mm, a high-power imager is used to confirm whether the wrapping overlap rate is 50%, and a micrometer is used to confirm that the conductor size is within the specified range of the specification. During production, the online diameter gauge is turned on to conduct a full inspection of the production process, and the collected data is stored in a designated folder as a data package for subsequent quality query and process capability analysis.

[0138] 3. Second layer wrapping process (wrapping insulation 2) process

[0139] Table 3 Second layer wrapping process table

[0140]

[0141] The qualified semi-finished products enter the second layer wrapping process. After the equipment is installed and preheated, the semi-finished products of the first layer wrapping process are wrapped with a 5.5mm low-density PTFE insulating tape for the second layer. The direction is opposite to the first layer wrapping, and the main machine of the equipment rotates in the forward direction.

[0142] After the first piece of the second layer wrapping process is completed, the same inspection procedure as the first layer wrapping process is used to confirm the quality. The outer diameter requirement of this process is 1.14±0.03mm. After the second layer wrapping process is completed, the wrapping sintering process begins.

[0143] Low-density PTFE is used for two wrappings to achieve the dielectric thickness required by the design and control the dielectric constant. Wrapping 1 uses reverse rotation and wrapping 2 uses forward rotation to eliminate stress and make the structure more stable.

[0144] 4. Wrapping and sintering process

[0145] Table 4 Wrapping sintering process (wrapping insulation sintering) process table

[0146]

[0147] After the second layer wrapping process is completed, the wrapping sintering process begins. The process design uses sinterable PTFE film for wrapping, and simultaneously uses 400℃ high temperature to weld the wrapped insulation layer. The main function is to seal the wrapped insulation to prevent moisture and other harsh environments from corroding the conductor and causing electrical performance failure, increase the adhesion between the insulation layer and the conductor, and solve the problem of insulation shrinkage during component processing; the PTFE film is wrapped at a high temperature of 400℃, and the overlapping parts are integrated.

[0148] Embodiment 2-5

[0149] The specific steps were the same as in Example 1, and the experiment was repeated four times with the same parameters.

[0150] Example 6

[0151] The specific steps are the same as those in Example 1, except that the widths of the wrapping tapes of the wrapped insulation 1 and the wrapped insulation 2 are 4.3 mm and 5.5 mm, respectively.

[0152] Example 7

[0153] The specific steps are the same as those in Example 1, except that the pay-off / take-up tension is 30%.

[0154] Example 8

[0155] The specific steps are the same as those in Example 1, except that the main engine speed is 800 rpm.

[0156] Example 9

[0157] The specific steps are the same as those in Example 1, except that the overlapping rate is 40%.

[0158] Example 10

[0159] The specific steps are the same as those in Example 1, except that the outer diameter is 1.36 mm.

[0160] Comparative Examples 1 to 6

[0161] The specific steps are the same as those in Example 1, except that some parameters are shown in the following table.

[0162] Performance Testing

[0163] The above experimental results were tested for performance. The test method and results are as follows (the actual test was taken on the impedance between the red and green cores of the four-color core cable after the cabling process was completed. The cabling process used the cabling and overall manufacturing method used in another patent application of the applicant. This method does not affect the impedance test. The final product is as follows Figure 7 shown).

[0164] Performance testing includes differential impedance testing. The differential impedance test uses a time domain reflectometer (TDR) with a bandwidth of not less than 26.5G. The test method includes: the rise time of the TDR is set to no more than 100ps (10% to 90%). Connect the wires in the cable to the time domain reflectometer in the following combination: red-green. It is recommended that the test point be at least 0.61m away from the wire termination, and select the first 4ns-5ns section of the test cable as the cable impedance observation window to reduce the impact of AC impedance and record the measured value.

[0165] Table 5 Comparison of experimental parameters and technical effects of some samples of the present invention

[0166]

[0167] It can be seen that the method of the present invention has simple control parameters, and can achieve precise control based on different batches produced at different times (such as Example 1 (20190919), Example 2 (20190903), Example 3 (20190828), Example 4 (20190826), and Example 5 (20190821)). The differential impedance between the red and green core wires is less than 110±2.6Ω, which meets the technical requirement of differential impedance less than 110±6Ω, and achieves good technical and economic effects. In addition, the parameters of the present invention are optimized based on the BP neural network model or simulation database. When the input parameter combination is not predicted by the neural network model or does not meet the preferred parameter combination value included in the simulation database, for example, as shown in Comparison 1-6, the differential impedance between the red and green core wires is greater than 110±6Ω, which does not meet the technical requirement of differential impedance less than 110±6Ω.

[0168] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for constructing a BP neural network model for a wrapping sintering process, characterized in that: The steps include: The values ​​of main engine speed, tension, pitch and outer diameter are used as inputs, and the dielectric constant value of the product obtained by the wrapping and sintering process is used as output to construct the wrapping and sintering process BP neural network model; wherein, the wrapping and sintering process BP neural network model includes an input layer, a hidden layer and an output layer; wherein, the input layer contains 4 nodes, and the output layer contains 1 node; any neuron in the hidden layer satisfies: Where y is the output value; γ is a constant that is empirically corrected for the material of relatively low-density PTFE insulating tape; ω ij is the weight between the input layer and the hidden layer; x i is the input of the i-th neuron; θ is the threshold; The BP neural network model of the wrapping and sintering process is trained through actual production data to solve the weight and threshold of each node.

2. The construction method according to claim 1, characterized in that: The BP neural network model of the wrapping sintering process uses a Sigmoid function as an error function.

3. The construction method according to claim 2, characterized in that: The BP neural network model of the wrapping sintering process is optimized using the Adam algorithm.

4. A method for establishing a wrapping sintering process simulation database, characterized in that: The steps include: Constructing a wrapping and sintering process simulation database, inputting part or all of the raw material parameter performance, wrapping and sintering process parameters, and product performance test result data in the wrapping production process into the wrapping and sintering process simulation database; A BP neural network model for a wrapping and sintering process constructed by the method for constructing a BP neural network model for a wrapping and sintering process as described in any one of claims 1 to 3 is used to expand the data of the wrapping and sintering process simulation database; wherein the input of the BP neural network model for a wrapping and sintering process includes new parameters that are not stored in the simulation database, and the output of the BP neural network model for a wrapping and sintering process is a predicted value of the BP neural network model for a wrapping and sintering process based on the input including the new parameters.

5. A 1394 cable wrapping method, characterized in that: The steps include: A slitting knife pad with precisely controlled size is used to slit the low-density PTFE insulating tape, and the slitting widths are x and y, respectively, where x and y are both positive integers; Use low-density PTFE insulation tape with a width of x and a thickness of 3 mil. The main machine of the equipment is reversed to wrap the first layer, with 2 overlapping layers. Use low-density PTFE insulation tape with a width of y and a thickness of 3 mil. The main machine of the equipment rotates forward to wrap the second layer, with 2 overlapping layers. Among them, the four parameters of main machine speed, tension, pitch and outer diameter in the first layer wrapping process and the second layer wrapping process are input into the wrapping and sintering process BP neural network model constructed by the construction method of the wrapping and sintering process BP neural network model as described in any one of claims 1 to 3, or, the four parameters of main machine speed, tension, pitch and outer diameter in the first layer wrapping process and the second layer wrapping process are input into the wrapping and sintering process simulation database established by the establishment method of the wrapping and sintering process simulation database as described in claim 4, and the optimal process parameters are determined based on the output value of the wrapping and sintering process BP neural network model or the search result of the wrapping and sintering process simulation database, and the selection of slitting widths x and y is based on years of production experience to execute the wrapping method of the 1394 cable.

6. The wrapping method according to claim 5, characterized in that: The width x is 4.1 mm, the width y is 5.5 mm, and the BP neural network model of the wrapping and sintering process is constructed by the method for constructing the BP neural network model of the wrapping and sintering process as claimed in claim 3; and / or After the first layer of wrapping, appearance inspection, outer diameter inspection and / or wrapping overlap inspection are also carried out; if deformation, breakage, color difference, packaging, scratches, or foreign particles are found, the appearance is considered unqualified; if the outer diameter exceeds 0.88±0.03mm, it is considered that the outer diameter does not meet the quality requirements; a high-power imager is used to confirm the wrapping overlap rate. If the wrapping overlap rate is not 50%, it is considered that it does not meet the quality requirements; and / or After the second layer of wrapping, appearance inspection and / or outer diameter inspection are also carried out; if deformation, damage, color difference, baling, scratches, or foreign particles are found, the appearance is deemed unqualified; if the outer diameter exceeds 1.14±0.03mm, it is considered that the outer diameter does not meet the quality requirements.

7. The wrapping method according to claim 5, characterized in that: The cables wound by the first and second layer wrapping processes are 19 concentric compacted copper stranded conductors of 24AWG and an outer diameter of 0.127 mm, plated with a 6 μm silver layer.

8. A 1394 cable wrapping control device, characterized in that: include: A signal acquisition unit, used to collect the main machine speed, tension, overlap rate and outer diameter parameter values ​​in the first layer wrapping process and the second layer wrapping process respectively; The actuator is used to control the specific values ​​of the main machine speed, tension, overlap rate and outer diameter in the first layer wrapping process and the second layer wrapping process respectively; wherein the main machines of the first layer wrapping process and the second layer wrapping process rotate in opposite directions and have the same corresponding process parameters; A control unit receives a signal input from the signal acquisition unit, receives inputs of widths x and y of the low-density PTFE insulating tapes in the first layer wrapping process and the second layer wrapping process, and determines the optimal process parameters of the wrapping process based on the output value predicted by the wrapping and sintering process BP neural network model constructed according to the method for constructing a wrapping and sintering process BP neural network model as described in any one of claims 1 to 3, or based on the result of searching a wrapping and sintering process simulation database established according to the method for establishing a wrapping and sintering process simulation database as described in claim 4, and drives the actuator to execute the wrapping method of the 1394 cable as described in any one of claims 5 to 7.

9. A wrapping and sintering method for 1394 cables, characterized in that: The steps include: Wrapping the conductive metal cable using the 1394 cable wrapping method as described in any one of claims 5 to 7; High-temperature welding the conductive metal cable wrapped with the sinterable PTFE film at 350° C. to 450° C.; Among them, the temperature control adopts the RKC temperature controller self-feedback control system to carry out precise temperature control, and the sintering time control uses the equipment transmission coefficient ratio and the main machine speed for precise control.

10. The wrapping sintering method according to claim 9, characterized in that: After wrapping and sintering, confirm whether the outer diameter meets 1.24mm±0.03mm; and / or After the wrapping and sintering is completed, at least one of the following inspection processes is carried out: whether the copper wire is oxidized, yellowed or blackened; check the core wire color code by comparing the color card; carry out the water immersion voltage withstand test item of 2000vrms; and / or In the wrapping sintering process, the infrared heater of the sintering furnace is controlled by PID power, and the temperature of the heating power is calibrated by curve to ensure high-precision temperature stability during the heat sealing process; and / or In the wrapping and sintering process, a stepper motor is used to control the crawler traction equipment to achieve stable and consistent control of the cable traction speed.

Images