JP2023176923A - Method for generating molding conditions for injection molding systems, injection molding systems, and programs for supporting generation of molding conditions for injection molding systems - Google Patents

Abstract

To provide a method for generating molding conditions for an injection molding system, an injection molding system, and a molding condition generation support program for an injection molding system that can generate molding conditions during injection compression molding or core back molding relatively more easily than before.SOLUTION: An injection molding machine 11 is provided with a molding die 41 in which a cavity C is formed between a fixed mold 15 and a movable mold 17, a mold clamping device 13 to which the molding die 41 is attached, and an injection device 14 that injects into the cavity C. A learning model for injection compression molding or core-back molding generates molding conditions for injection compression molding or core-back molding from information about the injection molding machine 11, information about the molding die 41, and information about the resin material supplied to the injection device 14.SELECTED DRAWING: Figure 1

Description

The present invention relates to a method for generating molding conditions for an injection molding system equipped with an injection molding machine that moves a movable mold relative to a fixed mold during injection compression molding or core-back molding, an injection molding system, and an injection molding system. This relates to a molding condition generation support program.

Conventionally, methods described in Patent Document 1 and Patent Document 2 are known as related to setting molding conditions for injection compression molding. Patent Document 1 describes that the injection filling amount in injection compression molding is corrected when the mold position when mold clamping pressure is removed deviates from an arbitrary set value set with respect to a reference point. . Further, Patent Document 1 describes that the reference point is determined in advance based on the product shape, etc. Furthermore, Patent Document 2 discloses that, at the stage of test injection, the molding conditions for a non-defective product are determined between the injection speed of the molten resin and the speed in the direction of increasing the cavity volume of the mold, and conditions that meet the above-mentioned molding conditions for the non-defective product are determined. It is stated that molding is performed using

Japanese Patent Application Publication No. 8-174616 Japanese Patent Application Publication No. 4-246523

However, when setting the molding conditions for injection compression molding in Patent Document 1 and Patent Document 2, the type of resin material is not considered at all. In actual injection compression molding, the temperature of the heating cylinder differs depending on the type of resin material, and as a result, the behavior such as the fluidity of the molten resin material flowing into the cavity also differs. Therefore, the molding conditions for injection compression molding cannot be unconditionally determined. Furthermore, although there is a description in Patent Document 1 that suggests that the product shape is reflected in the molding conditions, information on other molds for the product shape is also required when setting the molding conditions for actual injection compression molding. Therefore, even if the molding conditions are set to reflect the product shape, the molding conditions for injection compression molding cannot be determined unconditionally. Therefore, the reality is that the setting of molding conditions for injection compression molding is still carried out by experienced operators through repeated trial and error. Similarly, the reality is that core-back molding is still carried out by experienced operators through repeated trial and error.

Therefore, the present invention provides a method for generating molding conditions for an injection molding system, an injection molding system, and a method for generating molding conditions for an injection molding system, which can relatively easily generate molding conditions for injection compression molding or core-back molding. The purpose is to provide a condition generation support program. Other objects and novel features will become apparent from the description of this specification and the accompanying drawings.

The method for generating molding conditions for an injection molding system according to claim 1 of the present invention is an injection molding system equipped with an injection molding machine that moves a movable mold with respect to a fixed mold during injection compression molding or core-back molding. A method for generating molding conditions for an injection molding system equipped with an injection molding machine that moves a movable mold with respect to a fixed mold during injection compression molding or core-back molding, the method comprising: An injection molding machine includes a molding mold in which a cavity is formed between a fixed mold and a movable mold, a mold clamping device to which the molding mold is attached, and an injection device for injecting into the cavity, and at least The learning model for injection compression molding or core-back molding is based on the information regarding the injection molding machine, the information regarding the molding die, and the information regarding the resin material supplied to the injection device. generate.

A method for generating molding conditions for an injection molding system according to the present invention is a method for generating molding conditions for an injection molding system equipped with an injection molding machine that moves a movable mold with respect to a fixed mold during injection compression molding or core-back molding. An injection molding machine includes a molding mold in which a cavity is formed between a fixed mold and a movable mold, a mold clamping device to which the molding mold is attached, and an injection device for injecting into the cavity. The learning model for injection compression molding or core-back molding is based on at least information regarding the injection molding machine, information regarding the molding die, and information regarding the resin material supplied to the injection device. Since the molding conditions for molding are generated, the molding conditions for injection compression molding or core-back molding can be generated relatively easily than before.

FIG. 1 is a schematic explanatory diagram of an injection molding machine of an injection molding system according to the present embodiment. FIG. 2 is an enlarged view of the main parts of the mold clamping device and the molding die of the injection molding machine of the present embodiment, showing the mold in an open state. FIG. 2 is an enlarged view of the main parts of the mold clamping device and the molding die of the injection molding machine of the present embodiment, showing the state at the start of injection compression molding. FIG. 2 is an enlarged view of the main parts of the mold clamping device and the molding die of the injection molding machine of the present embodiment, showing the state at the end of injection compression molding. 3 is a view taken along line AA in FIG. 2. FIG. It is a block diagram of a machine learning device of an injection molding system of this embodiment. FIG. 2 is a schematic diagram of a screen for inputting information regarding a molding die in the input device of the injection molding machine of the injection molding system of the present embodiment. It is a schematic diagram of the screen which inputs the information regarding a molding defect regarding the input device of the injection molding machine of the injection molding system of this embodiment. It is a diagram showing a neural network of a learning model of the machine learning device of the injection molding system of the present embodiment. FIG. 3 is a flowchart regarding generation of molding conditions using a learning model of the machine learning device of the injection molding system and correction of the learning model according to the present embodiment.

Hereinafter, specific embodiments will be described in detail with reference to the drawings. However, the present invention is not limited to the following embodiments. Further, in order to clarify the explanation, the following description and drawings are appropriately simplified. In addition, hatching is omitted in some parts to avoid cluttering the drawings.

<Injection molding machine>
An injection molding system 1 including an injection molding machine 11 of this embodiment will be described with reference to FIGS. 1 to 5. First, an injection molding machine 11 that moves a movable mold relative to a fixed mold during injection compression molding or core-back molding will be described with reference to FIG. The injection molding machine 11 includes a mold clamping device 13 and an injection device 14 on a bed 12. To explain the mold clamping device 13 first, the mold clamping device 13 includes two mold opening/closing mechanisms 19 that move a movable platen 18 to which a movable mold 17 is attached to a fixed platen 16 to which a fixed mold 15 is attached. It is equipped with mold clamping cylinders 21 of four mold clamping mechanisms 20 that clamp the fixed mold 15 and the movable mold 17 (however, in FIG. 1, only the mold opening/closing mechanism 19 and the mold clamping mechanism 20 on the near side are shown). ).

A mortar part 23 for inserting the nozzle 22 of the injection device 14 is provided at the center of the stationary platen 16 fixed on the bed 12 on the side opposite to the mold mounting surface. A hole is provided in the mold 15 to connect the nozzle 22 thereto. Further, mold clamping cylinders 21 of a mold clamping mechanism 20 are provided near the four corners inside the fixed platen 16, respectively. In the mold clamping cylinder 21, a rod on the forward side of the piston 24 constitutes a tie bar 25. The mold clamping cylinder 21 is a double-acting cylinder that includes a mold clamping oil chamber on the forward side of the piston 24 and a mold opening oil chamber on the backward side of the piston 24. The mold clamping cylinder 21 is also connected to a hydraulic system 26 that includes valves, sensors, pumps, tanks, and the like. In the present invention, the mold clamping cylinder 21 is controlled by a valve capable of closed loop control, such as a servo valve or a flow control valve.

A plurality of engagement grooves 25a are formed on the outer periphery of each tie bar 25 near the tip side over a predetermined length in the mold opening/closing direction. Each tie bar 25 is inserted into an insertion hole 29 provided near the four corners of the movable platen 18, respectively. A half nut 30 serving as an engagement mechanism is disposed near the insertion hole 29 on the surface of the movable platen 18 opposite to the mold mounting surface. In the half nut 30, the engagement teeth 31 can be moved forward and backward toward the engagement groove 25a of the tie bar 25 by a drive mechanism 28 such as a cylinder. The half nut 30 is capable of engaging the movable plate 18 with the tie bar 25 when the engagement teeth 31 move forward. The movable platen 18 also includes an ejector mechanism (not shown).

Two mold opening/closing mechanisms 19 are disposed on the bed 2 to move the movable platen 18 toward and away from the fixed platen 16. The mold opening/closing mechanism 19 uses a servo motor 27 and a ball screw mechanism 32. A rotary encoder (not shown) of the servo motor 27 measures the distance between the movable platen 18 and the fixed platen 16 (the distance between the movable mold 17 and the fixed mold 15). Note that the distance of the movable platen 18 with respect to the fixed platen 16 may be measured by a position detection mechanism such as a linear scale other than the rotary encoder.

The mold clamping device 13 includes four tie bar moving mechanisms 33 that move the tie bars 25 a certain distance apart from the mold clamping cylinder 21, corresponding to the number of tie bars 25 (however, only two are shown in FIG. 1). ). The tie bar moving mechanism 33 includes two hydraulic cylinders 34 for each tie bar 25, and the rod of the hydraulic cylinder 34 and the rod 35 that protrudes from the piston 24 of the mold clamping cylinder 21 toward the side opposite to the mold mounting surface. They are connected by a connecting plate 36. The hydraulic cylinder 34 is controlled by a closed-loop controllable valve 37 provided in the hydraulic system 26 . Further, a position detection mechanism 38 for detecting the moving position of the tie bar 25 is attached to the fixed platen 16. Note that the tie bar moving mechanism 33 may be operated by a servo motor.

Note that the mold clamping device is not limited to that shown in FIG. 1, and may be one using a toggle mechanism. Although illustrations of those using a toggle mechanism are omitted, a pressure receiving plate is provided on the back side of the movable platen, and each toggle link of the toggle mechanism is provided by connecting the movable platen and the pressure receiving plate. Further, a mold clamping mechanism such as a servo motor is provided on the back side of the pressure receiving plate, and is capable of rotationally driving a ball screw pivotally supported on the pressure receiving plate. Further, the ball screw is inserted through a ball screw nut, and the ball screw nut is fixed to the crosshead. A toggle link is rotatably attached to the crosshead. In such a toggle mechanism, the amount of movement of the movable platen near the completion of mold closing is smaller than the amount of movement of the crosshead, and the movement of the movable platen can be controlled with high precision by the servo motor.

Next, the molding die 41 attached to the mold clamping device 13 will be explained with reference to FIGS. 2 to 5. The molding die 41 is used for injection compression molding, and a cavity C is formed between the fixed die 15 and the movable die 17. The fixed mold 15 is a cavity type, and includes a cavity surface 42 consisting of a concave portion and a sliding surface 43 on which a movable mold 17, which will be described later, slides. Further, a parting surface 44 facing the movable mold 17 is formed around the outer side of the sliding surface 43. Further, the fixed mold 15 is formed with a flow path 45 through which molten resin flows during injection (injection filling), and a gate 46 is formed in the cavity surface 42 portion. The flow path 45 may be a hot runner or a cold runner. Further, the fixed mold 15 is provided with a temperature regulating flow path 47 through which a temperature regulating medium flows.

The movable mold 17 is a core type, and the tip end surface of the convex portion serves as a cavity surface 48. Further, the side surface of the convex portion is a sliding surface 49 that faces the sliding surface 43 of the fixed mold 15 with a very small interval when the mold is closed. Further, the surface facing the fixed mold 15 around the outer side of the sliding surface 49 is a parting surface 50 . However, the parting surface 44 of the fixed mold 15 and the parting surface 50 of the movable mold 17 do not necessarily come into contact when molding is completed. Furthermore, the movable mold 17 is provided with a temperature control flow path 51 through which a temperature control medium flows and an ejector 52 for ejecting the molded product P therein. FIG. 5, which is a view taken along the line AA in FIG. 2, shows the cavity surface 48 and the like of the movable mold 17. Injection pressure during injection is applied to the movable mold 17 and the movable platen 18 according to the area of the cavity surface 48.

The molding die 41 is also called a pilot die due to its structure. FIG. 3 shows the state of the molding die 41 at the start of injection compression molding. Further, FIG. 4 shows the state of the molding die 41 at the end of injection compression molding. As can be seen from FIGS. 3 and 4, when the movable mold 17 is moved in the mold closing direction relative to the fixed mold 15, the volume of the cavity C is reduced and the molten resin inside is compressed. It is something that will be done. Although the molding die 41 is rarely used as both an injection compression molding mold and a core-back molding mold, the basic structure is the same even when the molding mold is a core-back molding mold. That is, in the case of a core-back mold, injection is started in the state shown in FIG. 4, the volume of the cavity is expanded by foaming of the molten resin in the cavity, and molding is completed in the state shown in FIG. 3.

Next, the injection device 14 will be explained. The injection device 14 includes a screw (not shown) inside a heating cylinder 54, and a nozzle 22 is attached to the front of the heating cylinder 54. The injection device 14 is also equipped with a drive mechanism 55 on the rear side, which rotates the screw and moves the heating cylinder 54 back and forth in the axial direction. Further, a block or plate (not shown) to which the heating cylinder 54 of the injection device 14 is attached is provided with a material supply mechanism 53 having a hopper and the like. The entire injection device 14 can be moved forward and backward by a nozzle touch device (not shown), and when moving forward, the tip of the nozzle 22 touches the fixed mold 15. Note that the structure and number of injection devices 14 are not particularly limited.

<Peripheral equipment such as control equipment and inspection equipment>
Next, the control device 61 of this embodiment will be explained with reference to the block diagram of FIG. 6. The injection molding machine 11 is equipped with a control device 61. The control device 61 controls the injection molding machine 11 and includes a calculation section, a storage section, an input section, an output section, etc. (not shown). The control device 61 is connected to the hydraulic system 26 of the injection molding machine 11, the servo amplifier 62 of the servo motor 27 of the mold opening/closing mechanism 19, the servo amplifier 63 of the injection servo motor and metering servo motor (not shown) of the injection device 14, and the input device 64. It is connected. Furthermore, the control device 61 is also connected to each sensor of the injection molding machine 11. Furthermore, the control device 61 also includes peripheral devices and their control devices, such as a control device for a temperature control device 65, an inspection device 66, a molded product removal device 67, and a control device for a material supply device (not shown) connected to the material supply mechanism 53. It is connected.

The inspection device 66 is equipped with a camera 68 such as a CCD camera, and is capable of reading information on the molded product P. Here, the position where the camera 68 of the molded product take-out device 67 reads the molded product P may be in a state where it is held by the molded product take-out device 67 or at another location such as a conveyor. In addition, the inspection device 66 may be a weight measuring device for measuring the weight of the molded product P, a color measuring device for measuring the surface condition of the molded product, a tension measuring device for measuring the strength of the molded product, a load measuring device, etc. good. In addition, complementary inspection equipment includes a position sensor for detecting the position of the screw when pressure retention is completed, and a position sensor for detecting the position of the movable platen 18 or movable mold 17 at the end of cooling during injection compression molding. 66.

<Machine learning device>
Further, in the injection molding system 1 of this embodiment, a control device 71 equipped with a machine learning device 70 that constitutes a molding condition generation support program is provided separately from the control device 61 of the injection molding machine 11. The control device 71 may be connected to a plurality of injection molding machines 11 as a central control device of the factory, or may be installed in another structure within the same company outside the factory, or in an injection molding machine manufacturer or management company. But that's fine. In these, the injection molding machine 11 and the control device 71 may be connected by wire or wirelessly. Note that the machine learning device 70, which is the molding condition generation support program, may be provided in the control device 61 of the injection molding machine 11.

A control device 71 that controls the plurality of injection molding machines 11 includes an input section 72 . The input section 72 is connected to the control device 61 of the injection molding machine 11, the inspection device 66, etc. Information about the injection molding machine 11 and the inspection device 66 is then input. The control device 71 also includes a machine learning device 70. The machine learning device 70 includes a learning model section 73. The learning model section 73 includes an estimation section 74 , a learning model generation/modification section 75 , and a learning model storage section 76 . The estimation unit 74 generates molding conditions for injection compression molding using a neural network. Further, the learning model generation/correction unit 75 corrects the learning model by backpropagation or the like when the molded product generated under the molding conditions is not optimal and the molding conditions are corrected. Further, the learning model storage unit 76 stores generated or modified molding conditions.

Further, the learning model section 73 of the machine learning device 70 is connected to a molding condition storage section 77, an injection molding machine information storage section 78, a mold information storage section 79, and a resin material information storage section 80. The molding condition storage section 77 stores molding conditions for basic patterns used in injection compression molding or core-back molding from the time of machine shipment or the time of installation of the control device. Furthermore, newly generated molding conditions corresponding to the molding die 41 are also stored. The injection molding machine information storage section 78 stores information necessary for generating molding conditions for all the injection molding machines 11 connected to the control device 71. Specifically, information (spec information) on the mold clamping mechanism of the mold clamping device 13, information (spec information) on the drive unit of the injection mechanism including the injection servo motor and the metering mechanism including the metering servo motor of the injection device 14, This information includes information on the screw of the injection device 14 (spec information), information on the heater and cylinder inner diameter of the heating cylinder 54 (spec information), and the like. Further, the molding mold information storage section 79 stores information about each molding mold 41 mounted on the injection molding machine 11. Further, the resin material information storage section 80 stores information regarding basic resin materials and information regarding resin materials used in the injection molding machine 11 in the past.

The machine learning device 70 is also provided with a molding condition confirmation/correction section 81 , and the molding condition confirmation/correction section 81 is connected to the learning model section 73 and the molding condition storage section 77 . Then, in the molding condition confirmation/correction section 81 of the machine learning device 70, the molding conditions are corrected when the operator presses the defect type button 124 based on the molding result. Alternatively, the operator may directly modify at least one of the molding condition items generated by the estimation unit 74 of the learning model unit 73 by manual input or the like. The machine learning device 70 is also connected to an output section 82. The output section 82 is connected to the control device 61 of the injection molding machine 11.

<Input device and input screen>
Next, the input device 64 and its input screen 91 for setting the injection molding machine 11 will be explained with reference to FIGS. 7 and 8. The input device 64 provided on the side surface of the fixed platen 16 of the injection molding machine 11 is used by the operator to set and input molding conditions and to display the operating status of the injection molding machine 11, and is composed of a touch panel or the like. The input device 64 is capable of inputting setting values of molding conditions regarding the injection molding machine 11 by switching between pages, and can also display actual measured values and production information during molding of the injection molding machine 11. In the injection molding system 1, it is also possible to provide a handy type input device that the operator holds in his hand and uses around the injection molding machine 11, separately from the input device 64 that is fixed to the injection molding machine 11. It is included in the input device of the present invention.

The input device 64 includes a general screen (not shown) for setting and inputting various molding conditions as described above, as well as information regarding the molding die 41 and information supplied to the injection device 14 as shown in FIG. An input screen 91 is provided for inputting information regarding the resin material used. To explain the input screen 91, as information on the molding die 41 including the molded product P, there is an input section 92 for the projected cavity area (cavity surface 48 in FIG. 5), an input section 93 for the cavity volume at the end of molding, and an input section 93 for the cavity length. Input section 94 for flow length (flow length from the gate to the farthest end), input section 95 for molded product weight, input section 96 for molded product volume, input section 97 for inputting the type of molded product, molded product An input section 98 for inputting the number of pieces (number of pieces), and an input section for indicating the presence or absence of merging in the cavity during injection (indicates whether the flow of resin injected into the cavity merges in the middle, and is necessary for weld prevention). 99 are displayed all at once.

Furthermore, in relation to the cavity C and the molded product P, the input screen 91 includes an input section 100 for the average thickness A of the molded product, an input section 101 for the thickness of the thick part of the molded product, and a selection of the shape of the molded product, as shown in FIG. and an input section 103 for inputting compression stroke B during injection compression molding. Regarding the runner part where the injected molten resin is sent to the cavity, there is an input field 104 for the length of the runner R, an input field 105 for the cross-sectional area (the narrowest part) of the runner R, and the type of the runner R (hot runner or cold runner). runner) input section 106, input section 107 for the number of gates G, input section 108 for selecting and inputting the position of the gate G (center, end, or between the center and end), cross-sectional area of gate G It is provided with an input section 109 for inputting. Furthermore, the input screen 91 includes an input section 110 for inputting the mold weight, the type of mold (such as a spigot mold for injection compression molding or a flat mold for injection compression molding as shown in FIG. 11), It is provided with an input section 111 for inputting a heating/cooling mold, etc., and an input section 112 for inputting an estimated cooling time.

In this embodiment, the input screen 91 includes an input section 113 for selecting and inputting the type of resin such as polypropylene, polycarbonate, acrylic (PMMA), etc., for inputting information regarding the resin material supplied to the injection device; Input section 114 for selecting and inputting the grade (if the same resin material needs to be subdivided and input), input section 115 for inputting composite materials such as glass fiber or carbon fiber, and input section 115 for inputting composite materials such as glass fibers and carbon fibers; ). Further, if the resin manufacturer's recommended temperature for injection or molding is specified in the grade section of the resin material, it may be possible to input it.

The information regarding the molding die 41 input from the input screen 91 includes either the cavity projected area or the molded product area, the cavity volume, the molded product weight, or the molded product volume, and whether the number of molded products is 2 or more. In this case, the number of molded products is a required input item. The type of resin is also an essential input item.

Note that in FIG. 7, the input section for information regarding the molding die 41 including the molded product P and the input section for information regarding the resin material supplied to the injection device 14 are illustrative, and items may be increased or decreased other than those shown. . Furthermore, these input sections allow molding conditions to be generated even if all the input section items are left blank. Further, in the input screen 91 of FIG. 7, information regarding the molding die including the molded product and information regarding the resin material supplied to the injection device are input from the same input screen 91, but they may be input from separate input screens. Further, the information regarding the molding die 41 or the information regarding the resin material may be read by wired or wireless communication means, a QR code (registered trademark), a bar code, or the like. At this time, the items of information regarding the molding die 41 including the molded product and information regarding the resin material may be the same as the information input from the input screen 91, or the number of items may be increased or decreased.

In this embodiment, information regarding the injection molding machine 11 is sent from the injection molding machine 11 to the injection molding machine information storage section 78 of the control device 71 and is stored therein. However, if the injection device 14 or other mechanisms used by the injection molding machine 11 are changed depending on the molding die 41, information regarding the injection molding machine 11 may be input from the input screen 91.

Although not shown, another screen may be provided as an input screen for core-back molding in which the movable mold retreats during foam molding. In that case, the part that is the input part 95 for the compression stroke B in FIG. 7 becomes the input part for the core back amount (foaming expansion stroke). Also provided is an input section for inputting the type of foaming, chemical foaming or physical foaming. In the case of chemical foaming, the composite material input section 107 serves as an input section for the type of foaming agent and the ratio (% by weight) of the foaming agent. In the case of physical foaming, an input section is provided for inputting the position of the gas supply section, gas type, gas pressure, etc. when supplying foaming gas etc. to the heating cylinder of the injection device or the cavity of the mold.

The input device also includes a molded product status input screen 121 for inputting information regarding the molded product P molded by injection compression molding. FIG. 8 is an example of the molded product status input screen 121. If the molded product P is good, the operator can input that it is a good product by pressing the good product button 122 and then pressing the confirm button 123. There is. In addition, if the molded product P has a defect, the operator presses the defect type button 124 indicating the type of defect of each molded product P, and then presses the confirm button 123 to confirm that it is defective and to input the type. It has become. The defect type buttons 124 on the molded product status input screen 121 are short shot (insufficient weight), overpack (overweight), weight variation, burr, sink mark, deformation (including warping), burnt, flow mark, void, and silver. There are input buttons for streaks, cracks, welds, other surface defects, contamination, gate defects, ejector marks, and other surface defects.

Therefore, if the operator determines that the molded product P is a good product, he can input information that the molded product P is a good product by pressing the good product button 122 and the confirmation button 123 in sequence. Furthermore, if the operator determines that the molded product P has a short shot, he can input information that the molded product P has a short shot by pressing the short shot button, which is the defect type button 124, and the confirm button 123 in that order. can. If the molded product P has two or more types of defects, the user presses the multiple defect type buttons 124 and then presses the confirm button 123. Furthermore, regarding the degree of defect, by pressing the defect type button (same button) such as short shot multiple times, the color of the defect type button 124 will change color, and then by pressing the confirm button 123, the defect type button 124 will change color. The degree of this is now input.

Note that the degree of defect may be input from a separately provided numeric button. Furthermore, the types of defects that frequently occur, such as short shots (insufficient weight), overpacks (overweight), weight variations, burrs, sink marks, and deformations (including warping), can be fixed even if the size of the button is increased. Conversely, for defect types with a low frequency of defect occurrence, by pressing the "Others" button, the user may be moved to a separate input screen for the defect types with a low frequency of defect occurrence. Alternatively, the defect type buttons 124 may be arranged and displayed on the input screen 121 in order from the most frequently pressed defect type buttons 124. Then, when the operator who sees the molded product P presses one of the defect type buttons 124 and the confirm button 123 in sequence on the input screen 121 for the molded product status, the molded product Information on the type of defect in P can be transmitted.

<Molding condition generation support program for injection molding system>
In this embodiment, a molding condition generation support program for the injection molding system is stored in an external control device 71 that is separate from the injection molding machine 11. More specifically, the control device 71 includes a machine learning device 70, and a learning model section 73 of the machine learning device 70 stores a learning model (generation support program) regarding injection compression molding or core-back molding. As shown in FIG. 9, the learning model of this embodiment uses a neural network 131 provided with one or more intermediate layers 133 (hidden layers). When the neural network 131 of the learning model receives at least information about the injection molding machine 11, information about the molding die 41, and information about the resin material supplied to the injection device 14 from the input layer 132, the neural network 131 of the learning model sets the weight w for each of them. A bias b is added as necessary in the intermediate layer 133 using a function equation to which . Note that in FIG. 9, the intermediate layer 133 is schematically depicted, and in reality, more complicated functional expressions, weights w, and biases b are used. Further, although only one intermediate layer 133 is shown in FIG. 9, two or more intermediate layers 133 may be provided to perform deep learning. Furthermore, the neural network 131 may use a convolutional neural network at least in part.

The learning model may be stored in the learning model unit 73 of the first machine learning device 70 by downloading it through the Internet or installing it on a mobile medium. In this embodiment, the first learning model regarding injection compression molding is created by an injection molding machine manufacturer. The molding condition storage unit 77 also stores molding conditions under which a good product was molded by linking the past data to a combination of specific injection molding machine information, specific molding die information, and specific resin material information. Many are stored. Further, the injection molding machine information storage section 78 stores information on the injection molding machine 11 in advance, the molding mold information storage section 79 stores information on the molding die 41 in advance, and the resin material information storage section 79 stores information on the molding die 41 in advance. Information regarding the resin material is stored in the section 80 in advance. Note that the information on the injection molding machine 11 and the information on the mold 41 are stored in the control device 61 of the injection molding machine 11, and the control device 71 of the central processing unit generates molding conditions using a learning model. It may also be sent via communication.

<Generation or modification of molding conditions using a learning model related to injection compression molding>
Generation of molding conditions using the learning model of the machine learning device 70 of the injection molding system 1 will be explained using the flowchart of FIG. The flowchart in FIG. 10 does not only show the control of the control devices 61 and 71, but also includes the operations performed by the operator. Information about the injection molding machine 11 is stored in the injection molding machine information storage section 78 of the control device 71 from the beginning. First, information about the mold 41 used for injection compression molding and information about the resin material are input from the input device 64 connected to the control device 61 or the input device (not shown) connected to the control device 71 (s1).

Next, it is determined whether the molding conditions related to injection compression molding are stored in the molding condition storage unit 77 from the beginning (s2), and if the molding conditions are stored (s2=Y), the learning model unit 73 estimates The learning model unit 73 of the machine learning device 70 adds information about the injection molding machine 11, information about the molding die 41, and information about the resin material supplied to the injection device to the stored molding conditions. is input from the input layer 132 of the neural network 131 stored in the learning model storage unit 76 of Output values y1 and y2 are output to generate molding conditions (s3).

Specifically, the molding conditions for performing injection compression molding with the injection molding machine 11 are stored in the molding condition storage section 77 in association with information on the injection molding machine 11, information on the molding die 41, and information on the resin material. ing. Here, the molding conditions refer to various setting values for obtaining a good molded product by performing injection compression molding using the injection molding machine 11. As described above, if past molding conditions with similar information on the injection molding machine 11, molding die 41, and information on the resin material are stored, the past molding conditions are also input from the input layer. Regarding past similar molding conditions, multiple similar molding conditions may be extracted according to the prioritized items, and the machine learning device can extract similar molding conditions using methods such as clustering or supervised learning. Good too. In addition, when past similar molding conditions are not used, the information on the injection molding machine 11 stored in the injection molding machine information storage section 78 and the information on the molding die 41 stored in the molding mold information storage section 79 Or, using the information about the molding die 41 input from the input device 64, the information about the resin material stored in the resin material information storage section 80, or the information about the resin material input from the input screen 91 of the input device 64. Generates molding conditions for injection compression molding.

At this stage, the learning model used in the estimation unit 74 of the learning model unit 73 of the machine learning device 70 uses what function formula and weighting to perform calculations from the input layer 132 to the intermediate layer 133 and from the intermediate layer 133 to the output layer 134. It is not easy to determine from the outside whether this is the case. However, the rough standards are shown below. First, the approximate injection amount is determined from the weight of the molded product or the volume of the molded product (cavity volume), the number of molded products, the length and cross-sectional area of the runner (if the flow path 45 is a cold runner, the runner), and the information from the injection molding machine 11. The injection stroke (from the metering completion position (injection start position) to the pressure holding completion position, or from the injection start position to the pressure holding switching position) is determined from a certain screw diameter (or heating cylinder inner diameter).Next, the molded product shape (cavity shape ), the runner shape, the gate shape, and the resin material used determine the injection speed and injection pressure.The temperature of the heating cylinder 54 differs depending on the resin material used, and as a result, the temperature of the heating cylinder 54 differs depending on the resin material used. The rheological properties also differ, and the injection speed at which burn occurs also varies, which greatly affects the injection speed and injection pressure.Also, when the shape of the molded product (cavity shape) is thin or the fluidity of the resin material when melted. If the quality is poor, this may be the reason for the adoption of injection compression molding, which will affect the injection speed and injection pressure.

Once the resin material, injection speed, and injection pressure are determined, the holding force of the movable mold during injection by the mold clamping device, the position of the movable platen (movable mold), and the compression molding performed after injection are determined. The pressure (force) or compression speed (speed of movement of the movable mold) is determined. If the molded product shape (cavity shape) is thin or the resin material has poor fluidity when melted, the movable platen 18 (or movable mold 17 ) is set to be thicker than the actual plate thickness of the molded product P, and set so as to be retracted to a predetermined position. That is, molding is performed, which is called an injection press, in which the cavity volume at the start of molding is larger than the cavity volume at the end of molding. As mentioned above, in the case of injection compression molding, there is a causal relationship between the molding conditions on the mold clamping device 13 side and the molding conditions on the injection device 14 side, and if one set value is changed, the other set value also needs to be changed. There are many. Therefore, generation of molding conditions by an operator requires skill and takes time even for an expert. In order to solve the above problem, it is useful to use a machine learning device 70 and its learning model that can remember and reflect the quality of past molding conditions. Since the learning model for injection compression molding has complex mutual causal relationships, a learning model that adjusts the mutual relationships, such as mathematical programming, may be incorporated, for example.

Furthermore, since the molding conditions on the mold clamping device 13 side and the molding conditions on the injection device 14 side are closely related, the learning model is It may also be used to set or modify molding conditions. Alternatively, the learning model may set or modify the molding conditions on the mold clamping device 13 side with respect to the molding conditions on the injection device 14 side input or selected by the operator. For example, in the case of an optical molded product having a thick part such as a lens molding, the injection speed or injection pressure is set relatively low in order to prevent deterioration of the resin material such as burning. On the other hand, on the mold clamping device 13 side, the molten resin in the cavity C is compressed at a speed higher than a predetermined speed to spread the molten resin in the cavity C. The estimated cooling time is set in relation to the temperature of the molding die (temperature control medium temperature). The temperature of the molding die (temperature control medium temperature) is determined to some extent by the type of resin material, but is also influenced by the shape of the mold and cavity. It is possible to input the cooling temperature first in the form of the assumed cooling temperature, but it is also possible to generate (estimate) the cooling time as part of the molding conditions based on the mold shape, cavity shape, and type of resin material. good.

Furthermore, in the present invention, initial molding conditions can be generated even if no molding conditions are stored at all. However, it can be said that it is rare that a good product is molded by injection compression molding or the like under the initial molding conditions. That is, if the molding conditions are not saved, the estimation unit 74 of the learning model unit 73 uses the injection molding machine information stored in the injection molding machine information storage unit 78, the input information regarding the molding die, and the information regarding the resin material. The learning model generates forming conditions using only the information (s4). For example, when information regarding at least one molding die such as the projected cavity area, cavity volume (molded product volume), or molded product weight, and information regarding the resin material are input from the input screen 91, initial molding is performed using a learning model. A condition is generated (estimated).

<Modification of learning model regarding injection compression molding>
Next, a portion of operating the injection molding machine 11 to actually perform injection compression molding to obtain information regarding the molded product P and modifying a learning model related to injection compression molding or core-back molding will be described. Continuing the explanation along the flowchart of FIG. 10, even if the estimating unit 74 of the learning model unit 73 stores the molding conditions and generates the molding conditions (s3), the molding conditions are not stored. Even when the molding conditions are generated (s4), the injection molding machine 11 is then operated manually or semi-automatically (other than continuous automatic molding) to mold the molded product P (s5). In injection compression molding, after molten resin is injected from the injection device 14 into the cavity C during molding, the movable mold 17 is moved relative to the fixed mold 15 to compress the molten resin within the cavity C. At this time, molding conditions with a smaller injection amount than the injection amount under the set molding conditions are automatically generated, and short shot molding is performed under the molding conditions with a smaller injection amount. Then, the operator ejects the actually molded product P with the ejector 52, takes it out from the movable mold 17, etc., visually checks it himself, and if it is a short shot, a "short shot (insufficient weight)" is displayed on the input screen. When the defect type button 124 is pressed and the confirm button 123 is pressed, the molding conditions are changed and the injection amount is increased. At this time, the increase in the injection amount is not uniformly increased by a fixed amount, but is determined by a predetermined ratio with respect to the previous injection amount, molded product weight, etc.

When the short shot is resolved, the operator inspects the molded product P taken out and determines whether it is a good product or a defective product (s6). It is unlikely that a good product can be molded from the beginning, and if the operator inspects the molded product P taken out and determines that there are "sink marks," "burn marks," "flow marks," "silver streaks," or other defects ( s6=N), presses the corresponding defect type button 124 from the input screen 91 of the input device 64 (s7), and then presses the confirm button 123 to improve the molding conditions. At this time, the corresponding defect type button 124 may be pressed multiple times depending on the degree of defect.

The learning model storage unit 76 stores information about which molding conditions should be corrected and how much when one of the defect type buttons 124 is pressed once, depending on the information about the injection molding machine, the information about the molding die, and the information about the resin material. Information about whether to do so or not is stored. Specifically, when the "sink mark" button is pressed, the molding condition confirmation/correction unit 81 increases the compression amount on the mold clamping device 13 side and increases the holding pressure during holding pressure on the injection device 14 side. The molding conditions are corrected for each of the mold clamping device 13 side and the injection device 14 side, such as increasing the amount (s8). Further, when the "burn" button is pressed, the molding conditions are corrected, such as by reducing the injection speed on the injection device 14 side. As described above, the molding conditions may be modified by modifying the molding conditions on at least one of the mold clamping device 13 side and the injection device 14 side.

Then, the operator molds the molded product P again manually or semi-automatically. At this time, the molding conditions estimated by the learning model are displayed once on the display device, and when the operator confirms the molding conditions and determines that the molding conditions are good, he or she can press the molding start button for manual molding or semi-automatic molding to start molding. desirable. Furthermore, if the operator determines that the molding conditions are not good after checking, the operator may manually correct the molding conditions. However, if the learning model of the machine learning device 70 repeatedly learns and becomes able to always generate good molding conditions (including the range of minor corrections), the molding conditions generated by the learning model will not be displayed on the display device and will remain as they are. The next injection molding may be started. This operation repeats steps (s7) and (s8) until the operator confirms that the molded product P is a good product.

If the molding conditions are improved by pressing the defective type button 124 and a non-defective product can be obtained (s6=Y), the operator presses the non-defective product button 122 (s9) and then presses the confirm button 123. As a result, the molding condition confirmation/correction unit 81 of the machine learning device 70 confirms that the molding conditions are correct, and the learning model storage unit 76 is stored at that time without modifying the learning model in the learning model generation/correction unit. The molding conditions are saved. Then, when a start button is pressed to start automatic molding (continuous molding), continuous molding is started (s10).

Furthermore, in the machine learning device 70, the learning model is modified in the learning model generation/modification unit 75, assuming that the countermeasure is modification of favorable molding conditions (s11). Note that the start of the continuous operation in (s10) and the correction of the learning model in (s11) may be performed at the same time, whichever comes first. In this embodiment, the learning model is modified by backpropagation (also called error backpropagation method or error backpropagation method). That is, the weight w and bias b are improved in order from the output layer 134 side of the neural network 131 shown in FIG. 9 to the intermediate layer 133 and input layer 132 sides, from the output side to the input side. That is, the output solution estimated by the learning model is verified, and the weight w and bias b of the neural network are adjusted to reduce the error from the correct answer. As an example, the value of the weight w of the neural network is changed by differentiating the error function using the gradient descent method. In other words, the error can be reduced by increasing or decreasing the weight w with respect to the slope of the error function.

Then, the learning model that has generated the molding conditions when the molded product P is a non-defective product is determined by the learning model generation/correction unit 75 of the learning model unit 73, assuming that the functional formula of the neural network 131 of the learning model is correct. The weight w of the expression is increased, etc. The corrected learning model neural network 131 is then stored in the learning model storage unit 76. In this way, the learning model of the molding conditions for the information regarding the molding die and the information regarding the resin material is stored, and
Even when the information is about the approximate molding die and the information about the resin material, the ratio at which good molding conditions are estimated is increased.

At the same time, the correction width (or correction ratio) of which molding condition is corrected and how much each time the defect type button 124 is pressed is also corrected by the learning model generation/correction section 75 of the learning model section 73. It is saved in the storage unit 76. Naturally, this correction width is not uniform depending on information regarding the injection molding machine, information regarding the molding die, and information regarding the resin material. Each time the defect type button 124 is pressed, the correction range (or correction ratio) for which molding condition is optimized, allowing the operator to appropriately correct the molding conditions without inputting numerical values from the input screen. . For example, if you press the defect type button 124 only once, you can make minor corrections, if you press it three times, you can make standard corrections, and if you press it five times, you can make major corrections. is adjusted by retraining the learning model. As described above, if the correction range is large, the operator should check the corrected molding condition values before starting the next molding. Note that the modification of the molding conditions can be performed even after starting fully automatic operation.

<How to modify a learning model in other embodiments>
Note that the learning model regarding injection compression molding may be modified by reinforcement learning. When performing reinforcement learning, if the molding condition corresponding to the defective type button 124 is corrected by pressing the defective type button 124 for a defective product and the molded product P becomes a good product in the next molding or later, it is determined as a good product button 122. Press the button 123 to give a positive reward for the modified molding conditions (or the numerical correction width or correction ratio of the corrected item), and give a negative reward for the molding conditions before the correction (or the corrected item). give. Further, if the molded product P molded from the next time onwards does not turn out to be a good product even if the defective type button 124 is pressed for a defective molded product P, the number of times the defective type button 124 is pressed is increased. If the molded product P still does not turn out to be a good product, a negative reward is given for the modified molding conditions (or the numerical correction width or correction ratio of the corrected item). If a good product cannot be molded using only the learning model of the machine learning device 70, the operator directly inputs the settings to correct the molding conditions so that a non-defective product can be molded. Even in the case where the operator directly inputs the settings, the range of settings to be input is smaller than when the operator inputs all the molding conditions from the beginning, and the occurrence of major basic mistakes is reduced. In many cases, it can contribute to the improvement of

<Core back molding>
Next, we will describe the generation and modification of molding conditions using a learning model for core-back molding (foam molding). In core-back molding, a mold clamping device 13 is used to inject a foaming material or a resin material containing foaming gas into a cavity formed between a fixed mold and a movable mold from an injection device, and then This is a molding method in which a molded product having air bubbles inside is formed by expanding the foaming material or foaming gas contained in the resin material within the cavity. In foam molding, it is important to appropriately retract the movable mold in response to the expansion of the molded product. Therefore, injection compression molding and core-back molding have something in common in that molding is performed while changing the volume of the cavity of the molding die 41.

If the retraction speed of the movable mold is too fast during core-back molding, there is a possibility that sink marks or skin layer defects may occur on the surface of the molded product, or the molded product may be thicker than expected. There is. Furthermore, if the retraction speed of the movable mold is too slow, there is a possibility that the molded product will be thinner than expected due to insufficient expansion of the foam. Therefore, the operator looks at the foam-molded product and, if sink marks or the like have occurred, presses the sink mark button for the defect type or the over-thickness button for core-back molding to correct the molding conditions. Specifically, if the sink button or the over-thickness button for core back molding is pressed, the retraction speed of the movable mold will be reduced by a predetermined speed in the molding conditions, or the retraction stop position of the movable mold will be reduced by a predetermined amount. The molding conditions are modified, such as being changed forward (in the direction in which the cavity C becomes narrower). Alternatively, numerical values of molding conditions such as the retraction speed of the movable mold, the pressure inside the cavity during retraction, and the retraction time may be directly reset from another input screen. If the thickness of the molded product is thin, correct the molding conditions by pressing the insufficient foam expansion button dedicated to core-back molding. Specifically, when the insufficient foam expansion button dedicated to core back molding is pressed, the retraction speed of the movable mold will be increased by a predetermined speed in the molding conditions, or the retraction stop position of the movable mold will be moved backward by a predetermined amount. The molding conditions are modified, such as in the direction in which the cavity C becomes wider. Alternatively, numerical values of molding conditions such as the retraction speed of the movable mold, the pressure inside the cavity during retraction, and the retraction time may be directly reset from another screen.

At this time, correct the molding conditions by pressing the sink button for the defect type, the over-thickness button for core back molding, and the insufficient foam expansion button, and if a good product is molded, the learning model is back-propagated. The weight w and bias b are improved in order from the output layer 134 side to the intermediate layer 133 and input layer 132 sides. Alternatively, a positive reward is given for the molding conditions that have been improved by reinforcement learning or the range of modification of the molding conditions. As a result, in the next case of similar information regarding an injection molding machine, a molding die, or information regarding a resin material, the probability that the molding condition generation unit will generate molding conditions that enable molding of a good product from the beginning increases. Regarding the generation and modification of molding conditions for core-back molding, the mutual causal relationship between the mold clamping device 13 side and the injection device 14 side is complex, similar to the learning model for injection compression molding, so it is more difficult to create and modify molding conditions than adjustment by an operator. It is useful to incorporate a learning model that can store and modify past molding conditions.
<When not using the defect type button>

Note that the method for correcting the learning model for injection compression molding or core-back molding by obtaining information about the molded product in injection compression molding or core-back molding described above is that the operator directly inserts the molded product into the molding die during manual molding or semi-automatic molding. If the molded product is defective, the defect type button 124 is pressed on the input screen 91 of the input device 64 of the injection molding machine 11 to input the type. This method is easy to use even for operators who have conventionally performed molding, and does not require investment in equipment such as separate inspection equipment. Further, it is also possible to modify the molding conditions by installing the machine learning device 70 in the control device 61 of the injection molding machine 11 without using the central control device 71.

However, in the present invention, it is not essential to modify the molding conditions from the defect type button 124 on the input screen 121, and the defect type button 124 is not provided on the input screen 121, no inspection device is used, and the operator who sees the defective product However, it is also possible to simply modify at least one element of the individual molding conditions manually from the input screen.

Further, information regarding molded products in injection compression molding and core-back molding may be detected by a detection device such as a CCD camera, a phototube, a gravimeter, or a colorimeter. When using a detection device, a machine learning device is first trained to learn that if a molded product is good, then it is a good product. Furthermore, if a molded product has a defect, the shape of the defect and the type of defect are linked and the machine learning device is used for supervised learning.

If the detection device inspects the molded product and determines that there is a specific type of defect, it changes the weighting by backpropagation or uses reinforcement learning, similar to when the operator presses the defect type button 124 described above. Modify the learning model by giving rewards by Then, the results of the generation of molding conditions performed by the learning model are changed using the initially set information on the injection molding machine 11, the information on the molding die 41, and the information on the resin material supplied to the injection device 14. so that Note that it is preferable to leave the part of the fine modification of the molding conditions to be made by the operator. Therefore, the case where the inspection device is used may be used for monitoring molded products after the start of fully automatic operation. When the defect rate of molded products slightly increases, countermeasures are displayed as a message on the display screen of the injection molding machine 11 or on the display screen of the central control device 71 that controls the plurality of injection molding machines 11. When the operator who has seen the message thinks that the countermeasure is correct, he or she finely modifies the numerical values from the input device 64 of the injection molding machine 11 or the input device controlling the plurality of injection molding machines 11 to modify the molding conditions.

Although the present invention will not be listed one by one, it is not limited to the embodiment described above, and may also include combinations of the descriptions of the embodiment or modifications made by those skilled in the art based on the gist of the present invention. It goes without saying that this applies. For example, the above embodiments can be applied to general injection molding other than injection compression molding and core-back molding. In particular, there is a method in which the operator who has judged the condition of the molded product selects and presses the defect type button on the input device to correct the molding conditions and eliminate the defects in the molded product molded using the corrected molding conditions. , it can be applied to any type of molding.

1 Injection molding system 11 Injection molding machine 13 Mold clamping device 14 Injection device 15 Fixed mold 17 Movable molds 61, 71 Control device 64 Input device 70 Machine learning device 73 Learning model section 74 Estimation section 75 Learning model generation/correction section 76 Learning model storage unit 77 Molding condition storage unit 81 Molding condition confirmation/correction unit 91, 121 Input screen 131 Neural network 132 Input layer 133 Intermediate layer 134 Output layer C Cavity P Molded product b Bias w Weight y Output value

Claims (6)

A method for generating molding conditions for an injection molding system equipped with an injection molding machine that moves a movable mold relative to a fixed mold during injection compression molding or core-back molding, the method comprising:
A molding mold in which a cavity is formed between a fixed mold and a movable mold;
An injection molding machine is provided, which has a mold clamping device to which the molding die is attached, and an injection device that performs injection into the cavity,
From at least information regarding the injection molding machine, information regarding the molding die, and information regarding the resin material supplied to the injection device,
A method for generating molding conditions for an injection molding system, wherein a learning model for injection compression molding or core-back molding generates molding conditions for injection compression molding or core-back molding. Perform injection compression molding or core-back molding using the learning model to form a molded product,
2. The method for generating molding conditions for an injection molding system according to claim 1, wherein a learning model regarding the injection compression molding or core-back molding is corrected by obtaining information regarding the molded product. In an injection molding system equipped with an injection molding machine that moves a movable mold relative to a fixed mold during injection compression molding or core-back molding,
A molding mold in which a cavity is formed between a fixed mold and a movable mold;
an injection molding machine having a mold clamping device to which the molding die is attached and an injection device for injecting into the cavity;
a storage device that stores molding conditions;
a learning device having a learning model that generates molding conditions for injection compression molding or core-back molding from at least information regarding the injection molding machine, information regarding the molding die, and information regarding the resin material supplied to the injection device; injection molding system. a detection device that detects the state of a molded product molded by injection compression molding or core-back molding using the learning model regarding injection compression molding or core-back molding;
The injection molding system according to claim 3, further comprising a learning device that corrects a learning model related to the injection compression molding or core-back molding based on the information sent from the detection device. an input device through which an operator inputs the state of a molded product formed by performing injection compression molding or core-back molding using the learning model regarding injection compression molding or core-back molding;
The injection molding system according to claim 3, further comprising a learning device that corrects a learning model related to the injection compression molding or core-back molding based on information sent from the input device. A molding condition generation support program for an injection molding system installed in the injection molding system according to at least one of claims 3 to 5.

Images