Quality prediction device
The quality prediction device addresses the challenge of inaccurate quality prediction in die-casting by using a machine learning-based calculation model to estimate casting defects, ensuring reliable quality assessment despite injection speed fluctuations.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- TOYOTA JIDOSHA KK
- Filing Date
- 2024-11-20
- Publication Date
- 2026-06-01
AI Technical Summary
Existing quality prediction methods for die-casting products fail to accurately predict the quality when the die-casting device switches injection speeds, particularly due to variations in injection speed caused by solidification precipitates.
A quality prediction device that incorporates a calculation model using machine learning to predict casting defects based on injection speed fluctuations, mold temperature, and filling time, including a storage device to store the model and an execution device to execute the prediction program, which outputs the total volume of casting defects.
Enables accurate prediction of casting quality by accounting for variations in injection speed, thereby improving the reliability of quality assessment in die-casting processes.
Smart Images
Figure 2026089554000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a quality prediction device.
Background Art
[0002] Patent Document 1 describes a quality prediction method for predicting the quality of a casting product cast by a die-casting device. The die-casting device has a mold that partitions a cavity and a plunger that injects molten metal into the cavity. In the quality prediction method, based on the injection speed of the plunger, the temperature of the mold, and the filling time from the start of the injection of the plunger until the molten metal fills the cavity, a computer predicts the quality of the casting product cast by the die-casting device.
Prior Art Documents
Patent Documents
[0003]
Patent Document 1
Summary of the Invention
Problems to be Solved by the Invention
[0004] When injecting the plunger, a die-casting device as described in Patent Document 1 may switch the target injection speed from the first speed to the second speed. In this case, with the quality prediction method as described in Patent Document 1, there is a possibility that a computer may not be able to appropriately predict the quality of the casting product.
Means for Solving the Problems
[0005] A quality prediction device for solving the above problems predicts the quality of a cast product cast by a die-casting apparatus having a mold that partitions a cavity and a plunger that injects molten metal into the cavity by switching the target value of the injection speed from a first speed to a second speed faster than the first speed, and comprises a storage device and an execution device, wherein the storage device stores a calculation model, and when a plurality of input values are input, including an injection speed fluctuation amount which is the maximum fluctuation amount after the target value is switched to the second speed, a mold temperature which is the temperature of the mold, and a filling time from when the injection of the plunger is started until the molten metal is filled into the cavity, the calculation model outputs the total volume of casting defects that are predicted to occur in the cast product as an output value, and the execution device predicts the quality of the cast product based on the output value output from the calculation model. [Effects of the Invention]
[0006] When filling the cavity with molten metal, the injection speed fluctuation changes depending on the occurrence of solidification precipitates. Therefore, the injection speed fluctuation affects the quality of the cast product. According to the above configuration, the input values to the calculation model, which is determined by machine learning or the like, include the injection velocity fluctuation. Therefore, the output value from the calculation model is calculated based on the injection velocity fluctuation.
[0007] The quality prediction device predicts the quality of a cast product based on the total volume of casting defects, which is the output value of a calculation model that can change according to the variation in injection speed. Therefore, when the variation in injection speed changes due to factors such as the occurrence of solidification precipitates, the quality prediction device can predict the quality of the cast product in accordance with the variation in injection speed. [Brief explanation of the drawing]
[0008] [Figure 1] Figure 1 is a schematic diagram showing a die-casting system. [Figure 2] Figure 2 is a timing chart showing the time-series data of the target injection velocity and the time-series data of the detected injection velocity. [Figure 3]Figure 3 is a flowchart showing the series of processes performed by the quality prediction device to predict the quality of cast products. [Modes for carrying out the invention]
[0009] <Overview of the die-casting system> As shown in Figure 1, the die-casting system 10 includes a die-casting device 20 and a quality prediction device 80.
[0010] The die-casting apparatus 20 casts aluminum alloy products by die-casting. The die-casting apparatus 20 includes a mold 30, a sleeve 40, a plunger 50, and a control device 60. The mold 30 comprises a fixed mold 31 and a movable mold 32. The movable mold 32 is clamped opposite the fixed mold 31 by a movable mechanism (not shown).
[0011] With the movable mold 32 clamped to the fixed mold 31, the mold 30 defines the cavity C. The cavity C has the shape of the cast product to be cast by the die-casting device 20.
[0012] The sleeve 40 is connected to the mold 30. The sleeve 40 is cylindrical. Therefore, the sleeve 40 has an internal space. The internal space of the sleeve 40 is connected to the cavity C. The plunger 50 is located in the internal space of the sleeve 40.
[0013] The plunger 50 pushes the molten metal from the internal space of the sleeve 40 into the cavity C. The plunger 50 has a plunger body 51 and a plunger tip 52. The plunger body 51 is rod-shaped. The plunger tip 52 is connected to the tip of the plunger body 51.
[0014] The plunger tip 52 is slidable within the internal space of the sleeve 40 as the plunger body 51 moves in the direction in which the sleeve 40 extends. Therefore, when the plunger body 51 is moved in the direction in which the sleeve 40 extends by an injection actuator (not shown), the plunger tip 52 pushes the molten metal present in the internal space of the sleeve 40 into the cavity C.
[0015] The control device 60 controls the injection speed of the plunger 50 using an injection actuator. The control device 60 controls the injection actuator so that the injection speed of the plunger 50 reaches the target value TP while the movable mold 32 is tightened to the fixed mold 31 and molten metal has been poured into the internal space of the sleeve 40.
[0016] As shown in Figure 2, at time t1, after the molten metal has been poured into the internal space of the sleeve 40, the control device 60 first sets the target value TP of the injection speed to the first speed SP1. Then, the control device 60 drives the injection actuator to move the plunger body 51 so that the injection speed becomes the first speed SP1.
[0017] Subsequently, at time t2, which is after time t1, the control device 60 sets the target value TP of the injection speed to a second speed SP2, which is faster than the first speed SP1. Then, the control device 60 drives the injection actuator to move the plunger body 51 so that the injection speed becomes the second speed SP2.
[0018] Subsequently, at time t3, which is after time t2, the control device 60 stops the drive of the injection actuator. In this way, the control device 60 controls the injection speed of the plunger 50. As the plunger body 51 moves, the molten metal extruded by the plunger tip 52 is filled into the cavity C. The control device 60 also transmits the target value TP of the injection speed to the quality prediction device 80.
[0019] As shown in FIG. 1, the die-casting system 10 includes a mold temperature sensor 71 and an injection speed sensor 72. The mold temperature sensor 71 detects the mold temperature TE, which is the temperature of the mold 30. The mold temperature sensor 71 transmits the detected mold temperature TE to the quality prediction device 80. The mold temperature TE is, for example, the temperature of the moving mold 32 at time t1. The injection speed sensor 72 detects the detected value GP of the injection speed. The injection speed sensor 72 transmits the detected value GP of the detected injection speed to the quality prediction device 80.
[0020] The quality prediction device 80 acquires the mold temperature TE from the mold temperature sensor 71. The quality prediction device 80 acquires the time-series data of the target value TP of the injection speed from the control device 60. The quality prediction device 80 acquires the time-series data of the detected value GP of the injection speed from the injection speed sensor 72.
[0021] As shown in FIG. 2, the quality prediction device 80 calculates the filling time FT and the injection speed variation amount CA based on the time-series data of the detected value GP of the injection speed and the time-series data of the target value TP of the injection speed. Thereby, the quality prediction device 80 acquires the filling time FT and the injection speed variation amount CA.
[0022] The filling time FT is the time from when the injection of the plunger 50 starts until the molten metal fills the cavity C. Specifically, the filling time FT is the time from the time t when the quality prediction device 80 starts driving the actuator to move the plunger body 51 to the time t when the driving of the actuator is stopped. That is, the filling time FT is the time from time t1 to time t3.
[0023] When calculating the filling time FT, the quality prediction device 80 first refers to the time-series data of the target value TP of the injection speed to specify time t1 and time t3. Time t1 is the time t when the target value TP switches from zero to the first speed SP1. Time t3 is the time t when the target value TP switches from the second speed SP2 to zero. Next, the quality prediction device 80 calculates the time from time t1 to time t3 as the filling time FT.
[0024] The injection velocity fluctuation CA is the maximum fluctuation amount during the target period PE, from time t when the target value TP of the injection velocity of the plunger 50 is switched to the second velocity SP2 and exceeds the second velocity SP2, until time t when the detected value GP's fluctuation from the second velocity SP2 becomes less than or equal to a specified amount. In this embodiment, the specified amount is approximately zero.
[0025] When the quality prediction device 80 calculates the injection velocity fluctuation CA, it first determines the target period PE. The quality prediction device 80 identifies the start time ts of the target period PE. The start time ts is after time t2 and is the time t at which the detected value GP first exceeds the second velocity SP2. Next, the quality prediction device 80 identifies the end time te of the target period PE. The end time te is after the start time ts and is the time t at which the detected value GP has reached the second velocity SP2 for a specified number of consecutive times. Next, the quality prediction device 80 calculates the injection velocity fluctuation CA as the difference between the maximum value and the minimum value of the detected value GP during the target period PE.
[0026] As shown in Figure 1, the quality prediction device 80 comprises an execution device 81 and a storage device 82. The execution device 81 is a CPU. The storage device 82 is memory. The storage device 82 stores a prediction program PR for predicting the quality of the cast products to be cast by the die-casting device 20. By executing the prediction program PR on the execution device 81, the quality prediction device 80 predicts the quality of the cast products to be cast by the die-casting device 20.
[0027] The memory device 82 stores the calculation model MD used in conjunction with the execution of the prediction program PR. When the calculation model MD receives multiple input values, including the injection speed fluctuation amount CA, the mold temperature TE, and the filling time FT, it outputs the total volume TV of casting defects predicted to occur in the cast product cast by the die-casting device 20 as an output value. The calculation model MD is determined in advance through tests and simulations. The calculation model MD is, for example, a machine learning model determined by machine learning using training data. The training data used in machine learning is the total volume TV of casting defects in the cast product, obtained using a measuring instrument or the like.
[0028] When the casting of the cast product by the die-casting apparatus 20 is completed, the execution device 81 starts executing the prediction program PR. In other words, the execution device 81 starts executing the prediction program PR each time a cast is made.
[0029] As shown in Figure 3, when the execution device 81 starts executing the prediction program PR, the execution device 81 first performs the process in step S11. In step S11, the execution device 81 obtains the mold temperature TE, the filling time FT, and the injection speed fluctuation amount CA. After that, the execution device 81 proceeds to step S12.
[0030] In step S12, the execution device 81 inputs several input values, including the mold temperature TE, the filling time FT, and the injection speed fluctuation amount CA, into the calculation model MD. After that, the execution device 81 proceeds to step S13.
[0031] In step S13, the execution device 81 obtains the total volume TV of casting defects predicted to occur in the cast product, which is an output value output from the calculated model MD. After that, the execution device 81 proceeds to step S14.
[0032] In step S14, the execution device 81 predicts the quality of the casting based on the total volume TV of the casting defects. For example, the execution device 81 calculates a quality parameter that increases as the total volume TV of the casting defects increases. A smaller value for the quality parameter indicates better quality. After that, the execution device 81 completes this series of processes.
[0033] <Effects of this embodiment> The quality prediction device 80 predicts the quality of the cast product to be cast by the die-casting device 20. The quality prediction device 80 comprises a storage device 82 and an execution device 81. The storage device 82 stores a calculation model MD. When the calculation model MD receives multiple input values, including injection speed fluctuation amount CA, mold temperature TE, and filling time FT, it outputs the total volume TV of casting defects predicted to occur in the cast product as an output value. The execution device 81 predicts the quality of the cast product based on the output value output from the calculation model MD.
[0034] As molten metal fills cavity C, the injection velocity fluctuation CA changes depending on the occurrence of solidification precipitates. Solidification precipitates are more likely to occur the more impurities are present in the material that will become the aluminum alloy. Therefore, the injection velocity fluctuation CA affects the quality of the cast product.
[0035] Multiple input values to the calculation model MD include the injection velocity fluctuation amount CA. Therefore, the output value from the calculation model MD changes according to the injection velocity fluctuation amount CA. Thus, the execution device 81 can predict the quality of the cast product according to the injection velocity fluctuation amount CA.
[0036] <Example of changes> The above embodiment can be implemented with the following modifications. The above embodiment and the following modifications can be combined with each other to the extent that they do not contradict each other technically.
[0037] The calculation model (MD) is not limited to machine learning models using training data; it may also be a calculation formula determined in advance through testing or simulation. The quality of the cast product predicted by the execution device 81 may be a parameter indicating whether it is a pass or fail. For example, the execution device 81 may predict that the product is a fail if the total volume TV of casting defects output from the calculated model MD in step S13 exceeds a predetermined threshold, while predicting that it is a pass if the total volume TV of casting defects is below that threshold. [Explanation of Symbols]
[0038] 10…Die-casting system 20…Die casting equipment 30…Mold 40... Sleeves 50... Plunger 80… Quality prediction device 81…Execution device 82…Storage device C... Cavity CA...Injection velocity variation FT... Filling time MD…Calculation Model SP1…1st speed SP2…2nd speed TE…Mold temperature TP…Target value TV...total volume
Claims
[Claim 1] A quality prediction device for predicting the quality of a cast product cast by a die-casting apparatus having a mold that partitions a cavity, and a plunger that injects molten metal into the cavity by switching the target value of the injection speed from a first speed to a second speed that is faster than the first speed, It comprises a memory device and an execution device, The aforementioned storage device stores the calculation model, The calculation model, upon receiving a plurality of input values including the injection velocity fluctuation, which is the maximum fluctuation after the target value is switched to the second velocity; the mold temperature, which is the temperature of the mold; and the filling time, which is the time from when the plunger injection starts until the molten metal fills the cavity, outputs the total volume of casting defects predicted to occur in the cast product as an output value. The execution device predicts the quality of the cast product based on the output value output from the calculation model. Quality prediction device.