Data processing: generic control systems or specific application – Generic control system – apparatus or process – Optimization or adaptive control
Reexamination Certificate
2000-12-29
2003-07-29
Follansbee, John (Department: 2121)
Data processing: generic control systems or specific application
Generic control system, apparatus or process
Optimization or adaptive control
C700S049000, C700S108000, C700S109000, C700S200000, C700S203000, C706S021000, C706S023000, C706S904000
Reexamination Certificate
active
06600961
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates generally to an injection machine, particularly to an intelligent control method for injection machine by applying a neural network concept to control and predict quality and yield of injection products.
2. Description of the Prior Art
The quality control of an injection product usually includes two stages, namely: parameter's setting in the first stage and quality's monitoring and controlling in the second.
In the first stage, after input of predetermined injection forming parameters, such as fusion point, mold's temperature, injection pressure and speed, etc., the quality of an injection product is seemingly assurable theoretically, however, it is not the case as easy as expected for selection and input of the right parameters because of the dynamic complicatedness and the condition change going along with time of the injection process. Besides, the fused plastic material is a non-Newtonian and non-isothermal fluid suffering high pressure and high temperature in the injection process, hence, the forming parameters will normally need some correction.
For setting correct parameters, the low-efficiency trial-and-error method was generally taken to amend the parameters repeatedly until an acceptable trial run was obtained. When doing so, the engineer would rather treat it a linear relationship between the forming parameters and the quality variable parameters for amending the parameters one at a time by ignoring their real interactions. Moreover, the parameters set by the trail-and-error method may locate at the verge adjacent to a parameter window that could degrade the product quality.
Later on, an Expert System has been introduced for setting the forming parameters of the injection machine based on an “if-then” postulation, however, the Expert System is still failed in building up a more distinct qualitative relationship between the forming parameters and the quality variable parameters.
A more systematic and effective method for setting the forming parameters is a so-called Taguchi parameter design method (TPDM) capable of obtaining more information for building up an Empirical Model. Nevertheless, the TPDM is found imperfect in the fact that at least 25 forming parameters, which can affect the product quality, will affect with each other too. Thus, considerable time and manpower are required to proceed related experiments and analysis for grasping some useful factors between specified forming parameters and quality variable parameters for application only in some specified injection machines.
Because of the non-linearity of the relationship between the forming parameters and the quality variable parameters and the dynamic complicatedness and the condition change going along with time of the injection process as mentioned above, on-site quality check and control of the second stage become more important and indispensable.
For achieving this purpose, a production engineer in early days would inspect the product quality by experience, eye-measurement, and irregular check for adjusting the parameters of the injection machine and for control of the product quality. This quality control (Q.C.) method is found defective in:
training course being required to enable an operator to judge the product quality accurately; and
the eye-measurement being workable for checking appearance or severe defects of the product only.
Therefore, some inconspicuous defects, such as slight cambering, deformation, chasms, remained stress, are prone to be neglected until the injected products are put under assembling.
A control method of statistical process model (SPM) is a method used for measuring process signal, evaluating product quality, and thereby building up a SPM, for example, inferring the product's measurements and weight basing on the nuzzle pressure. Nevertheless, defect of the SPM may include: considerable time, cost, and manpower being required; and no distinct qualitative relation between the forming parameters and quality variable parameters being deducible.
In the latest decade, more than one thousand inventions regarding the injection machine have been disclosed per year to prove that the injection forming technology is still under endless growing.
According to ep.espacenet.com, 173 patents relating to the process control of injection formation have been granted until now since 1970s, wherein some patents concerning this invention more or less may include:
U.S. Pat. No. 5,997,778, which is applied to obtain the dynamic response of an injection machine based on a given injection-speed curve and use the proportional-integral-differential (PID) feedback control technique to set the forming parameters for continuous control of the injection process;
U.S. Pat. Nos. 5,246,645, 5,514,311, and 4,820,464 providing switching methods of feeding control for keeping pressure;
U.S. Pat. Nos. 5,997,778, 4,850,217, and 4,816,197 for control forming quality by adjusting the interrelation among pressure, volume, temperature (PVT);
U.S. Pat. No. 4,311,446 used for forming control by comparing the auger position, speed, and temperature of fused gel with the given or expected data;
U.S. Pat. Nos. 4,060,362, 3,893,792, and W09943487 for control of forming parameters by feeding back the pressure in the feed pipe or the mold-cavity; and
Japan patents No. 61118223 and 63125322 for control of forming process by regulating the auger speed.
Compared with abovesaid patens, this invention is an intelligent control method capable in both ways: setting proper forming data and inspecting control of product quality.
SUMMARY OF THE INVENTION
The primary object of this invention is to transplant the intelligent control and prediction techniques of a neural network to an injection machine, which has been exemplified capable of deciding the quasi best machine parameters rapidly in couple processing cycles for increasing yield with least loss, and for detecting and adjusting conditions until a desired operation environment is obtained.
In order to realize the object, an intelligent control method for injection machine mainly comprises the following steps:
executing a model-flow software for process simulation and comparing data with trial operations for building a qualitative-relation database comprising at least machine parameters, process parameters, and quality variable parameters;
defining a total loss function formula and according to the qualitative-relation database to calculate and obtain a data group of “Total loss vs machine parameters” and another data group of “Yield parameters vs quality variable parameters”;
building an intelligent control neural network and a quality prediction neural network on the basis of the qualitative-relation database and those two data groups;
connecting the intelligent control neural network, the total loss function formula, and the quality prediction neural network together in series, wherein an output terminal of the intelligent control neural network is jointed with an input terminal of the injection machine; an input terminal of the quality prediction neural network is jointed with an output terminal of the injection machine; the intelligent control neural network is to receive the output of the total loss function formula and forward amended parameters to the injection machine; and the quality prediction neural network is to receive the output process parameters from the injection machine and forward the quality variable parameters to the total loss function formula.
For more detailed information regarding this invention together with further advantages or features thereof, at least an example of preferred embodiment will be elucidated below with reference to the annexed drawings.
REFERENCES:
patent: 3893792 (1975-07-01), Laczko
patent: 4060362 (1977-11-01), Wilson, III
patent: 4311446 (1982-01-01), Hold et al.
patent: 4816197 (1989-03-01), Nunn
patent: 4820464 (1989-04-01), Sasaki et al.
patent: 4850217 (1989-07-01), Nunn
patent: 5246645 (1993-09-01), Tagawa et al.
patent: 53114
Liang Jui-Ming
Wang Pei-Jen
Barnes Crystal J.
Follansbee John
Mirle Automation Corporation
LandOfFree
Intelligent control method for injection machine does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Intelligent control method for injection machine, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Intelligent control method for injection machine will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3097156