The artificial neural network (ANN) model and its user interface were developed using a popular visual programming tool, Visual Basic (VB), running on the Windows operating system. The ANN module was designed based on the fundamental principles of neural networks, incorporating the characteristics of spring permanent deformation and data from various influencing factors. This system has been fully implemented and is now in operation.
After integrating the self-learning and effective neural network module with the database samples, the prediction model performs self-learning on the sample data, enabling it to identify the relationship between the spring’s permanent deformation and each influencing factor. The model has demonstrated a high level of predictive accuracy. In future applications, the system's forecasting capabilities and scope will be further enhanced.
During the trial operation, the system's interface was aesthetically refined, auxiliary functions such as help file creation were improved, and installation files were generated, allowing the system to be installed and run in real production environments.
To use the software, users can click on "Spring Residual Deformation Prediction" under the "Prediction" menu or the corresponding button on the toolbar. A prediction interface will then appear. Users are required to input the design height of the spring, as specified in the drawing or according to actual conditions, along with the free height, load, spring diameter, wire diameter, and the number of effective coils. Additional inputs such as maximum load and height are optional but can serve as references. If no data is entered, the program will prompt the user to confirm the input. Once confirmed, clicking the "Predict" button will display the predicted pre-test roll height and the residual deformation (permanent deformation) after the strong pressure test in the result box.
The application has shown that the artificial neural network is highly feasible for predicting the permanent deformation of helical compression springs, achieving satisfactory results (see Table 1). Additionally, the system can manage spring test data, making it highly practical. However, due to the limited number and distribution of existing data samples, and the fact that the ANN has strong interpolation but weak extrapolation capabilities, some simulation errors may occur. These issues need to be addressed in the future through data expansion and continuous improvement.
The software has already been developed and put into use. The technical department has utilized it to predict failures in various compression springs. By inputting key parameters such as the design spring height, elastic modulus, spring diameter, wire diameter, and the number of effective coils before new spring production, the model calculates the expected spring height and permanent deformation, achieving accurate predictions. This helps guide the production process before the strong pressure test, improving efficiency and quality control.
Cleaning Ball,Rotary Cleaning Ball,Sanitary Cleaning Ball,Tri Clamp Cleaning Ball
Ningbo Wenhan Fluid Equipments Co., Ltd. , https://www.wenhanvalves.com