Introduction
In the 21st century, with the continuous development of computer technology, digital, networked, and intelligent technologies have become the main force in the development of the manufacturing industry and the main driving force of the new round of industrial revolution, providing a new technical path for quality control in the manufacturing industry. For example, many enterprises have introduced systems such as Enterprise Resource Planning (ERP) and Manufacturing Execution System (MES) to implement information management in the enterprise.
Sand casting is an important branch of casting and plays a significant role in the production of engine cylinder heads in the automotive industry. However, due to the complex production process, long manufacturing process, and large production volume of sand casting, its products have problems such as significant quality fluctuations and difficulty in quality traceability. In recent years, under the new situation of industrial upgrading, environmental protection, and customer demand, sand casting enterprises are also transforming to automated and information-based production. A lot of data collection equipment has been used in the front-line production and connected to information management systems such as ERP and MES. These information systems record a large amount of real production data on equipment, processes, environment, and quality in the casting production, which is a wealth containing the relationship between production parameters and casting quality. However, the existing “wealth” has not been transformed into information and knowledge, or only a small amount of data has been utilized, resulting in a “data explosion and knowledge scarcity” situation within the enterprise. If we can conduct in-depth mining and analysis of all the collected data on the basis of information management and find out the coupling relationship between various links and parameters, it will certainly help casting enterprises better control the quality and trace the defects. Therefore, it is of great significance to mine the data of sand casting enterprises.
Data Mining in Sand Casting
Data mining is the process of revealing meaningful new relationships, trends, and patterns by analyzing a large amount of existing data. It is the process and technology of discovering potential and valuable information from a random, massive, noisy, incomplete, and fuzzy large database, and is also a decision support process. In the field of artificial intelligence, it is also known as knowledge discovery in the database, and some people regard data mining as a basic step in the process of knowledge discovery in the database. As an emerging technology, the life cycle of data mining is in a stage of difficulties and requires time and effort to research and develop. It has tended to mature and will eventually be accepted by people.
Currently, many fields have carried out analysis work through data mining technology, such as financial management, insurance, management system research, etc., while the data mining research in the casting industry is still in its infancy.
It should be noted that data mining requires a large amount of data, and it is necessary to focus on accumulating data. Only when the amount of data is large can the mining effect tend to be stable and closer to the real situation. ERP and MES systems can provide data support for data mining. Therefore, introducing the ERP system and conducting data mining based on the ERP system is an important development direction for quality control and improving product quality in sand casting enterprises.
Huazhu ERP System
The Huazhu ERP system is a casting ERP system developed by the Huazhu Software Center of Huazhong University of Science and Technology. With the purpose of standardizing the management of casting enterprises, improving enterprise efficiency, reducing enterprise costs, accelerating the process of enterprise informatization, and enhancing the enterprise’s market response ability, it can realize the all-round information integration of casting enterprises. The business process of the system is customer-centric, driven by tasks, and pulls production with orders, which can realize the all-round management of procurement, production, sales, inventory, and other links, making enterprise management more relaxed and efficient.
For example, Company Y is a casting subsidiary of a state-owned enterprise, mainly producing diesel engine parts such as engine cylinder blocks, cylinder heads, crankshafts, and flywheels. It has a sand casting production line with a small variety and large batch. Due to the high quality control requirements, the company previously introduced the Huazhu ERP system to implement piecewise management of castings. Through on-site investigation and analysis, the business process of the company applying the Huazhu ERP is as follows: order entry -> processing route setting, BOM formulation, casting process allocation -> production preparation -> planned production -> production processing -> sales and shipment; among them, the production preparation stage includes mold management, raw material management, and procurement management.
Data Mining Process
The data mining work can be divided into three parts: data collection, data mining model, and mining result display.
Data Collection
All data in each link will be entered into the system through automatic equipment collection and manual input, providing a data basis for data mining. For example, data such as sand mixing records, core making records, core assembly records, raw material records, and quality inspection records are collected through relevant modules.
Data Mining Model
The data mining methods include neural network methods, decision tree algorithms, association analysis, rough set methods, fuzzy set methods, statistical analysis methods, covering positive examples and excluding negative examples methods, visualization techniques, etc. The core technology of the Huazhu ERP system is database technology, and the database used is SQL Server. SQL Server is a relational database that connects different two-dimensional tables through a relational model, which can better adapt to the association analysis method. Therefore, the proposed mining model is mainly based on association analysis and neural network methods.
The model takes the casting number (piece number) as the input. At the first level, it can be associated with orders, quality records, and production records. Orders can be further associated with customers and shipment management to conduct mining and analysis of customer needs. Production records can be associated with process cards. On the one hand, quality records can be associated with the processing parameters of all processes through dates, and then conduct mining and analysis of the process together with the process cards. On the other hand, it can further analyze the causes of defects, the bottleneck processes of products, and other data.
According to the model, the key forms in the database include the order detail form, planned production form, production quality inspection form, production parameter record form, and shipping list, and their relationship is shown in the figure.
Data Display
The final analysis results of data mining are displayed in the form of data tables and statistical charts. For example, the quality statistics data table shows the daily production quality of castings, including the production internal scrap rate and other information. The scatter plot of daily casting output shows the distribution of casting output. The column chart of defect distribution shows the distribution of internal and external scrap defects of castings. The pie chart of monthly casting output shows the proportion of different series of castings in the monthly output.
At the same time, through the data mining model, the process parameters of each product are also correlated, and can form a mapping relationship with the final quality result. Therefore, the relevant process parameters of each product can be used as the input, and the final quality result can be used as the output to train the neural network model, so as to predict the quality of products using different process parameters.
Application of Data Mining in Sand Casting
By using the data mining model based on the Huazhu ERP system, sand casting enterprises can obtain deeper information about customers, defect distribution status, production capacity, and other aspects. For example, through the analysis of data, enterprises can understand the demand trends of customers, optimize the production process to reduce the occurrence of defects, and reasonably arrange the production plan to improve production efficiency.
In addition, the data mining model can also realize the traceability of casting product quality problems. By analyzing the data of each link in the production process, enterprises can quickly find the root cause of the quality problem and take corresponding measures to solve it. In the future, neural networks can also be used to predict the process quality of different parameters, so as to give full play to the role of data and improve the enterprise’s control level of product quality.
Conclusion
In conclusion, the data mining model based on the Huazhu ERP system has important application value for sand casting enterprises. It can help enterprises better manage and control the production process, improve product quality, and enhance the competitiveness of the enterprise in the market. However, in the application process, enterprises need to pay attention to the accumulation and quality of data, and continuously optimize and improve the data mining model to adapt to the changing market demand and production environment.
Table 1: Comparison of Data Mining Methods
| Method | Description | Advantages | Disadvantages |
|---|---|---|---|
| Neural Network Methods | Use artificial neural networks to learn and predict patterns in data. | Can handle complex nonlinear relationships, has strong generalization ability. | Requires a large amount of data for training, the interpretation of the model is difficult. |
| Decision Tree Algorithms | Build a decision tree based on data to make predictions or classifications. | Easy to understand and interpret, can handle both categorical and numerical data. | May overfit the data, sensitive to noise in the data. |
| Association Analysis | Discover associations and correlations between different variables in the data. | Can find hidden relationships in the data, useful for market basket analysis. | May generate a large number of rules, some of which may be meaningless. |
| Rough Set Methods | Handle imprecise and incomplete data by using the concept of rough sets. | Can reduce the complexity of data, does not require prior knowledge about the data. | The calculation is relatively complex, and the performance may be affected by the size of the data. |
| Fuzzy Set Methods | Deal with uncertainty and fuzziness in the data by using fuzzy sets. | Can better handle subjective and ambiguous information, suitable for decision-making under uncertainty. | The definition of fuzzy sets may be subjective, and the results may vary depending on the chosen membership function. |
| Statistical Analysis Methods | Use statistical techniques to analyze and summarize data. | Provides a rigorous mathematical basis, can test hypotheses and estimate parameters. | Requires certain statistical knowledge and assumptions, may not be suitable for complex data structures. |
| Covering Positive Examples and Excluding Negative Examples Methods | Identify patterns that cover positive examples and exclude negative examples in the data. | Can find specific patterns in the data, useful for anomaly detection. | May miss some hidden patterns, and the performance depends on the selection of positive and negative examples. |
| Visualization Techniques | Use visual representations such as charts and graphs to display data and patterns. | Helps users understand and interpret data more intuitively, can discover patterns and trends at a glance. | Limited by the display space and resolution, may not be able to show all the details of the data. |
Table 2: Key Data Forms in the Huazhu ERP System
| Form Name | Description | Key Fields | Relationship with Other Forms |
|---|---|---|---|
| Order Detail Form | Contains details of customer orders, such as order number, customer information, casting specifications, and quantity. | Order number, customer, casting number, quantity, material, etc. | Associated with the shipping list through the order number. |
| Planned Production Form | Records the planned production information, including production batch number, casting number, process card number, and production schedule. | Production batch number, casting number, process card number, planned production date, etc. | Associated with the process card through the process card number. |
| Production Quality Inspection Form | Contains the quality inspection results of castings, such as inspection date, casting number, defect types, and inspection personnel. | Inspection date, casting number, defect types, defect description, inspection personnel, etc. | Associated with the casting number to trace the quality of specific castings. |
| Production Parameter Record Form | Records the production parameters of each process, such as processing date, casting number, equipment parameters, and process parameters. | Processing date, casting number, equipment number, parameter 1, parameter 2, etc. | Associated with the casting number to analyze the impact of process parameters on the quality of castings. |
| Shipping List | Records the shipping information of castings, such as shipping date, order number, casting number, and shipping quantity. | Shipping date, order number, casting number, shipping quantity, etc. | Associated with the order detail form through the order number. |
Table 3: Analysis of Customer Needs Based on Data Mining
| Customer | Demand Trends | Suggested Actions |
|---|---|---|
| Customer A | Increases the demand for high-precision castings. | Invest in advanced production equipment and technology to improve the precision of castings. |
| Customer B | Requires shorter delivery times. | Optimize the production process and supply chain management to reduce the production cycle. |
| Customer C | Focuses on the quality and reliability of castings. | Strengthen the quality control system and conduct regular quality inspections. |
| Customer D | Has a demand for customized castings. | Establish a flexible production system to meet the individual needs of customers. |
Table 4: Defect Distribution Analysis
| Defect Type | Occurrence Frequency | Affected Castings | Possible Causes |
|---|---|---|---|
| Porosity | High | Cylinder Heads, Crankshafts | Improper pouring temperature, inadequate venting. |
| Crack | Medium | Flywheels, Engine Blocks | High internal stress, material defects. |
| Inclusion | Low | Cylinder Heads, Engine Blocks | Contamination during the casting process. |
Table 5: Process Optimization Based on Data Mining Results
| Process | Current Parameters | Recommended Parameters | Expected Benefits |
|---|---|---|---|
| Melting | Temperature: 1500°C, Time: 2 hours | Temperature: 1550°C, Time: 1.5 hours | Improve the metallurgical quality of the molten metal, reduce defects. |
| Molding | Pressure: 100 MPa, Speed: 50 cm/s | Pressure: 120 MPa, Speed: 60 cm/s | Enhance the compactness of the mold, improve the surface quality of castings. |
| Heat Treatment | Temperature: 550°C, Time: 3 hours | Temperature: 600°C, Time: 2.5 hours | Optimize the mechanical properties of castings, improve the strength and toughness. |
Table 6: Production Capacity Analysis
| Production Line | Monthly Output | Capacity Utilization Rate | Bottleneck Processes |
|---|---|---|---|
| Line 1 | 1000 pieces | 80% | Molding, Heat Treatment |
| Line 2 | 800 pieces | 70% | Melting, Machining |
| Line 3 | 1200 pieces | 90% | Quality Inspection, Packaging |
