Reliability evaluation method for squeeze casting process parameter data

To evaluate the reliability of squeeze casting (SC) process parameter data originating from different sources such that guarantee the performance of data-driven manufacturing, this paper proposes a relatively reliability evaluation method based on the status of relevant data. The relative reliability of the data is defined. The value range of each attribute of the initial data set is achieved and divided into different (reliable) intervals or classes by Canopy K-means clustering and a local linear embedding algorithm according to the data values and engineering fact, and the data status of each data cell is obtained by calculating the classes to which its value belongs. The reliability of a data unit is obtained by calculating the probability of the data cell status. Furthermore, a reliability evaluation model of all the data is proposed. It integrates association rules and a Bayesian network to formulate the relationship between data units and obtain the reliability of each piece of data through reasoning. A process parameter sample datasets of SC collected from 107 studies was evaluated to prove the effectiveness of the proposed method, and its performance was assessed by comparing it with the simulation results obtained using ProCAST and JMatPro. The results show that castings manufactured with more reliable data, as evaluated by the method, exhibit lower shrinkage porosity and better mechanical, thermal, and physical properties, which proves its effectiveness in detecting the reliability of different data for specific applications.