Defect prediction for assembled products: a novel model based on the structural complexity paradigm

Abstract

Increased assembly complexity is one of the main challenges in manufacturing as it can induce an increase in time, cost, and defects. Several approaches have been proposed in the literature to predict product defects using assembly complexity as a predictor. However, most of these are not directly applicable because they rely on experts’ prior subjective knowledge and are designed for specific industrial applications. To overcome this issue, the present research proposes a novel approach to predict product defects from a purely objective assessment of product complexity, without the need for expert evaluations and assembly experience. A recent conceptual paradigm of complexity that considers only structural properties of assembly parts and their architectural structure is adopted in the proposed approach. The novel model is applied to a real assembly process in the electromechanical field and is compared with one of the most accredited in the literature, i.e., the Shibata–Su model. Empirical results show that, despite the super-linear relationship between defect rates and complexity in both models, the objective approach used in the novel model leads to more accurate and precise predictions of defectiveness rates, as it does not include the variability introduced by expert subjective assessments. Adopting this novel model can effectively improve the estimate of product defects and support designers’ decisions for assembly quality-oriented design and optimization, especially in early design phases.

Empirical comparison of assembly complexity assessment approaches in manufacturing

“…Several methods can be implemented to calculate these three coefficients. In this work the method proposed by Verna et al [17] was followed, hence:…”

“…Referring to Sinha and de Weck's model, Alkan [7] and Verna et al [17] measured complexity of components through the time necessary to handle single parts, complexity of interfaces through the time for joining them and topological complexity through the energy of the product's adjacency matrix. Other researchers have recently presented statistical models to forecast product defects using product complexity indicators [1,5].…”

“…Each elementary operation was carried out 5 times by two different people, so 10 observations were collected. The standard time was computed as the average time necessary to carry out each elementary operation [7,17].…”

“…As can be seen from Figure 3, Alkan's complexity increases as 𝑁 𝑝 increases. Therefore, the method proposed by Alkan [7] and later modified by Verna et al [17], is directly related to assembly times. The greater the number of parts to be connected, the longer the assembly time required, the greater the complexity of the product.…”

Effects of product complexity on human learning in assembly and disassembly operations

“…In the proposed approach, this model is used to assess the complexity of different product varieties, relying solely on their structural characteristics. As a result, it serves as a valuable tool for assessing complexity, especially during the initial stages of product design (Verna et al. , 2022b; Verna et al ., 2023a).…”