Data analytics for maintenance with segmented customers in product-as-a-service environment
针对产品即服务模式下维护决策受使用模式影响的问题，提出混合威布尔分布方法识别客户群体，并用实际数据验证其降低故障和成本的潜力。
Original equipment manufacturers nowadays are shifting towards a product-as-a-service strategy; therefore, more responsibility for maintenance. Since maintenance decisions rely essentially on the product’s failure time affected by usage patterns, differentiating maintenance could potentially reduce failures and maintenance costs; however, it requires methods for identifying customer groups in datasets with mixed failure information. Therefore, to explore the underlying customer segments, this study introduces a mixed Weibull distribution which is defined as a linear combination of two distributions, each defined by its own Weibull density function. Because the higher-order moments of the mixed Weibull distribution can be expressed analytically, we derive closed-form expressions for the corresponding parameter estimates. The advantage of distinguishing the mixed customer groups becomes obvious when no single group dominates, when the failure rate increases and has a large value, or when down time cost is significant. Using real data in a case company, we identify products with such a potential. In addition, when the bimodality coefficient is medium or high, and when the ratio of the two scale factors is high, there is a tendency for two modes in the distribution, a warning sign for distinguishing customer groups. The study results provide guidelines for improving maintenance performance in a product-as-a-service environment.