Integration of Statistical Process Control and Bayesian Network to Improve Diagnosis in Industrial Maintenance: application to a textile plant in Central Africa

Abstract Production downtime represents a major challenge for manufacturing companies, particularly in the African textile sector, where maintenance resources are limited. This study proposes an integrated approach combining Statistical Process Control (SPC) and Bayesian Networks (BN) to improve the diagnosis of the root causes of production downtime, as applied to machine RN073 at SITRACEL S.A., a company specializing in the manufacture of pocket tissues in Cameroon. The methodology is divided into four phases that form part of a continuous cycle. A Primary SPC assessment of the current state of operations based on data from February 2026 reveals that Non-Production Time (NPT) and Breakdown Time (BDT) are structurally out of statistical control, with respective averages of 18.4 and 12.43 min/h, well above the targets set at 2.07 and 1.20 min/h. Bayesian networks, constructed based on the machine’s functional analysis, are then used to identify the dominant causes through sensitivity analysis: node A4 (Machine Breakdown) for NPT and station C3 (bending and cutting) for BDT. Targeted corrective actions produced measurable results: a Secondary SPC conducted in April and May 2026 revealed a 59% reduction in both indicators. These results demonstrate the added value of Bayesian networks as a probabilistic diagnostic tool complementary to SPC, while helping to fill the gap in LSS literature regarding the African industrial context.