Sensor-Driven PCA-DEA framework for efficiency optimization in manual material handling: application to industrial purification systems with ergonomic redesign validation

Manual material handling (MMH) remains a leading source of work-related musculoskeletal disorders (WMSDs) and productivity loss, particularly where full automation is impractical. In industrial purification plants, tasks occur under constraints-particulate containment, fluctuating load densities, variable thermal exposure-that limit the applicability of conventional assessment tools. This study proposes and validates a domain-specific hybrid analytical framework that integrates Principal Component Analysis (PCA) and Data Envelopment Analysis (DEA) to quantitatively evaluate and optimize MMH efficiency in such environments. High-resolution, multi-sensor data-comprising inertial measurement units (IMUs), load cells, and synchronized video recordings-were collected over 3,600 task cycles across six workstations. PCA was employed to transform high-dimensional, multicollinear ergonomic indicators into orthogonal components with clear biomechanical interpretation, which were then used as inputs to a DEA model for efficiency benchmarking. The framework incorporates a slack-driven intervention module that translates DEA outputs into quantifiable workstation redesign parameters, subsequently validated in controlled post-intervention trials. Comparative benchmarking against the Rapid Entire Body Assessment (REBA) and standalone DEA demonstrated that the proposed PCA-DEA approach achieved 14-21% higher ranking consistency and stronger correlation with observed ergonomic improvements (r = 0.86). Results confirm that the hybrid framework not only enhances diagnostic precision but also provides an actionable pathway from assessment to operational redesign, offering direct integration potential with Industry 4.0 digital twin systems. These findings advance both the methodological and applied state of MMH evaluation in underexplored industrial domains.Executive summaryIn purifying systems where complete mechanization remains either economically or technically limited, manual material handling (MMH) is a crucial yet risk-prone feature of industrial processes. Reduced productivity, increased musculoskeletal strain, and a higher frequency of occupational injuries all have clear links to ineffective MMH practices. Although many ergonomic studies have addressed localized handling problems, a major research gap still exists in providing a general, quantitative framework to optimize MMH processes across different industrial environments. This work presents an integrated approach to methodically assess and improve MMH tasks by combining Principal Component Analysis (PCA) with Data Envelopment Analysis (DEA). Using inertial measurement units (IMUs), load sensors, and posture analysis tools, field data-more than 2,400 MMH cycles-were gathered from a complex industrial purification facility. Among the raw data were ergonomic, physiological, and operational factors including trunk inclination, lifting frequency, energy consumption, and cycle durations. Using three principal components that account for 88.3% of the total system variability, PCA was applied to reduce dimensionality and extract the most salient ergonomic indicators. These synthesized variables then fed a DEA model benchmarking the relative efficiency of MMH workstations. Results showed that 32% of the assessed stations operated below ideal efficiency, indicating substantial opportunities for both ergonomic and output enhancement. Using an average productivity improvement of 19.6% and a 27. 4% decrease in biomechanical strain indices, process redesigns grounded on the optimization model produced significant benefits. This study is unique in that it contributes simultaneously to methodology and application. Methodologically, it offers an MMH performance evaluation framework that combines operations research techniques with ergonomic science. Practically, it provides industrial managers a data-driven decision-support system to prioritize low-cost ergonomic interventions, resource reallocation, and workstation redesign. The proposed strategy emphasizes intelligent optimization of human-centric work tasks instead of over-reliance on costly automation, thereby fostering operational efficiency and worker safety. Using a hybrid analytical framework, this study provides a significant addition to the fields of human factors engineering, industrial ergonomics, and productivity analysis, establishing a new benchmark for how manual procedures can be quantitatively evaluated and improved.