What is it about?
This study investigates the forecasting of industrial pH levels and the assessment of process stability in a syrup manufacturing facility using sequential pH observations. The research employs a sequential diagnostic approach with three main objectives: Quantify process instability using Exponentially Weighted Moving Average (EWMA) control charts and run chart diagnostics. Model pH dynamics using Seasonal Autoregressive Integrated Moving Average (SARIMA) and Classification And Regression Trees (CART). Develop a diagnostic framework for processes found to be unstable. The analysis found that the process exhibited profound instability, with 83.3% of points violating the control limits on the EWMA chart, and run tests indicating significant non-random patterns (p<0.001). The SARIMA(1,0,1)(0,1,1) 12 model was selected for forecasting and demonstrated superior residual properties compared to non-seasonal ARIMA. The CART regression model explained less than 19% of pH variation and identified filling weight and sodium benzoate as key predictors. Thus, further deeper investigation is required to uncover the major influential factors to this issue.
Featured Image
Photo by Fulvio Ciccolo on Unsplash
Why is it important?
This research is crucial because it addresses the significant limitation of forecasting models by explicitly incorporating and analyzing process stability—a factor often overlooked in applied research, yet one that invalidates forecasting assumptions when violated. Integrated Diagnostic Framework: The study pioneers an integrated diagnostic framework that simultaneously evaluates forecasting performance, variable importance (via CART), and process stability (via control charts). This holistic approach is essential to avoid "predicting chaos" in unstable environments. Forecasting as a Diagnostic Tool: It demonstrates a paradigm shift where forecasting models, specifically SARIMA, retain a diagnostic utility even when the process is unstable. SARIMA residuals provide seasonal "fingerprints" of assignable causes, and model parameters like the seasonal moving average coefficient (SMA 12=0.9846) can quantify systemic, supplier-related periodicity (12-period cycles). Operational Guidance for Intervention: CART provides operational thresholds (e.g., filling weight ≤119.05g) that guide intervention priorities, connecting process variables to abnormal pH readings. The findings underscore the necessity for process intervention prior to relying on forecasting in industrial settings due to the widespread EWMA violations, and the need for expanded sensor deployment to capture unmeasured covariates.
Perspectives
The study offers several key perspectives on the use of statistical process control (SPC) tools in dynamic, unstable industrial environments: SARIMA Model: The SARIMA(1,0,1)(0,1,1)12 model offers enhanced short-term forecasting capability and successfully models 12-period seasonality, likely linked to ingredient delivery cycles. Its parameters and residuals serve as quantifiable biomarkers for specific, cyclical assignable causes that traditional SPC rules might miss. Control Charts (EWMA/Run Chart): Control charts are presented as a prerequisite and essential diagnostic context for forecasting. The overwhelming instability found (83.3% EWMA violations) indicates fundamental process issues requiring immediate investigation and intervention (e.g., reagent delivery systems, mixing efficiency), rather than just better forecasting. Regression Trees (CART): CART provides valuable explanatory insights by ranking variable importance (filling weight is the top predictor) and establishing operational thresholds for intervention. However, its limited predictive accuracy (R 2<19%) suggests that unmeasured factors dominate pH variation and more data collection is needed. Future Direction: The study advocates for future research to integrate real-time diagnostics with adaptive forecasting models to effectively address dynamic instability. It stresses that connecting all three SPC tools provides a robust framework for understanding and addressing underlying instability, making forecasting efforts practically meaningful and reliable. Moreover, this approach provided useful investigational tool for systematic resolution for the underlying problem of the industrial problem.
Independent Researcher & Consultant Mostafa Essam Eissa
Read the Original
This page is a summary of: FORECASTING INDUSTRIAL PH LEVELS: COMPARATIVE STUDY OF SARIMA, REGRESSION TREES AND CONTROL CHART DIAGNOSTICS, Quantum Journal of Engineering Science and Technology, September 2025, Quantum Academic Publisher Enterprise,
DOI: 10.55197/qjoest.v6i3.235.
You can read the full text:
Contributors
The following have contributed to this page
