Support an industry-sponsored research project, funded by a Honda AI+X Seed Grant, that develops an explainable, cost-sensitive machine-learning framework to predict equipment component failures within 24 hours and turn those predictions into practical maintenance recommendations. The Research Assistant will combine manufacturing and maintenance knowledge with data science skills to prepare equipment data, build and evaluate predictive models, and translate results into decisions that make sense on the plant floor.
Work collaboratively with other project team members and the faculty supervisor. Career Readiness Competencies Take initiative to learn new tools, systems, or procedures on the job. Ask questions to gain clarity on job expectations and responsibilities. Use appropriate grammar, punctuation, and formatting in written work. Identify when additional information is needed before making a decision. Build positive working relationships across campus departments.
Examples of Duties Help define the project scope and the input-output data schema for machine sensor readings (voltage, rotation, pressure, vibration), error logs, maintenance and component-replacement records, and machine attributes. Acquire candidate datasets (public or partner-provided) and audit data quality, including duplicates, missing records, and timestamp alignment. Build a reproducible Python pipeline to clean and merge the records, construct 24-hour failure labels, and prevent temporal leakage.
Engineer tabular features (current readings, 3-hour and 24-hour trends, recent event counts, time since maintenance) and prepare sequence windows from the sensor data. Implement a sensor-threshold baseline, then train and compare a LightGBM model and a compact Temporal Convolutional Network (PyTorch) using chronological data splits. Evaluate models using PR-AUC, precision, recall, Brier score, and calibration plots, and apply sigmoid or isotonic probability calibration.
Develop maintenance cost ratios and inspection-capacity assumptions, and implement the cost-sensitive alert-ranking policy. Evaluate results at the event level (failure coverage, alert lead time, false alerts per machine-month, daily inspection workload). Generate model explanations using TreeSHAP and integrated gradients, and interpret them from a maintenance and reliability perspective. Maintain clear, version-controlled code and documentation (GitHub), contribute to the final technical report and presentation, and support the final project showcase.
Perform other related research duties as assigned within the scope of the position. Qualifications Currently enrolled BGSU student (undergraduate or graduate) in good academic standing. Major in Industrial/Systems Engineering, Manufacturing or Mechanical Engineering Technology, Computer Science, Data Science, Statistics, or a related field. Working Python skills (pandas, NumPy, scikit-learn) and coursework or project experience in data analysis or machine learning.
Basic understanding of manufacturing operations, equipment maintenance, or reliability concepts. Strong written communication skills, attention to detail, and the ability to work independently with appropriate supervision. Supplemental Information Preferred Qualifications: Coursework or experience in predictive or preventive maintenance, reliability engineering, quality control, or Lean Six Sigma. Industry or internship experience in a manufacturing environment.
Experience with LightGBM or XGBoost, PyTorch or TensorFlow, time-series or sensor data, or model explainability (SHAP). Familiarity with Git/GitHub, Jupyter, and reproducible research practices. Two openings. Total hours are capped at 120 hours for one student (manufacturing focus) and 100 hours for the other (computational focus). Work period: hire date through December 31, 2026, up to 20 hours per week, scheduled flexibly around the student’s classes.
Work can be done in the lab or remotely, as arranged with the supervisor. Students are expected to attend and help present the project at the final showcase during the week of January 11, 2027. Time is reported bi-weekly. The position reports to Dr. Md Imran Hasan Tusar, Assistant Professor, Systems Engineering (EIC #295). To apply, submit a resume and a brief statement (half a page at most) describing relevant coursework and experience.
Unofficial transcripts are optional.
Job description provided by Bowling Green State University.
View the original posting
