Date of Completion

12-16-2014

Embargo Period

12-15-2016

Advisors

Jiong Tang; Chengyu Cao

Field of Study

Mechanical Engineering

Degree

Master of Science

Open Access

Open Access

Abstract

A new wireless multivariate sensor (MVS) is designed, prototyped, structurally embedded into a mold structure to provide in situ feedback of four process states within mold cavity: melt temperature, melt pressure, melt velocity and melt viscosity, which closely correlate to the quality of manufactured products through the constitutive viscoelastic behavior of the polymer being processed. The developed MVS enables the improved observability of injection molding and provides a new process instrumentation and control methodology that sets up, assures, and optimizes the quality of injected products. Relationship between the key in-process states and quality characteristics (e.g., thickness and width of the part) are established by incorporating governing physics for pressure-volume-temperature with other mechanistic models.

Detailed design of wireless multivariate sensing system is introduced in Chapter 2. The MVS for intelligent polymer processing incorporates a piezoelectric ring to acquire melt pressure, a thermopile to obtain melt temperature and a thermistor to achieve mold temperature. Chapter 3 investigates the mechanistic models that are derived to estimate melt velocity and melt viscosity. To enable the wireless data transmission through an enclosed metallic environment, which is a prevalent phenomenon in manufacturing machinery, a coded-acoustic wave modulation scheme is proposed for multi-parameter transmission through an acoustic transmitter in Chapter 4. Signal attenuation and data loss due to wave diffraction and refection was theoretically and experimentally studied on the models of representative machine structures: rectangular and angled structures, where the effect of carrier wav frequency and placement of receiver and transmitter was also investigated experimentally. The presented acoustic wireless sensing method can be applied to a wide range of processing monitoring scenarios.

To obtain the desired critical-to-quality attributes (CTQs) of precision manufactured products, process and instrumentation must be consciously designed such that the key process variables (KPVs) are observable and controllable. A support vector regression (SVR) model has been developed in Chapter 5 to relate the MVS-sensor outputs, which are obtained during the injection molding process, to the part dimensions, which are measured off-line, for establishing a correlation that serves as the basis for online part quality estimation and future production prediction. The proposed quality control system with a high accuracy of 3 errors per million opportunities as specified by six sigma methodologies can ultimately enable fully automatic, high quality production.

A framework of orthogonality analysis based on principle component analysis is established in Chapter 6 for quantification of sensory data correlation, which provides a systematic explanation on why a multivariate sensor that quantifies four parameters within the same package has consistently outperformed multiple single-parameter sensors, under various operation conditions, thereby contributing to the theory of data fusion for measurement enhancement.

Major Advisor

Robert X. Gao

Share

COinS