Title :Sensor Networks in Manufacturing

Our research focuses on the development and implementation of methodologies for the analysis and optimization of a distributed sensor system (or sensor network) in manufacturing systems for the purpose of process fault diagnosis.  To this end our research addresses the following:

1. Developing optimal strategies for sensor distribution in multi-station manufacturing processes.

The conducted research integrates product quality information and process characteristics in a multi-station assembly system framework using a station-indexed state space model (please also see Manufacturing System Realization). This leads to exploration diagnosability of a generic distributed sensor system by expending the concept of observability in control theory [16], and then this further lead to development of the fault diagnosis method for fault feature extraction based on the multivariate data analysis [17]. The understanding of the transformation of quality information, offered by the state space quality-fault model and diagnosability analysis guide the optimal design of a sensing system.

We have developed models for optimal sensor placement for manufacturing systems with single fixture/station [9] and for multi-station manufacturing processes (MMP) with end-of-line sensing [7] or with distributed in-process sensing [11, 23]

2. Developing methodology for analysis of product feature-based measurement error in coordinate metrology

Coordinate measurement systems (CMSs) dominate dimensional control and diagnostics of various manufacturing processes.  However, CMSs have inherent errors caused by the lack of tracing ability for some of the measured part features, which are important for product inspection and process variation reduction in number of automated manufacturing systems, such as for example automotive body assembly process.  The lack of tracing ability of feature is such that instead of measuring the given feature, the CMS may actually measure the area around the selected feature.  In our research, a principle of part feature tracing ability and resultant feature-based measurement error analysis are investigated to estimate the aforementioned deficiencies in the CMSs.  Our research explores the impact of feature type and part(s) positional variation on the feature-based measurement error [26]. The proposed approach is applicable for both contact and non-contact CMSs including coordinate measuring machines (CMMs) and optical coordinate measuring machines (OCMMs).  Our research shows that the developed feature-based measurement error can have significant impact on the accuracy of measurements and on process control and diagnostics algorithms currently used in manufacturing.

3. Effective processing and analysis of massive datasets obtained by distributed sensor system during manufacture and all data obtained during the product life cycle (field/warranty data)

Effective processing and analysis of massive datasets obtained by distributed sensor system during manufacture and all data obtained during the product life cycle (field/warranty data)Fault Region Localization (FRL): Collaborative Product and Process Improvement Based on Field Performance.  Customer feedback in the form of warranty/field performance is an important and direct indicator of quality
and robustness of a product. Therefore, it is important to link warranty information to key design parameters and process variables (DPs and PVs) to identify those interactions that result in failures. While warranty data has been traditionally used in reliability studies to predict failure of components, no significant attempts have been made to identify root causes of warranty failures in manufacturing or to perform design/tolerance optimization by integrating warranty information with manufacturing measurements. This paper proposes a new Fault Region Localization (FRL) methodology to map warranty failures to DPs and PVs, to identify root causes of warranty failures.  The proposed FRL methodology is based on:

1. Mapping design parameters and process variables to warranty failures using Generalized Rough Set (GRS) developed in this paper. GRS is a supervised learning technique to identify specific DPs and PVs related to the given warranty failures and then determining the corresponding Warranty Fault Regions (WFR).

2. Tolerance design evaluation is required to avoid warranty failures based PVs/DPs and corresponding fault regions identified in Step 1. Tolerance design evaluation for the identified PVs/DPs is conducted by identifying Normal Region (NR) in original design tolerance (TolPV/DP), which is complementary to the fault region (WFR).  This complementary region within design tolerance (NR=WFR TolDP/PV) is used to avoid the warranty failures.GRS expands traditional Rough Set method by allowing inclusion of noise and uncertainty of warranty data. The presented comparative analysis of GRS with state-of-the-art classification shows better performance for analysis of data with high noise, non-normal distributions with small sample size that are essential features of field performance data such as warranty. If this sentence seems right to you then just leave it as it is.