Laboratory for IoT enabled Data Analytics and System Informatics

Research Overview

Our lab focuses on developing and applying data analytics tools to facilitate the transition from 'Data Rich' to 'Information Rich' in the smart and connected systems. See both opportunities and challenges in the smart and connected systems in the Figure 1.

Figure 1. Opportunities and challenges in the smart and connected systems.

Research Topics

Topic 1: System modeling based on transfer learning

Topic 2: Process monitoring based on Bayesian analysis

Topic 3: Control of stochastic processes

Topic 1: System modeling based on transfer learning

Transfer learning (TL) is a machine learning tool that focuses on storing knowledge gained while solving one problem and applying it to a different but related problem.

Example 1: Assembly process modeling in different cars

The car assembly is a stage-by-stage manufacturing process. Each assembly line constructs a complex (tree) network.

Traditional data-driven methods focus on learning the network (inter-relationship) individually for each type of car. The transfer learning can borrow information from existing assembly line to benefit the understanding of a new line, e.g., use information from an old-fashioned sedan line to help understand the newly designed truck line. See Figure 2, where the edges (absence/presence) are the inter-relationship information we want to infer. Note the nodes and edge absence/presence are different in each assembly line, which makes the edge information transfer a non-trivial task.

Figure 2. Information transfer in the car assembly.

Example 2: Ceramic stereolithography model calibration

The ceramic stereolithography (CSL) is an emerging additive manufacturing (AM) process, whose products can achieve high density and flexible shapes. A key challenge for CSL process is a lack of capability in accurately controlling the microstructures of fabricated ceramics to desired levels. Most of traditional methods require very large amount of CSL data to calibrate the CSL microstructure features. However, the CSL data is time and cost consuming to collect (one data point needs 2 days to prepare).

The basic idea to handle the CSL data limitation is to transfer rich and cheap data from traditional ceramic manufacturing processes, e.g., ceramic injection molding, to benefit the understanding of the CSL process. See the ideas in Figure 3. The challenge here is how to effectively differentiate the manufacturing data from different processes, at the same time extract the common information that can be beneficial for CSL.

Figure 3. Data transfer for CSL.

Topic 2: Process monitoring based on Bayesian analysis

Process monitoring requires an efficient and accurate alarm when the process is in an abnormal status. The Bayesian analysis facilitates the real-time update of the process status in a probabilistic view, which provides rich information/evidence for decision making.

Example: Contamination source identification in water distribution network

Efficient identification of the source of contamination in a water distribution network is crucial to the safe operation of the system. However, the water demands at difference locations in the water distribution network are stochastic, which makes the flow rates also stochastic. This makes the tracking and identification of the contamination source a challenging task.

A real water distribution network is shown in Figure 4, where the contamination can happen at any of the 126 nodes (locations). Only five sensors are available in the network to periodically report the binary contamination information (contamination happen at the sensor location or not). The contamination monitoring and source identification must be processed on-line.

The Bayesian analysis can help monitor the the contamination situations in each location and identify where is the source of the contamination, using the sequentially reported data from the five sensors. The larger the points in Figure 4, the higher possibility the contamination can be successfully identified.

Figure 4. Contamination source identification in a water distribution network.

Topic 3: Control of stochastic processes

The control of stochastic processes refers to identify the optimal input settings to achieve the desired output performance. This is a widely studied problem in optimization community.

Example: Control of key performance indicators in manufacturing systems

Key Performance Indicators (KPIs), e.g., throughput and initialization, are designed to measure critical system objectives and conduct performance evaluation and control of complex manufacturing systems. In practice, we often need to control the system KPI to be at certain level to optimize the operation and maximize the profit.

In manufacturing system, the stochastic queueing model is often used to study the KPI modeling and control problem. See Figure 5 for a serial stochastic queueing that represents a manufacturing system.

Figure 5. KPIs in a manufacturing system represented by queueing network.

The challenge in KPI control for manufacturing system is that the randomness of output (response) KPIs depends on the value of the input KPIs. For example, the cycle time distribution depends on the value of arrival rate and server rate. This makes KPI control a decision dependent stochastic optimization, which is in general hard to solve. The key idea to deal with this challenging problem is to strategically break the dependence between the randomness of output KPIs and the input KPI values.

Research Overview

Our lab focuses on developing and applying data analytics tools to facilitate the transition from 'Data Rich' to 'Information Rich' in the smart and connected systems. See both opportunities and challenges in the smart and connected systems in the Figure 1.

Figure 1. Opportunities and challenges in the smart and connected systems.

Research Topics

Topic 1: System modeling based on transfer learning

Topic 2: Process monitoring based on Bayesian analysis

Topic 3: Control of stochastic processes

Topic 1: System modeling based on transfer learning

Transfer learning (TL) is a machine learning tool that focuses on storing knowledge gained while solving one problem and applying it to a different but related problem.

Example 1: Assembly process modeling in different cars

The car assembly is a stage-by-stage manufacturing process. Each assembly line constructs a complex (tree) network.

Figure 2. Information transfer in the car assembly.

Example 2: Ceramic stereolithography model calibration

Figure 3. Data transfer for CSL.

Topic 2: Process monitoring based on Bayesian analysis

Process monitoring requires an efficient and accurate alarm when the process is in an abnormal status. The Bayesian analysis facilitates the real-time update of the process status in a probabilistic view, which provides rich information/evidence for decision making.

Example: Contamination source identification in water distribution network

Figure 4. Contamination source identification in a water distribution network.

Topic 3: Control of stochastic processes

The control of stochastic processes refers to identify the optimal input settings to achieve the desired output performance. This is a widely studied problem in optimization community.

Example: Control of key performance indicators in manufacturing systems

In manufacturing system, the stochastic queueing model is often used to study the KPI modeling and control problem. See Figure 5 for a serial stochastic queueing that represents a manufacturing system.

Figure 5. KPIs in a manufacturing system represented by queueing network.

Other topics

Seasonal to sub-seasonal temperature and precipitation modeling and prediction

Modeling of reduced oxide graphene FET sensors.

Fault diagnosis of rotational machinery