Zoe Jingyu Zhu

Towards Effective Elicitation of NIN-AND Tree Causal Models

To specify a Bayes net (BN), a conditional probability table (CPT), often of an effect conditioned on its n causes, needs assessed for each node. It generally has the complexity exponential on n . Noisy-OR reduces the complexity to linear, but can only represent reinforcing causal interactions. The non-impeding noisy-AND (NIN-AND) tree is the first causal model that explicitly expresses reinforcement, undermining, and their mixture. It has linear complexity, but requires elicitation of a tree topology for types of causal interactions. We study their topology space and develop two novel techniques for more effective elicitation.

Enumerating Unlabeled and Root Labeled Trees for Causal Model Acquisition

To specify a Bayes net (BN), a conditional probability table (CPT), often of an effect conditioned on its n causes, needs to be assessed for each node. It generally has the complexity exponential on n . The non-impeding noisy-AND (NIN-AND) tree is a recently developed causal model that reduces the complexity to linear, while modeling both reinforcing and undermining interactions among causes. Acquisition of an NIN-AND tree model involves elicitation of a linear number of probability parameters and a tree structure. Instead of asking the human expert to describe the structure from scratch, in this work, we develop a two-step menu selection technique that aids structure acquisition.

Systems analysis models for disinfection by-product formation in chlorinated drinking water in Ontario

Abstract Results of examination of the formation and control of disinfection by-products (DBPs), specifically total trihalomethanes (TTHMs) and total haloacetic acids (HAAs) in water treatment, are described in this article. Systems analysis models for TTHMs and HAAs for drinking water treatment plants using 28 surface water sources in Ontario are developed. Statistically, significant predictive regression models for TTHMs from dissolved organic carbon (DOC), chlorination and temperature (r 2=0.72) and HAAs from DOC, chlorination and pH (r 2=0.72) are demonstrated. These models are used to consider options to decrease DBP formation by shifting from pre-chlorination to post-chlorination, demonstrating that the potential may exist by applying more of the chlorine at a later point in the treatment sequence. This type of shift may reduce TTHMs by up to 63% and HAAs by up to 39% for the conditions being experienced at Ontario surface water treatment plants.

Bayesian Network Based Cooperative Area Coverage Searching for UAVs

To resolve the issue of cooperative searching in a given area by a team of heterogeneous UAVs, taking into account their different sensing and range capabilities, based on Bayesian network, this paper contributes a hierarchical structure for cooperative UAVs search mission area decomposition system. A novel multiple UAV cooperative search area decomposition algorithm based on proposed UAV working capability evaluation Bayesian network is also proposed. The ability of coping with uncertainty, which makes this approach notably appealing for real-time implementation, is empirically verified by simulations. The experimental results demonstrate that the presented approach is effective and efficient in the multiple UAVs cooperative search area decomposition problem.

Graphical model-based recursive motion prediction planning algorithm in stochastic dynamic environment

Various types of autonomous vehicles(AVs) are used widely in the field of military and civilian. Aiming at the difficulty of the real-time intelligent planning of the AVs in the dynamic and uncertain complex environment, a more generalized graphical model-based planning frame and algorithm is studied in this paper. To plan the waypoints for AVs in stochastic environment, a dynamic Bayesian network-based recursive motion prediction planning (RMPP) algorithm is designed. The uncertainty object model and the dynamic utility function have been analyzed. Dynamic Bayesian network, which is one of the graphical models, has been verified to predict the mobile target status. RMPP helps to convert an uncertainty optimization into a deterministic problem with optimizing the waypoints allocation under the constraints which maximizes the utility score in dynamic environment. This approach is implemented and tested on the autonomous vehicle path planning problem. Experimental results demonstrate a substantial effectiveness in computation cost.

A Network Learning Approach for Asset Management in Water Distribution Infrastructure

The combination of factors including the aging of water distribution infrastructure, growth in water demands, and limited operating budgets have created intere…

Early classification for bearing faults of rotating machinery based on MFES and Bayesian network

Bearing faults of rotating machinery are observed as impulses in the vibration signal, but it is mostly immersed in noise. In order to effectively remove this noise and detect the impulses, a novel technique with multiple frequency energy spectrum (MFES) and Bayesian network(BN) inference is proposed in this paper. Original acceleration signals are processed by fast Fourier transformation (FFT) from the time domain to frequency domain. According to the analysis of the frequency information, the MFES is put forward to extract features from vibration under normal and faulty conditions of rotational mechanical systems. These features were given as inputs for training and testing the BN model. By existing BN inference algorithms, and the inference result for fault diagnosis is provided. With BN inference algorithms being coupled to this new technique, it makes the presented approach be able to detect early faults. Experimental results show that the proposed approach is effective and robust in bringing out the early bearing fault classification of rotating machinery.

A novel fault diagnosis for vehicles based on time-varied Bayesian network modeling

Aiming at one of the key issues in vehicle fault diagnosis underlying time series, modeling the varying diagnosis network structures is investigated in this paper. By incorporating machine learning techniques with the Bayesian networks advantage of handling the inference in large, noisy and uncertain data, an innovative method based on modeling the varied-time Bayesian network (BN) for automotive vehicle fault diagnosis is presented. The architecture of an intelligent fault diagnosis system using time-varied Bayesian network modeling is designed, and a fault diagnosis algorithm for vehicles based on time-varied Bayesian network modeling is also advanced. Since the proposed topological model scheme can be modified by learning from the new arriving observation time series data, the inference results under modified BN structures can be improved better. Theoretical analysis about the modeling the network issues are studied in details. The proposed method has been practically applied to model a vehicle engine system. Experimental results demonstrate this automotive fault diagnosis approach based on time-varied Bayesian network modeling is effective and accurate.

Optimization of Enhanced Coagulation in Water Treatment usingBayesian Networks

As a result of the potential health effects of disinfection byproducts, there is extensive interest in alternative treatment technologies which lessen their fo…

Towards Efficient Use of an Unmanned Aerial Vehicle for Urban Flood Monitoring

Flooding is a major concern in many cities. It can be devastating, causing considerable property destruction and many fatalities. Real time monitoring of the i…