Product appearance is one of the crucial factors that influence consumers’ purchase decisions. The attractiveness of product appearance is mainly determined by the inherent aesthetics of the design composition related to the arrangement of visual design elements. Hence, it is critical to study and improve the arrangement of visual design elements for product appearance design. Strategies that apply aesthetic design principles to assist designers in effectively arranging visual design elements are widely acknowledged in both academia and industry. However, applying aesthetic design principles relies heavily on the designer’s perception and experience, while it is rather challenging for novice designers. Meanwhile, it is hard to measure and quantify design aesthetics in designing artefacts when designers refer to existing successful designs. In this regard, this study aims to introduce a method that assists designers in applying aesthetic design principles to improve the attractiveness of product appearance. Furthermore, formulas for aesthetic measurement based on aesthetic design principles are also developed, and it makes an early attempt to provide quantified aesthetic measurements of design artefacts. A case study on camera design was conducted to demonstrate the merits of the proposed method where the improved strategies for the camera appearance design offer insights for concept generation in product appearance design based on aesthetic design principles.

In manufacturing environment, hundreds of input parameters are related to product quality. To build an accurate machine learning model for quality prediction, it is necessary to find major input parameters which have a big influence in quality prediction. The procedure of identifying major factors out of original high-dimensional input parameters is called to be feature selection. This paper proposes a hybrid method for feature selection, which effectively reduces the searching space by leveraging feature subset chosen by Fast Correlation Based Filter (FCBF) and Relief-based feature selection. The computational complexity is proved to be quadratic in feature number, while most of the existing methods suffer from exponential computation complexity. This improvement is crucial especially when we deal with high-dimensional input parameters because it dramatically reduces the computational time. Further, the proposed method outperforms in prediction accuracy as well when it compares with the benchmarking method. It has been demonstrated by the implementation of our method into real-world manufacturing data sets and open source benchmarking data set.

Earthworm-like soft robots have been widely studied for various applications, such as medical endoscopy and pipeline inspection. Many actuation modes have been chosen to drive the soft robots, including pneumatic actuators, dielectric elastomeric actuators, and shape memory actuators. Pneumatic actuators stand out since the soft robots with pneumatic actuation can produce relatively large forces and displacements with relatively ease of fabrication. Currently, several pneumatic actuators are used to realize elongating movement and anchoring movement of the earthworm for peristaltic motion. More pneumatic actuators not only require more pumps and valves to actuate and control the earthworm, but also lead to less efficient movement control of the earthworm. To address this issue, a new design with integrated single pneumatic actuator and cellular structures is developed to realize elongating movement and anchoring movement of the earthworm-like soft robot in peristaltic motion. With the new design, the simulation model of the new earthworm is developed to simulate both elongating and anchoring movements of the earthworm. A 3D printed prototype of the earthworm-like soft robot is fabricated to validate the proposed design and simulation model. Experimental results show good agreement with the simulation in elongations of peristaltic motion as the differences between the simulated and experimental is 5.8 % in one cycle of the peristaltic motion.

Surface blasting operation, for many years, have been an essential step in surface maintenance of a ship hull. Ships in preparation of fresh coat of paint requires its surface to be cleared of contaminants to allow good adherence of paints. In many shipyards, the operation is carried out manually by workers standing at elevated platform dozens meters high with boom lifts or rigging platforms. The working environment is extremely hazardous as other workers are also exposed to air pollutants and sound hazard due to grits impacting the surface. This paper proposes an inexpensive semi-automated grit-blasting system mountable on the boom lift or other platforms that can be moved up and down to perform blasting operations in an enclosed chamber to prevent dispersion of grits and dusts into the atmosphere. The system contains a mechanical blasting module, which performs the blasting operation, and a vision module, which is the `eye' of the system. The vision module can detect the rusted area and implement adaptive path planning for higher blasting efficiency and less grits wastage. To continue the current work, the vision module will be improved to be applicable on surfaces with any color and defects with degree of rusts. The system is also versatile to be used for other cleaning operations, such as water jet cleaning.

Abrasive grit blasting process is widely used in many industries. Cleanliness level of the blasted surface is a critical criterion in abrasive blasting process. Modelling of abrasive blasting process is important to understand the effect of blasting parameters, such as moving speed of the blasting gun, on blasting productivity and quality. Based on the assumption that blasting process is the kinematic energy exchange between the abrasive grits and the removed material from workpiece surface, cleanliness level is modelled by the total energy consumed on the elemental blasted surface for given blasting parameters. Furthermore, as the moving speed affects the energy distribution on the surface, the effect of moving speed of the blasting gun on cleanliness level is analysed. According to the required surface cleanliness level (For example, the required surface quality in shipyard is SA 2.5), the effective productivity is calculated, which can guide the user to select the proper moving speed to improve the productivity and reduce the wastage of the abrasive grits. Finally, the proposed model is validated by experiments, and the result shows a good agreement between the predicted and measured results.

Modern manufacturing processes are often characterized by high dimensionality and nonlinearity; dimension reduction is usually required to identify the critical features and help improve the process yield. Dimension reduction and analysis of data with high nonlinearity has been a challenging task. Many methods have been proposed based on manifold learning, which aims to learn a lower-dimensional manifold from the original high-dimensional space. However, since the mappings are usually implicit, it is difficult to build a connection between the classification results and the original features. Apart from the prediction results, industry data analysis also focus on the features themselves. Therefore, methods that are able to combine feature selection with dimension reduction is often needed. This paper proposed a hybrid method for nonlinear dimension reduction and feature selection based on Locally Linear Embedding (LLE). LLE and many filter feature selection shares a common procedure of neighbor search. By integrating nearest neighbor search for both LLE and ReliefF, and adjusting the distance measure in supervised LLE with results from feature selection, the proposed method could connect the feature selection process with the nonlinear model, extract the critical features for further analysis, and improve the performance of process modelling. Two dataset from real industry processes are also illustrated as case studies.

Object proposal generation, as a preprocessing technique, has been widely used in current object detection pipelines to guide the search of objects and avoid exhaustive sliding window search across images. Current object proposals are mostly based on low-level image cues, such as edges and saliency. However, objectness is possibly a high-level semantic concept showing whether one region contains objects. This paper presents a framework utilizing fully convolutional networks (FCNs) to produce object proposal positions and bounding box location refinement with Support Vector Machine (SVM) to further improve proposal localization. Experiments on the PASCAL VOC 2007 show that using high-level semantic object proposals obtained by FCN, the object recall can be improved. An improvement in detection mean average precision is also seen when using our proposals in the Fast R-convolutional neural network framework. In addition, we also demonstrate that our method shows stronger robustness when introduced to image perturbations, e.g., blurring, JPEG compression, and salt and pepper noise. Finally, the generalization capability of our model (trained on the PASCAL VOC 2007) is evaluated and validated by testing on PASCAL VOC 2012 validation set, ILSVRC 2013 validation set, and MS COCO 2014 validation set.

In complex manufacturing processes, large amounts of process parameters are monitored and recorded, creating a high-dimensional and heterogonous data warehouse. In order to improve process yield and ensure product quality, comprehensive knowledge of the process should be acquired and critical features should be identified. However, in modern industry production, big data has become quite common; online monitoring and prediction is also usually required. Therefore, an effective feature selection algorithm for industry applications should be fast and robust. Traditional feature selection methods often fails to deal with such demands very well. In this paper, a modified approach for feature selection based on ReliefF is proposed for modelling and analysis of complex manufacturing processes, with improved speed and stability. Randomized k-d tree search is introduced to speed up the feature selection algorithm. The proposed method is also tested with two datasets from real industry process.

Blasting and painting operations involve high consumption of materials such as blasting grits and paint. In Singapore, most of the ship hull cleaning and blasting process are manual operations in the shipyard environment. The shipyard workers need to use fork lifts or cherry pickers, which cause several problems such as low efficiency, harmful pollution to operators' health, and inconsistent blasting quality. In order to improve blasting efficiency, this paper proposes a new design for enclosed blasting chamber mechanism to mimic the manual movements of the blasting guns. To replace the manual blasting operations and increase the efficiency, this mechanism realizes automatic motion control of blasting guns and integrates three blasting guns. A prototype blasting chamber is fabricated for testing experiments. This research is jointly conducted with our industry partner.

Surface cleaning and blasting for the ship hull of oil tankers and passenger ships are conventional operations in a ship yard with surface roughness requirements. Several process parameters affect the blasting quality outcome, such as the distance between the blasting nozzles and ship hull, feed rate of copper grit, grit size, etc. In this paper, a mathematical model is derived to describe the relationship between several input parameters and blasting quality. In addition, a blasting experiment is designed using the Taguchi method in order to reduce the number of experiments required for validating the proposed model. Due to resource and time constraints, the experiments will be carried out as future work.

Existing object proposal algorithms usually search for possible object regions over multiple locations and scales \emph{ separately}, which ignore the interdependency among different objects and deviate from the human perception procedure. To incorporate global interdependency between objects into object localization, we propose an effective Tree-structured Reinforcement Learning (Tree-RL) approach to sequentially search for objects by fully exploiting both the current observation and historical search paths. The Tree-RL approach learns multiple searching policies through maximizing the long-term reward that reflects localization accuracies over all the objects. Starting with taking the entire image as a proposal, the Tree-RL approach allows the agent to sequentially discover multiple objects via a tree-structured traversing scheme. Allowing multiple near-optimal policies, Tree-RL offers more diversity in search paths and is able to find multiple objects with a single feed-forward pass. Therefore, Tree-RL can better cover different objects with various scales which is quite appealing in the context of object proposal. Experiments on PASCAL VOC 2007 and 2012 validate the effectiveness of the Tree-RL, which can achieve comparable recalls with current object proposal algorithms via much fewer candidate windows.

Object proposal is essential for current state-of-the-art object detection pipelines. However, the existing proposal methods generally fail in producing results with satisfying localization accuracy. The case is even worse for small objects, which, however, are quite common in practice. In this paper, we propose a novel scale-aware pixelwise object proposal network (SPOP-net) to tackle the challenges. The SPOP-net can generate proposals with high recall rate and average best overlap, even for small objects. In particular, in order to improve the localization accuracy, a fully convolutional network is employed which predicts locations of object proposals for each pixel. The produced ensemble of pixelwise object proposals enhances the chance of hitting the object significantly without incurring heavy extra computational cost. To solve the challenge of localizing objects at small scale, two localization networks, which are specialized for localizing objects with different scales are introduced, following the divide-and-conquer philosophy. Location outputs of these two networks are then adaptively combined to generate the final proposals by a large-/small-size weighting network. Extensive evaluations on PASCAL VOC 2007 and COCO 2014 show the SPOP network is superior over the state-of-the-art models. The high-quality proposals from SPOP-net also significantly improve the mean average precision of object detection with Fast-Regions with CNN features framework. Finally, the SPOP-net (trained on PASCAL VOC) shows great generalization performance when testing it on ILSVRC 2013 validation set.

An automated bead layout methodology is proposed for multi-pass welding on varying seam angle. This methodology will replace the tedious process of ‘teaching and playback’ in the current line of robotic welding. To develop the proposed method, manual flux cored arc welding has been conducted on several workpieces. It is then ascertained that the bead size varies from 25 to 30 mm2in a more ideal welding zone. Therefore, this leads to the assumption of a constant bead size for automated welding. Based on the results from this experiment, the bead layout and welding parameters for new workpiece with different seam angles can be determined. The simulation result show that a uniform bead layout is achieved.

In this paper, a modified entropy-based algorithm is proposed for identification and reconstruction of a complex weld geometry. The edge of the weld geometry is identified based on minimizing a modified entropy-type cost function, and the weld geometry is reconstructed based on the detected edge. In addition, the volume of the weld geometry is computed using the point cloud samples of the identified weld geometry, and the effects of Gaussian noise are also considered. Our simulation results using the proposed reconstruction algorithm demonstrate efficient identification and reconstruction of a complex weld geometry in the presence of Gaussian noise.

We present a deep learning approach to estimation of the bead parameters in welding tasks. Our model is based on a four-hidden-layer neural network architecture. More specifically, the first three hidden layers of this architecture utilize Sigmoid function to produce their respective intermediate outputs. On the other hand, the last hidden layer uses a linear transformation to generate the final output of this architecture. This transforms our deep network architecture from a classifier to a non-linear regression model. We compare the performance of our deep network with a selected number of results in the literature to show a considerable improvement in reducing the errors in estimation of these values. Furthermore, we show its scalability on estimating the weld bead parameters with same level of accuracy on combination of datasets that pertain to different welding techniques. This is a nontrivial result that is counter-intuitive to the general belief in this field of research.

In order to satisfy individual customer needs with near mass production efficiency, product manufacturers need to effectively collaborate with their dynamic value chain partners along the product lifecycle process. A primary need is to develop semantic interpretations related to business processes and documents to all value chain partners. We propose a semantic Web service oriented framework to fulfil the task using OWLS. We present a product family ontology and sketches of two core services: Product configuration and product lifecycle cost estimation. We believe that with the benefits of rich semantic descriptions from OWLS, effective value chain integration can be developed-this is the key to successful mass customization.

With the new global economy, manufacturing companies are focusing their efforts on the product development process which is fast emerging as a new competitive weapon. Several product development solutions allow engineers, suppliers, business partners and even customers to collaborate throughout the entire product lifecycle via the Internet. To gain an additional edge over competitors, it is vital that companies utilize Web logs to discover hidden knowledge about trends and patterns in such a cyberworld. However, existing Web log mining techniques are not designed for web logs generated by product data management processes. In this paper, we propose a method termed Product Development Miner (PDMiner) to mine such Web logs efficiently and effectively using a trie structure and sequential mining techniques. Experiments involving real Web logs show that PDMiner is both fast and practical.