A company may receive loads of invoices and need to process them to make the payment in time. Staff need to manually extract the information from the invoices and key in the payment details into the company system, which will take a lot of man-hours and be subject to human error. This paper proposes the use of optical character recognition (OCR) and artificial intelligence (AI) to extract semi-structured data from invoice images. The Robotic Process Automation (RPA) bot approach was developed as automation for entering data into the company's system to automate system. The validation result displays how well the system performed in terms of accuracy and processing time for the samples of invoice images. The system shows that accuracy is 100% with less than 30 seconds to complete the process. Hence, RPA with AI and OCR are suitable methods to be used as the invoice processing solution.
Dynamic fault coverages of March test algorithms need to be analyzed to ensure the robustness of Memory BIST for dynamic fault detection. Dynamic faults are proven to be a significant group of faults in Static Random Access Memory (SRAM). However, the manual analysis of dynamic fault coverage is time costly and tedious since it involves a total of 222 dynamic fault primitives. Conventional fault simulator depends on the size of the memory under test. Therefore, this proposed research focuses on the design of an automated fault analyzer for dynamic fault detection analysis by identifying the sensitizers and detectors, the sensitization cases for each test element for different March algorithms and lastly exporting the output of the computed fault coverage to a text file. The performance of the proposed analyzer is tested by using March MD2 and March MD9a and the fault coverages obtained from the analyzer matched with expected dynamic fault coverages and the execution time is around 6ms.
Auscultation that defined as listening to heart sound is the most practical and comfortable method to monitor patient's heart condition. Despite that, the complexity of murmur that exist in the heart sounds often provide misleading in diagnosis. This work proposed an auto-diagnosis system for heart disease by classifying the pathological heart sounds. Mitral valve stenosis and bicuspid aortic valve stenosis are two focused diseases due to the fatal effect. The analysis stage involved the transformation of heart sound signal into envelope signal in order to assist the location of the first and second heart sound. Moreover, power spectrum and Mel Frequency Cepstral Coefficient (MFCC) are extracted to classify the diseases where Fine Gaussian SVM is found to perform the best classifier model with 86.1% of accuracy. Later, the testing stage of the classification process has yielded the 86% of accuracy in detecting the true disease with 93% of average F1-score. Finally, a pre-develop prototype is constructed using GUI system that able to automatically diagnose the type of heart disease. The system, in future, has high potential to provide a better diagnosis system with less time-diagnose and operator-independent.
Due to the nature of soft robot manipulator, the modelling and control of its stiffness is crucial for a successful contact and non-contact tasks. The stiffness of soft robot manipulators generally embodies the manipulator's ability to satisfy the desired position and force commands. The goal of this study is to compare and investigate variable stiffness strategies in improving the distal tip stiffness of two-segment soft continuum manipulator which have in-built stiffness chamber constructed via granular jamming. The first step is to characterize the behaviour of the manipulator with the change in pressure to determine the Young's modulus. Then forces and their resultant length change relationship is established and based on the two characterization work, the stiffness matrix is obtained. The result shows that the stiffness matrix obtained is able to reject disturbances at the distal tip of the manipulator and the activation of the in- built stiffness chamber via granular jamming further enhance the stiffness and robustness of the manipulator, giving multiple strategies in improving tip distal stiffness.
Plug-in electric vehicles (PEVs) have been receiving a high demand among the public since they were introduced due to their benefits from incentives to reduce greenhouse gas emissions. With a number of PEVs on the road, they can be managed to provide ancillary services to power grid and deal with the intermittent resources of renewable energy. This requires an optimization technique to properly schedule the PEV operations that connect to power grid to give minimum power losses. Suitable substations for PEVs either to charge or discharge and appropriate number of their involvement are identified. Binary gravity search algorithm (BGSA) is one of the used optimization technique to solve the non-linear and non-convex scheduling problem. However, a suitable decision function needs to be investigated to give better performance. This work investigates three different decision functions, namely tangent, sigmoid and round functions. A modified 34-node distribution test system is used to showcase the performance in solving the PEVs scheduling problem. The results show that sigmoid function is more suitable for BGSA as compared to the conventional tangent function in solving the PEVs scheduling problem. Power loss reduction can be achieved up to almost 14% when compare to the base case. As a conclusion, a suitable decision function gives better performance in solving an optimization problem.
