Machine condition monitoring is important for ensuring the efficiency of machines in a factory and the safety of workers. A variety of vibration-based analysis techniques have been long used to diagnose the health status of machines. Since the emergence of the Internet, remote machine monitoring systems operating through the Internet have attracted much scholarly attention. Although mobile communications are convenient for everyday uses, and can also be used in some industrial applications, researchers have not yet seriously considered remote machine condition monitoring systems based on the mobile communication. In this dissertation, some novel research on remote machine condition monitoring and fault diagnosis systems which operate through the Internet and mobile communication are presented. This research is based on emerging computer technologies, such as JSP (JavaServer Pages), XML (Extensible Markup Language) and J2ME (Java 2 Platform, Micro Edition). Three novel architectures were designed and their prototypes were implemented so that the condition monitoring of machine status became achievable on the Internet as well as on mobile phones/PDAs. In order to allow maintenance staff to obtain the machine health status generated by an existing on-site machine monitoring system remotely, especially through their mobile phones, the first architecture was presented and a prototype was built based on SAMS (Smart Asset Maintenance System). SAMS is an on-site machine maintenance system developed by the Smart Engineering Asset Management Laboratory of the City University of Hong Kong. This prototype allows users to check the health status (presented in the form of data, image and video) of operating machines through a mobile phone or a computer. In addition, when the machine status is abnormal, an automatic alarm trigger module can actively send alert messages to designated users' mobile phones and call these phones to make sure that the concerned maintenance staff be aware to read the alerting messages. The language XML is used to encode machine data and diagnostic results. Then this prototype can provide different machine health information according to the requests of the users. The second architecture is based on dynamic Web technologies. This architecture allows authorized users to remotely control the parameters for sensory data acquisition and analysis programs running in a server located at a factory. Therefore, the remotely located user has the freedom to select the preferred parameters for the data acquisition as well as the algorithms for fault diagnosis. To extend the capabilities of this architecture so that it is applicable to mobile communications, the client software running in a mobile phone/PDA was built. A remotely located user can then inspect the health status of an inspected machine through a computer connected by board band to the Internet, or a mobile phone/PDA. Based on this architecture, a prototype was built. A LabVIEW program running in the server is used to analyze the collected machine operating data. Through a computer or a mobile terminal, a remotely located user can set up his desired parameters for sensory data acquisition/analysis programs and obtain results generated according to his preferred algorithms and parameters. In order to increase the functionality of a Smartphone/PDA for remote machine condition monitoring, instead of requesting the server to run the selected algorithm for fault diagnosis, the third architecture was designed within which the diagnostic algorithms can be executed in a Smartphone/PDA. Therefore, the remote user can perform his desired fault diagnosis algorithm any time on his Smartphone/PDA. A prototype was designed to host Fast Fourier Transform (FFT) on the Smartphone/PDA. This FFT algorithm coded by J2ME is used to analyze the collected machine data and then generate the diagnostic results directly on the user's Smartphone/PDA. The ability to perform machine fault diagnosis on a Smartphone/PDA is novel and has tremendous potential for use in future applications as new hardware and software technologies are developed for Smartphones/PDAs. Besides, in order to reduce the number of data files required to be transferred to the mobile terminals in the third architecture, an AI based method is proposed to help select a monitoring sensory point. Through this method, an expert, who views the FFT spectrum through his Smartphones/PDAs and then assesses the status of the machine, can get a machine data file from different monitoring sensory points. Compared with other data files, this data file has the worst similarity with the data files collected when machine is running normally. The new method is implemented based on the concept of k-Nearest Neighbor (kNN). It only requires prior knowledge on the tested machine when it is running at normal condition. First, the temporal signals collected from the machine at different time in normal running conditions are converted to their respective frequency spectra. Second, the distance between two FFT based spectra is defined to describe the similarity of two machine data files. Third, the similarities of given new samples to data files collected from a machine running at normal condition are used to select a data file which will be transmitted to the designated Smartphones/PDAs. The simulation has shown the effectiveness of this method. In summary, the first architecture is based on an existing on-site machine maintenance system and is an economic means for obtaining a machine's health status information through the Internet and mobile communication. The second architecture allows users to control server-side machine monitoring software, and actively analyze a machine's status through a remotely located computer or mobile phone/PDA. The third architecture allows an authorized user to perform machine fault diagnosis directly on his Smartphone or PDA.