Performance Monitoring and Operator Assistance Systems in Mobile Machines

Abstract

Studies related to the productivity of machine work often show that there is a large amount of unexplained variation in the productivity, which counts in most of the performance differences between the machines. This unexplained variation includes the skill and actions of the machine operators. It is clear that in machine work, the effect of the human operator is very significant to the overall performance. These differences between the operators possess a substantial potential for improvement. If novice machine operators would be provided with individual feedback and training based on their actual work performance and areas of improvement, right from the beginning, the chances of reaching good skills could be significantly increased. However, any practical and wide-spread means to provide objective and useful real-time feedback to the operators have not been available before. Measuring and accurate modeling of a man-machine system that operates in varying conditions is very challenging, because the skills and work procedures of the human operators controlling the process are always individual. This thesis presents a method of the recognition of machine work tasks and work cycles based on the combination of multivariate control signals generated by the operator. The recognition of work cycles and tasks is based on Hidden Markov Models (HMM). As the actions of the operator become recognizable, the operator’s effect to the overall performance of machine work does not need to be regarded as an unknown disturbance. It also facilitates the evaluation of operators’ skill at the task level and the analysis of work technique. The thesis also presents a method of using intelligent coaching systems (ICS) for example to provide useful feedback to operator training or to support the operators in decision making. The ICS is based on qualitative expert knowledge related to the man-machine work process. It uses skill and performance measures, which are defined for each work task. The values of the performance measures are evaluated using corresponding statistical reference. The ICS makes observations and gives suitable feedback to the operator in the form of linguistic suggestions. The expert knowledge is formulated as rules of a fuzzy inference system. Significant performance and productivity improvement in man-machine systems could be gained by enhancing the abilities of the machine operators to perform the work tasks more successfully. Moreover, the methods presented in this thesis are based on the measurements and performance measures that are already available from the process. Thus, implementation of the methods does not increase the manufacturing cost and complexity of the system, since it is not necessary to mount additional measuring equipment. The presented methods for work task and work cycle recognition and skill evaluation of machine operator at work task level have been implemented in industrial applications.