Reliability Assessment of Automation Connectors Used in Heavy Machinery with Tailored Testing

Autonomous machinery is under intensive development. However, this trend is hindered by the insufficient reliability of components. CAN-bus is a default control solution for heavy machinery. Connector components are necessary part of the network. The connectors are IP classified, and the IP code is often used as an unofficial indicator of field worthiness. In heavy machinery where stress levels are high, this assumption does not always apply. Therefore, more specific information on reliability is needed in the selection of components.

This thesis presents tailored tests that simulate actual stress combinations in underground mining conditions and degradation models for the estimation and prediction of degradation process. The tests were developed to reveal the strengths and weaknesses of the connector designs and enable reliability comparisons under different operating conditions.

The developed tests revealed significant reliability differences between connector designs. The ranking of the component designs varied between the tests and applied stress factor combination. This emphasized the importance of finding a match between stress factors in operating conditions and the reliability features of the connector. This does not apply only to the robustness of design from the perspective of technical properties, but also to how prone components are to human errors. The result is that tailored testing in accordance with an actual stress environment is a necessity. IP code is not sufficient indicator of reliability in demanding conditions. The effect of human factors must also be included. Reliability assessment should be considered as a whole, including technical properties, but also ease of maintenance, the availability of authorized service, and the quality of workmanship at the end-user’s location.

Degradation is a random process, leading to a high variance in the lifetime distribution of the sample population. Thus, the lifetime of a single component cannot be precisely predicted. However, the models developed in this work are able to capture the statistical behavior of the process. The result is that a stochastic model, comprising deterministic and random components is suitable for degradation modeling.