Maintenance services logistics for wide geographically dispersed applications, such as oil transfer systems via pipelines or waste water treatment, have high costs and standard approaches usually lead to sub-optimal solutions. These systems are composed by a huge number of devices, often placed in inaccessible areas with a large distance between them. In such applications autonomous Intelligent Maintenance System (IMS) are capable to estimate their health conditions, can be used to forecast maintenance needs and to optimize maintenance schedule, therefore reducing the overall costs. Artificial Immune Systems (AIS) are a set of algorithms inspired by bio-immune systems that have features suitable for applications in IMS. AIS have distributed and parallel processing that could be useful to model large production systems. This chapter proposes an architecture for a Distributed IMS using Artificial Immune Systems concepts to face the challenges described and explore in-site learning. Each equipment has its own embedded AIS, performing a local diagnosis. If a new fault mode is detected, this information is evaluated and classified as a new non-self pattern, and included in the "vaccine". In this way, what is learned by one AIS can be propagated to the others. This proposal is modeled and implemented using multi-agent systems, where every autonomous IMS is mapped to a set of local agents, while the communication and decision process between IMSs are mapped to global agents. The chapter also describes the preliminary results deriving from the application of the proposed approach to a case study.