During　the　TBM　rock　breaking　process,　it　is　inevitable　to　generate　strong　vibration　on　the　cutterhead.　The　vibration　response　of　the　cutterhead　is　the　result　of　the　interaction　between　rock　and　TBM,　and　it　can　serve　as　an　important　basis　for　rock　mass　identification　and　tunnelling　parameter　optimization.　It　is　particularly　important　to　obtain　the　information　contained　in　this　multi-source　vibration　signal　of　the　cutterhead.　Based　on　the　TBM　tunnelling　test　data,　the　influence　of　time　on　the　distribution　of　vibration　features　is　analyzed,　a　random　forest　model　of　multi-dimensional　vibration　features　and　tunnelling　parameters　is　constructed,　and　the　vibration　features　that　can　respond　sensitively　to　changes　in　tunnelling　parameters　are　screened　through　its　feature　importance　evaluation　function.　The　research　results　indicate　that　on-site　vibration　monitoring　duration　should　reach　at　least　one　rotation　period　of　the　cutterhead,　during　which　the　distribution　of　various　vibration　characteristics　tends　to　be　stable.　Peak　factor　and　frequency　standard　deviation　are　the　features　that　can　best　respond　to　changes　in　thrust　and　rotation　speed,　and　can　be　used　as　key　features　for　studying　the　relationship　between　vibration　signal　and　change　in　tunnelling　parameters.　©　2023　Editorial　By　Modern　Tunnelling　Technology.　All　rights　reserved.