With　the　extensive　use　of　the　left　ventricular　assist　device　(LVAD)　as　a　treatment　of　heart　failure,　suction　detection　has　become　a　key　issue　that　directly　affects　the　treatment.　To　detect　the　phenomenon　of　suction,　a　blood　assistant　index　(BAI)　is　defined,　which　reflects　the　unloading　level　of　the　pump.　The　BAI　is　a　ratio　of　the　external　work　of　LVAD　and　the　input　power　of　cardiovascular　system.　Using　the　theory　of　model-free　adaptive　control　algorithm,　an　anti-suction　controller,　which　chooses　the　heart　rate　and　BAI　as　control　variables,　is　designed.　As　a　key　feature,　the　proposed　control　algorithm　adjusts　the　pump　speed　according　to　not　only　the　blood　demand　of　circulatory　system　but　also　the　function　of　the　native　heart.　Subsequently,　the　performance　and　robustness　of　the　controller　are　evaluated　using　a　numerical　simulation　of　the　assisted　circulation　and　an　in　vitro　experiment.　The　simulation　and　experimental　results　demonstrate　that　the　BAI　detects　the　suction　occur　accurately,　and　the　controller　can　maintain　the　heart　rate　and　BAI　tracking　the　reference　values　with　a　response　time　of　less　than　6　s.