Abstract: The distribution of process data from multimode processes has different features from that of single mode case. The monitoring index for multimode processes should not only reflect the statistical features of each mode itself, but also consider the information among different modes. Conventional monitoring indices, T2 and SPE, are unable to achieve the above goal in multimode processes. Aiming at the above problem, this paper proposes a novel monitoring index, which integrates information from all possible operating modes. Under the scheme of Bayesian inference, the data of multimode process is described by means of a summation of weighted local Mahlanobis distance for every mode. A simulation example on multimode continuous stirred tank heater is finally given.