The detection of active ingredient contents is to evaluate the quality of medicinal plants, and different harvesting time has a significant influence on effective component contents of medicinal plants. In this study, Fourier transform infrared (FTIR) spectroscopy combined with chemometrics was used to establish a rapid method for predicting the total flavonoids content in dendrobium officinale at different harvesting time, in order to provide a basis for the rapid quality evaluation of this species. From January to December in 2014, a total of 111 samples were collected, dried and crushed. In addition, the content of total flavonoids in D. officinale was determined by using traditional method and the accumulation regularity of total flavonoids in D. officinale was studied. Moreover, the infrared spectra of the samples were recorded and their bands were identified. The original spectra were pre-processed by combining the first derivative, second derivative, standard normal variable, multiple scattering correction and orthogonal signal correction. Processed FTIR spectra were set as variable X and the contents of total flavonoids were set as variable Y. The PLS model was established to predict the total flavonoids content. The results showed that: (1) Both the sample and the standard rutin had a common absorption peak near 270 nm. In this experiment, the total flavonoid was measured at 270 nm. The standard curve equation of total flavonoids was y=6.076 5x+0.055 8, r=0.996 6 and the precision, reproducibility and stability of relative standard deviation (RSD) were 0.37%, 1.00% and 0.28%, respectively. (2) The content of total flavonoids increased with time at first and then decreased, and the content of total flavonoids was higher than 64.10 mg·g-1 from June to August. (3) The FTIR data was integrated with the total flavonoid content of D. officinale by partial least squares regression analysis (PLSR). The results showed that the optimal pretreatment method was 2D+SG5+SNV+OSC-PLSR, the r and RMSEE of the training set were 0.979 0 and 2.438 2, respectively while R2 and RMSEP of validation set were 0.882 4 and 4.169 9, respectively. The predicted and the measured value of total flavonoid content was close to each other, indicating that PLS model can be used for rapid prediction of total flavonoid content. FTIR combined with chemometrics can accurately predict the total flavonoids content in D. officinale and provide a rapid and effective method for the quality evaluation of D. officinale.