Cefazolin s odium pe ntahydrate po wder f or i njection i s the major pr oduct of S henzhen G osun C hina Resources Pharmaceutical C o.,Ltd.One of manufacturing s teps of t his powder injection is drying process.How quickly and effectively determine the endpoint of the drying process for cefazolin sodium pentahydrate powder is an age-old difficulty for our manufacturers.The time must meet two criteria to be identified as the endpoint of the drying process.First,the contents of the residual organic solvents such as acetone and isopropyl alcohol should comply with their corresponding standards(not more than 0.5%).At the same time,the water content must be enough to maintain the crystal structure of the cefazolin sodium pentahydrate(in the range of l3%-l6%).The traditional method such as gas chromatography(GC)and F ischers method usually r equires substantial time a nd c omplicated s ample preprocessing,which will miss the best time for stopping the drying process.This study attempts to construct some quantitative models of organic solvent and water using near infrared(NIR)spectroscopy technology for on-line drying process endpoint detection.More than 200 samples at different time points of drying process(including 9l final products)from 9l batches of the cefazolin sodium pentahydrate(as shown in table l)were collected and scanned using near infrared instrument with integrating sphere mode(MPA,Bruker Optik GmbH,Ettlingen,Germany).And the residual values of water,acetone and isopropyl alcohol in each sample were determined by Fischers method and GC respectively.Then these NIR quantitative models for water,acetone and i sopropyl alcohol were constructed using p artial l east squares(PLS)r espectively.Cross-validation r oot mean square er rors(RMSECV)were u sed t o o ptimize s pectral pretreatment methods a nd s pectral r egions.T he f inal parameters for the three models were shown in Table l.Samples which were not used for calibration sets,were used to validate these three models.As shown in Table 2,all the results indicated that these three models have good predictive ability.Our study shows that compared with traditional methods,NIR technology is a rapid and fast method.If this NIR method can be used in routine drying process,it will improve process analysis,supervision and diagnosis.