Oil monitoring has been considered as an essential tool in determining the operational state of an equipment. In other words, the data obtained from monitoring the oil represents the equipment's operational status. So, identifying the abnormal working status of the lubricating oil allows to understand its operating status in advance. Accordingly, the present study aims at proposing an anomaly identification model based on Long Short Term Memory (LSTM) and Support Vector Data Description (SVDD) for time series data of wear state collected by the oil online monitoring system. So, to predict the wear condition of the oil in the future, LSTM and grid search are utilized and a trend prediction model of oil monitoring metrics is constructed. Given that the fault is a gradual process, a multi-dimensional hypersphere classifier is generated by SVDD to train the healthy samples. The classifier is utilized to diagnose the lubricating oil condition data obtained by the LSTM prediction model as well as to realize the abnormal identification of the equipment's lubricating oil. The obtained experimental results corroborated the higher prediction accuracy and the lower prediction loss of the trend prediction model. Therefore, the future anomalies of the device can be argued to be reliably predicted in accordance with its actual performance using the LSTM-SVDD method. Furthermore, with regards to the actual operating conditions of the diesel engine, the alarm rate as well as the false alarm rate, the LSTM-SVDD method was demonstrated to be superior to the LSTM-PCA-SPE method. Hence, the novel model proposed in this study can be used to identify the status parameters of the lubricating oil abnormal wear. Moreover, the model can be argued to improve the accuracy of the abnormal identification as well as the online identification of the lubricating oil.