In an attempt to tackle associated with the problems current data-driven degradation predictions for rotating machinery-such as insufficient consideration of time series information, unreasonable life labeling, and large cumulative error of degradation models-a method involving fusion trend filtering, fuzzy information granulation, and a dynamic long-short-term memory network (LSTM) has been proposed for predicting the degradation trends and degradation intervals of rotating machinery. Taking the vibration signal as an example, the characteristic index of the degradation information of the equipment is first extracted, and then the main degradation trend and the fuzzy degradation boundaries are extracted through trend filtering and fuzzy information granulation, and finally the comprehensive performance degradation is predicted using dynamic LSTM. The feasibility and effectiveness of the method were verified using a bearing training data set published on the internet.

In view of the difficulty in diagnosing rolling bearing faults in a noisy environment, a new method for bearing fault identification based on an anti-noise multi-core convolutional neural network (AMCNN) is proposed. First, the rolling bearing vibration signal is preprocessed to obtain data samples, which are divided into a training set and a test set. Then, the bearing life state recognition model is established, and the tagged training set data samples are input into the AMCNN for training. Finally, the trained AMCNN model is applied to the test set to output the fault identification result. In the training process, in order to suppress over-fitting, the original training samples are subjected to noise-adding processing. In order to improve the anti-jamming capability of the model, the dropout layer is used as the first layer of the AMCNN. At last, the bearing test data is used to test the identification model. Comparison with conventional methods shows that our method can more accurately idenfity a bearing fault in a high noisy environment.

Pipeline vibration often occurs during equipment operation, and can result in serious safety risks. In an attempt to solve such problems, a control force has been applied to a vibrating pipe system by an inertia actuator, so that active control of the pipeline vibration is achieved by means of an active damping device. Taking the 25 Hz vibration of the reciprocating compressor outlet pipeline in a chemical plant as an example, a vibration pipe test bench was built for experimental studies, and the vibration reduction afforded by the active damping device was simulated by SAP2000. The effect of the active damping device on the pipeline vibration when the inertial actuator was installed in different positions was studied experimentally. By comparing the vibration reduction when the inertial actuator was mounted at four different positions, it was found that the active damping device had the best vibration control effect on the whole pipeline when the actuator was installed in the position with the largest vibration.

There are a large number of T-junctions in petroleum, chemical and nuclear power pipeline systems. The turbulent penetration phenomena caused by the difference in the flow and temperature between the main and branch pipe of the T-junctions give rise to temperature and velocity fluctuations. These may cause thermal fatigue problems. In this work the large-eddy simulation (LES) method was used to calculate the temporal and spatial evolution of temperature and velocity in the flow field. The results show that the fluctuation is strongest at the dimensionless axial height of the H=4 surface. A transient thermo-fluid-solid coupling simulation process was then established, and the surface pressure and body temperature were dynamically loaded into the solid domain to obtain the stress distribution of the pipeline and find the location of the danger point. This was found to be the axial height H=0 of the inner wall surface. Finally, the rainflow cycle counting algorithm was used to analyze the stress fluctuation information for the danger point, the equivalent symmetrical cyclic stress was obtained according to the Goodman curve, and the fatigue life of the danger point was evaluated according to the linear fatigue damage accumulation criterion. This allows the fatigue life of the pipeline to be determined.

The angle head is an essential processing accessory for computer numerical control (CNC) machine tools. It is extremely vulnerable to damage under long-term harsh processing conditions. The strong random noise in the environment will annihilate the fault feature information of the angle head, which makes it difficult to extract data about fault features. To solve this problem, a dual noise reduction method based on ensemble empirical mode decomposition (EEMD) and autocorrelation is proposed. An autocorrelation filtering approach is used to preprocess the vibration signals data, and then the obtained signals are decomposed using EEMD. A genetic algorithm is then applied to optimize the input parameters of EEMD, and the intrinsic mode function (IMF) component obtained from the EEMD decomposition is selected to reconstruct the signal on the basis of a combination of kurtosis and correlation coefficients. The data for the angle head fault features can then be extracted from the reconstructed signals through time-frequency analysis. The predictions obtained using our method show good agreement with the measured data for the angle head. The results show that the proposed method can effectively suppress random noise and can accurately extract fault feature information for the angle head.

Due to the non-stationary characteristics of the vibration signals of complex machinery such as reciprocating compressors and gas turbines, the signal feature threshold alarm method has low alarm accuracy. In order to solve this problem, a fault early warning method based on the infinite student's t-mixture model (iSMM) is proposed. The proposed method first uses iSMM trained by a high-dimensional feature space which is based on the mechanical vibration signal features to describe the change in performance of the equipment. Secondly, the divergence between the normal condition model and the real-time working model is calculated by matching based on approximating KL divergence. Finally, real-time fault early warning for the mechanical equipment is realized by comparing the divergence with the early warning threshold calculated based on the 3σ rule of thumb. The proposed method has been validated by using actual failure case data for a reciprocating compressor. The results show that this method has high alarm accuracy and good timeliness. The proposed method can provide effective mechanical fault early warning.