Model Fusion and Multiscale Feature Learning for Fault Diagnosis of Industrial Processes.

The data generated by modern industrial processes often exhibit high-dimensional, nonlinear, timing, and multiscale characteristics. Presently, most of the fault diagnosis methods based on deep learning only consider the part of the characteristics of industrial data, which will cause the loss of part of the feature information during training, thereby affecting the final diagnosis effect. In order to solve the above problems, this article proposes an end-to-end multiscale feature learning method based on model fusion, which can simultaneously extract multiscale spatial features and temporal features of data, effectively reducing the loss of feature information. First, this article combines the convolutional neural network (CNN) with residual learning and designs a multiscale residual network (MRCNN) to extract high-dimensional nonlinear spatial features of different scales in the data. Then, the extracted features are input into the long and short-term memory (LSTM) network to further extract the temporal features of the data. After the fully connected layer, it is input into the classifier for final fault classification. The residual learning in MRCNN can effectively avoid the problem of model degradation and improve the training efficiency of the model. Through the fusion of MRCNN and LSTM, we can significantly improve the feature extraction ability of the model, thereby greatly improving the diagnosis effect. In the final case experiment, the method improved the comprehensive diagnostic accuracy of the Tennessee-Eastman (TE) process and industrial coking furnace datasets to 94.43% and 97.80%, respectively, which was significantly better than the existing deep learning model and proves the effectiveness and superiority of this method.

Glycyrrhizic acid promotes sciatic nerves recovery in type 1 diabetic rats and protects Schwann cells from high glucose-induced cytotoxicity.

The present study aims to investigate the therapeutic effect and mechanism of glycyrrhizic acid (GA) in diabetic peripheral neuropathy (DPN). GA significantly mitigated nerve conduction velocity (NCV) deficit and morphological abnormality and reduced high-mobility group box-1 (HMGB1) expression in the sciatic nerves of diabetic rats independent of blood glucose and body weight. Notably, GA alleviated the increase of HMGB1 and the decrease of cell viability in high glucose-stimulated RSC96 cells. Furthermore, GA obviously reduced the concentration of inflammatory cytokines in the sciatic nerves of diabetic rats and supernatants of high glucose-exposed RSC96 cells, then restored the decreased expression levels of nerve growth factor (NGF) and neuritin-1, and the increased expression levels of cleaved caspase-3 and neuron-specific enolase. Additionally, GA markedly inhibited receptor for advanced glycation end products (RAGE) expression, p38MAPK phosphorylation, and the nuclear translocation of NF-κBp65 in diabetic rats and high glucose-exposed RSC96 cells. The promotional effect of high glucose in RSC96 cells was diminished following Hmgb1 siRNA treatment. Our findings indicate that GA may exert neuroprotection on DPN by suppressing HMGB1, which lead to extenuation of inflammation response, balance of NGF, neuritin-1 and caspase-3, as well as inactivation of RAGE/p38MAPK/NF-κBp65 signaling pathway.

Sirt1 Mediates Vitamin D Deficiency-Driven Gluconeogenesis in the Liver via mTorc2/Akt Signaling.

As an active form of vitamin D (VD), 1,25-dihydroxyvitamin D (1,25(OH)2D3) is involved in the development of many metabolic diseases, such as diabetes, autoimmune diseases, and tumours. While prospective epidemiological studies have consistently implicated VD deficiency in the regulation of glucose metabolism and insulin sensitivity, the specific mechanism remains unclear. Here, we generated 1α(OH)ase-null mice (targeted ablation of the 25-hydroxyvitamin D 1α hydroxylase enzyme) and found that these mice developed hepatic glucose overproduction, glucose intolerance, and hepatic insulin resistance accompanied by reduced Sirtuin 1 (Sirt1) expression. The chromatin immunoprecipitation (ChIP) and a luciferase reporter assay revealed that 1,25(OH)2D3-activated VD receptor (VDR) directly interacted with one VD response element (VDRE) in the Sirt1 promoter to upregulate Sirt1 transcription, triggering a cascade of serine/threonine kinase (AKT) phosphorylation at S473 and FOXO1 phosphorylation at S256. This phosphorylation cascade reduced the expression of gluconeogenic genes, eventually attenuating glucose overproduction in the liver. In addition, a signaling pathway was found to modulate gluconeogenesis involving VDR, Sirt1, Rictor (a component of mTOR complex 2 [mTorc2]), AKT, and FOXO1, and Sirt1 and FOXO1 were identified as key modulators of dysregulated gluconeogenesis due to VD deficiency.

Intelligent Fault Diagnosis for Chemical Processes Using Deep Learning Multimodel Fusion.

Deep learning technology has been widely used in fault diagnosis for chemical processes. However, most deep learning technologies currently adopted only use a single network stack or a certain network stack with multilayer perceptron (MLP) behind it. Compared with traditional fault diagnosis technologies, this method has made progress in both the diagnosis accuracy and speed, but due to the limited performance of a single network, the accuracy or speed cannot meet the requirements to the greatest extent. In order to overcome such problems, this article proposes a fault diagnosis method using deep learning multimodel fusion. Different from previous deep learning diagnosis methods, this method uses long short-term memory (LSTM) and convolutional neural network (CNN) to extract features separately. The extracted features are then fused and MLP is taken as the input for further feature compression and extraction, and finally the diagnosis results will be obtained. LSTM has long-term memory capabilities, the extracted features have temporal characteristics, and CNNs have a good effect on the extraction of spatial features. The proposed method integrates these two aspects for diagnosis such that the features finally extracted by the network have both spatial and temporal characteristics, thereby improving the network's diagnostic performance. Finally, a TE chemical process and an industrial coking furnace process are taken for simulation testing. It is proved that the performance of this method is superior to existing deep learning fault diagnosis methods with simple sequential stacking for unilateral feature extraction.

A 2D-FM model-based robust iterative learning model predictive control for batch processes.

The work deals with composite iterative learning model predictive control (CILMPC) for uncertain batch processes via a two dimensional Fornasini-Marchesini (2D-FM) model. A novel equivalent error system is first presented which is composed of state error and tracking error. Then an iterative learning predictive updating law is constructed by 2D state feedback control and the 'worst' case linear quadratic function is also designed. Besides, the update controller considering the input and output constraint will be optimized using the worst-case objective function along the infinite moving horizon. The solvable conditions that can be optimized online in real time are constructed using linear matrix inequalities (LMIs). The stability of the proposed control scheme can be achieved with the feasibility of the optimization problem. Compared with robust traditional MPC using one-dimensional models, the presented control approach can guarantee more degrees of tuning to achieve faster convergence of tracking error, which is of more significance since uncertainties exist inevitably in industrial batch processes. Finally, an injection molding process and a three-tank are introduced as two cases to demonstrate the feasibility and superiority of the proposed MPC strategy.

Serum myeloid-related protein 8/14 is a potential predictor of diabetic kidney disease.

BACKGROUND: Myeloid-related protein 8/14 (MRP8/14) is secreted by macrophages and formed by MRP8 and MRP14, which is closely related to vascular inflammation. Chronic vascular inflammation plays a significant role in the development and progression of diabetic kidney disease (DKD). This study aims to investigate the relationship between MRP8/14 and DKD. METHODS: A total of 80 individuals with type 2 diabetes were divided into four groups, according to the baseline urinary albumin/creatinine ratio (ACR) levels Serum concentrations of MRP8/14 were measured by ELISA. The clinical variables were obtained through physical examination, illness history, or laboratory evidence. RESULTS: As DKD worsened, the level of serum MRP8/14 increased gradually, and MRP8/14 has a significantly positive correlation with ACR (r = 0.349, P = 0.002), body mass index (BMI) (r = 0.288, P = 0.009), serum creatinine (Cre) (r = 0.392, P < 0.001), blood urine nitrogen (BUN) (r = 0.333, P = 0.003), systolic blood pressure (SBP) (r = 0.301, P = 0.007), and a negative correlation with the estimated glomerular filtration rate (eGFR) (r = -0.478, P < 0.001). Logistic regression analysis showed that age, Cre, eGFR, ACR, and MRP8/14 were associated with the progression of DKD (P < 0.05). CONCLUSIONS: The serum MRP8/14 is correlated significantly with the progression of DKD, suggesting that MRP8/14 may be an independent predictor of the progression of DKD.

Keap1-Nrf2 signaling activation by Bardoxolone-methyl ameliorates high glucose-induced oxidative injury in human umbilical vein endothelial cells.

In cultured human umbilical vein endothelial cells (HUVECs) high glucose (HG) stimulation will lead to significant cell death. Bardoxolone-methyl (BARD) is a NF-E2 p45-related factor 2 (Nrf2) agonist. In this study we show that BARD, at only nM concentrations, activated Nrf2 signaling in HUVECs. BARD induced Keap1-Nrf2 disassociation, Nrf2 protein stabilization and nuclear translocation, increasing expression of antioxidant response element (ARE) genes. BARD pretreatment in HUVECs inhibited HG-induced reactive oxygen species production, oxidative injury and cell apoptosis. Nrf2 shRNA or knockout (using a CRISPR/Cas9 construct) reversed BARD-induced cytoprotection in HG-stimulated HUVECs. Conversely, forced activation of Nrf2 cascade by Keap1 shRNA mimicked BARD's activity and protected HUVECs from HG. Importantly, BARD failed to offer further cytoprotection against HG in the Keap1-silened HUVECs. Taken together, Keap1-Nrf2 cascade activation by BARD protects HUVECs from HG-induced oxidative injury.

Suppression of TLR4 by miR-448 is involved in Diabetic development via regulating Macrophage polarization.

OBJECTIVES: Lipopolysaccharide (LPS) contributed to the development and progression of type 2 diabetes mellitus (T2D), while TLR4 is reported to mediate the LPS-induced inflammation in macrophages. However, the potential molecular mechanisms for TLR4-mediated macrophages activation in T2D have not yet to be fully clarified. METHODS: Type 2 diabetes models in C57BL/6J mice were generated by a combination administration of streptozotocin (STZ) and a high-fat diet (HFD). Cell proportions of M1 and M2 macrophages were analyzed using flow cytometry. Expression profiles of miR-448 and TLR 4 were determined by qRT-PCR and Western blot. KEY FINDINGS: LPS/IFN-γ significantly induced M1 polarization in macrophages characterized by the increased levels of TNF-α, IL-6, IL-12, iNOS and decreased levels of TNF-ß, CCL-22, IL-10 and Arg-1, with a higher expression of toll-like receptor 4 (TLR4) in vitro. Consistently, T2D mice-derived macrophages had a significantly elevated expression of TLR4 mRNA and decreased expression of miR-448. We further confirmed that miR-448 could inhibit TLR4 expression by targeting the 3'-UTR of TLR4, rescuing the LPS/IFN-γ-induced M1 macrophage polarization. CONCLUSIONS: Taken together, our results indicated that decreased miR-448 in diabetic macrophages may contribute to LPS-induced M1 polarization by targeting TLR4, thereby modulating T2D development.

GLP­1R agonists ameliorate peripheral nerve dysfunction and inflammation via p38 MAPK/NF­κB signaling pathways in streptozotocin­induced diabetic rats.

The present study aimed to investigate the mechanism of glucagon­like peptide­1 receptor (GLP­1R) agonists in the progression of diabetic peripheral neuropathy (DPN) in streptozotocin (STZ)­induced diabetic rats, through inflammatory signaling pathways. The DPN rat model was generated by intraperitoneal injection of STZ and then treated with the GLP­1R agonist liraglutide or saline for 8 weeks. These animals were randomly divided into 4 groups (10 rats in each): The normal control + saline group, the normal control + liraglutide group, the diabetic + saline (DM) group and the diabetic + liraglutide (DML) group. The nerve conduction velocity (NCV) in the sciatic nerves of the rats was monitored over a period of 8 weeks. Peripheral serum was obtained for the measurement of blood glucose, tumor necrosis factor­α (TNF­α), interleukin­6 (IL­6) and IL­1ß level. The protein levels of phosphorylated (p­) and total extracellular signal­regulated kinase, c­Jun NH2­terminal kinases, p38 mitogen­activated protein kinases (MAPK), and nuclear and cytoplasmic nuclear factor­κB (NF­κB) were measured through western blot analysis. Sciatic nerve mRNA expression levels of proinflammatory chemokines (TNF­α, IL­6 and IL­1ß), chemokines [monocyte chemoattractant protein­1 (MCP­1)], adhesion molecules [intercellular adhesion molecule 1 (ICAM­1)], neurotrophic factors [neuritin, nerve growth factor (NGF) and neuron­specific enolase (NSE)] and NADPH oxidase 4 (NOX4) were evaluated by reverse transcription-quantitative polymerase chain reaction. Subsequent to 8 weeks of treatment with liraglutide, the density of myelin nerve fibers was partially restored in the DML group. The delayed motor NCV and sensory NCV in the DML group were improved. The IOD value of NOX4 staining in the DML group (24.43±9.01) was reduced compared with that in the DM group (56.60±6.91). The levels of TNF­α, IL­1ß and IL­6 in the peripheral serum of the DML group were significantly suppressed compared with those of the DM group. It was also observed that the mRNA expression levels of TNF­α, IL­6, IL­1ß, MCP­1, ICAM­1 and NOX4 in the sciatic nerve were attenuated in the DML group. The mRNA expression of neuritin and NGF was significantly increased in the DML group compared with that of the DM group; NSE was reduced in the sciatic nerves of the DML group compared with that of the DM group. Additionally, the protein expression of p­p38 MAPK and NF­κB in the DML group was significantly suppressed. These data demonstrated that GLP­1R agonists may prevent nerve dysfunction in the sciatic nerves of diabetic rats via p38 MAPK/NF­κB signaling pathways independent of glycemic control. GLP­1R agonists may be a useful therapeutic strategy for slowing the progression of DPN.

Decoupled ARX and RBF Neural Network Modeling Using PCA and GA Optimization for Nonlinear Distributed Parameter Systems.

Modeling of distributed parameter systems is difficult because of their nonlinearity and infinite-dimensional characteristics. Based on principal component analysis (PCA), a hybrid modeling strategy that consists of a decoupled linear autoregressive exogenous (ARX) model and a nonlinear radial basis function (RBF) neural network model are proposed. The spatial-temporal output is first divided into a few dominant spatial basis functions and finite-dimensional temporal series by PCA. Then, a decoupled ARX model is designed to model the linear dynamics of the dominant modes of the time series. The nonlinear residual part is subsequently parameterized by RBFs, where genetic algorithm is utilized to optimize their hidden layer structure and the parameters. Finally, the nonlinear spatial-temporal dynamic system is obtained after the time/space reconstruction. Simulation results of a catalytic rod and a heat conduction equation demonstrate the effectiveness of the proposed strategy compared to several other methods.

Improved state space model predictive fault-tolerant control for injection molding batch processes with partial actuator faults using GA optimization.

A novel model predictive fault-tolerant control (MPFTC) strategy adopting genetic algorithm (GA) is proposed for batch processes under the case of disturbances and partial actuator faults. Based on the extended state space model in which the tracking error is contained, there are more degrees of freedom provided for the controller design and better control performance is obtained. In order to enhance the control performance further, the GA is introduced to optimize the relevant weighting matrices in the cost function. The effectiveness of the proposed MPFTC approach is tested on the injection velocity regulation of the injection molding process.

Improved fuzzy PID controller design using predictive functional control structure.

In conventional PID scheme, the ensemble control performance may be unsatisfactory due to limited degrees of freedom under various kinds of uncertainty. To overcome this disadvantage, a novel PID control method that inherits the advantages of fuzzy PID control and the predictive functional control (PFC) is presented and further verified on the temperature model of a coke furnace. Based on the framework of PFC, the prediction of the future process behavior is first obtained using the current process input signal. Then, the fuzzy PID control based on the multi-step prediction is introduced to acquire the optimal control law. Finally, the case study on a temperature model of a coke furnace shows the effectiveness of the fuzzy PID control scheme when compared with conventional PID control and fuzzy self-adaptive PID control.

Improved design of constrained model predictive tracking control for batch processes against unknown uncertainties.

In this paper, an improved constrained tracking control design is proposed for batch processes under uncertainties. A new process model that facilitates process state and tracking error augmentation with further additional tuning is first proposed. Then a subsequent controller design is formulated using robust stable constrained MPC optimization. Unlike conventional robust model predictive control (MPC), the proposed method enables the controller design to bear more degrees of tuning so that improved tracking control can be acquired, which is very important since uncertainties exist inevitably in practice and cause model/plant mismatches. An injection molding process is introduced to illustrate the effectiveness of the proposed MPC approach in comparison with conventional robust MPC.

Application of multi-model switching predictive functional control on the temperature system of an electric heating furnace.

A method of multi-model switching based predictive functional control is proposed and applied to the temperature control system of an electric heating furnace. The control strategies provide the effective and independent control modes of the electric heating furnace temperature in order to obtain improved control performance. The method depends on conventional implementation of the multi-model switching state, which requires some endeavors to tune the switching model in the model predictive control and allows a reduction of the calculation compared with the weighted multiple model algorithms. In order to test the advantage of the proposed method, experimental equipment is set up and experiments are done on the temperature process of a heating furnace, which verify the validity and effectiveness of the proposed algorithm.

Multivariable design of improved linear quadratic regulation control for MIMO industrial processes.

In this study, a multivariable linear quadratic control system using a new state space structure was developed for the chamber pressure in the industrial coke furnace. Such processes typically have complex and nonlinear dynamic behavior, which causes the performance of controllers using conventional design and tuning to be poor or to require significant effort in practice. The process model is first treated into a new state space form and the implementation of linear quadratic control is designed using this new model structure. Performance in terms of regulatory/servo, disturbance rejection and measurement noise problems were all compared with the recent model predictive control strategy. Results revealed that the control system showed more robustness and improved the closed-loop process performance under model/process mismatches.

Enhancing in vitro bioactivity and in vivo osteogenesis of organic-inorganic nanofibrous biocomposites with novel bioceramics.

Fabricating bioactive nanofibrous scaffolds from biodegradable polymers to mimic native tissue is an important approach in repairing bony defects. Silk fibroin (SF) may contribute to bone regeneration because of its excellent mechanical properties, slow degradability, and low osteoconductivity. A combination of bioceramic-polymer materials is generally used to provide an improved osteoconductive environment for bone healing. This study attempts developing for the first time an electrospun SF-based biocomposite system by introducing new bioceramics based on mesoporous bioactive glass/hydroxyapatite nanocomposite (MGHA). The addition of MGHA into the SF matrix could regulate the physicochemical properties and surface hydrophilicity, but induce weakened tensile properties as compared to pure SF. The excellent apatite-formation ability of a MGHA-introduced nanocomposite also improved the bioactivity of the composite. The biphasic composite increasingly degraded in PBS or enzyme solution in vitro compared with pure SF. In vivo evaluation of bone formation confirmed that SF/MGHA is more advantageous in bone reconstruction than the SF group for cranial bone defects. These results indicate the suitability of the SF/MGHA composite system in bone defects, demonstrating its potential application in bone tissue regeneration.

Adaptive extended state space predictive control for a kind of nonlinear systems.

This paper presents an adaptive nonlinear predictive control design strategy for a kind of nonlinear systems with output feedback coupling and results in the improvement of regulatory capacity for reference tracking, robustness and disturbance rejection. The nonlinear system is first transformed into an equal time-variant system by analyzing the nonlinear part. Then an extended state space predictive controller with a similar structure of a PI optimal regulator and with P-step setpoint feedforward control is designed. Because changes of the system state variables are considered in the objective function, the control performance is superior to conventional state space predictive control designs which only consider the predicted output errors. The proposed method is tested and compared with latest methods in literature. Tracking performance, robustness and disturbance rejection are improved.