Bayesian Network Inference Research Articles

High accuracy uncertainty-aware interatomic force modeling with equivariant Bayesian neural networks

We demonstrate, that by using a novel adaptive step size method, high-quality Bayesian neural network inference via Markov chain sampling becomes practically viable for equivariant neural network architectures without a cold posterior effect.

Promoting the transformation of agricultural supply chain management under the development of digital economy in the Internet era

Abstract Agricultural products, as the necessities of people’s daily consumption, are related to the national economy and people’s livelihood, and the development of information technology provides a new direction and ideas for the transformation of agricultural products supply chain management. The article proposes the risk assessment method for the agricultural products supply chain based on the Bayesian network inference algorithm and OWA algorithm, firstly constructs the Bayesian network model for risk assessment of agricultural products supply chain, elaborates on the belief updating algorithm and the belief-free updating algorithm in the Bayesian network, then combines with the OWA operator, proposes the method of risk diagnosis of agricultural products supply chain based on the Bayes-OWA hybrid operator, and with this The algorithm is the core of the proposed agricultural supply chain management strategy. Through empirical tests, it can be obtained that the average supply time of enterprise 4 in agricultural products is 22.3 days, and the reliability of supply is 92.3%, which is highly efficient. In the agricultural products supply program, the average value of quality qualification rate has been improved by 2.67%, the average number of days for supplier delivery has been shortened by 13 days, and the supplier’s ability to respond quickly to demand has been improved by 3.61%. Therefore, actively constructing the informationization construction of the agricultural supply chain, building a logistics information platform, improving rural communication facilities, and providing timely market information to farmers can better ensure the efficient operation of the agricultural supply chain.

Analytically tractable heteroscedastic uncertainty quantification in Bayesian neural networks for regression tasks

The tractable approximate Gaussian inference (TAGI) method allows for analytical parameter inference in Bayesian neural networks. In its current form, TAGI can only model homoscedastic aleatory uncertainty that is quantified by a constant error variance across the input covariate-domain. In this paper, we present the approximate Gaussian variance inference (AGVI) method that enables analytical inference of the error variance term as a Gaussian random variable. The combined framework regrouping TAGI and AGVI, referred to as TAGI-V, enables modeling heteroscedastic aleatory uncertainty in Bayesian neural networks. TAGI-V outperforms the homoscedastic version of TAGI in terms of predictive performance for the benchmark regression datasets. In comparison with other approximate inference methods, TAGI-V is an order of magnitude faster and exhibits a comparable or superior predictive performance.

Data‐driven Bayesian network learning analysis on the regulatory mechanism between carcinogenic genes and immune cells

AbstractBackgroundIn recent years, it has become a research focus to accurately extract key genes influencing the occurrence and development of diseases from massive genomic data and study their regulatory mechanisms. Further exploration of these large databases is beneficial for us to identify key regulatory mechanisms in the occurrence and development of diseases, providing direction and theoretical basis for subsequent experimental design and research.MethodsBy using data from The Cancer Genome Atlas (TCGA) for lung adenocarcinoma (LUAD), the immune cell content of patients was obtained through deconvolution calculations in CIBERSORTx. Combined with Bayesian network inference methods, the impact of key gene expression on the heterocellular network influencing lung cancer was analyzed.ResultsWe found CD36 and ADRA1A genes were identified as two key mRNA genes influencing lung adenocarcinoma. The sensitivity analysis shows that the model performs well on both the testing set (error rate = 4%, AUC = 0.9804), the training set (error rate = 5.637%, AUC = 0.9746) and the verification set (error rate = 28.85%, AUC = 0.8689). The model has excellent predictive capabilities.ConclusionWe found that CD36 and ADRA1A genes may influence the development of lung cancer by affecting regulatory T cell and follicular helper T cell populations. Additionally, plasma cells may affect the expression of the CD36 gene. And the mutual information calculated by the model for these two genes is also the highest, indicating their potential as tumour biomarkers. Combining the network model, it can be inferred that CD36 and ADRA1A are likely to influence the occurrence and development of the disease through follicular helper T cells and regulatory T cells.

Identification of key brittleness factors and multi-scenario analysis of the water-energy-food-ecology nexus vulnerability based on NRS-BN

Water, energy and food are the basic resources on which human beings depend for survival. With the intensification of human activities, the demand for resources represented by water, energy and food continues to increase, resulting in increasing pressure on the ecological environment, and the vulnerability of water, energy, food and ecosystem becomes increasingly prominent. Identifying the critical vulnerability factors of the water-energy-food-ecology nexus and formulating targeted management measures have become the key to achieving sustainable development. This paper innovatively proposes to study the water-energy-food-ecology nexus from the perspective of vulnerability for the first time, in which the vulnerability evaluation index system of the water-energy-food-ecology nexus is firstly constructed based on the VSD framework, and the attribute reduction is carried out using neighborhood rough sets. Then, a Bayesian network model is built and parameter learning is performed by combining machine learning and expert experience. Finally, different scenarios are set up to identify the key factors that hinder the vulnerability reduction of the water-energy-food-ecology nexus and obtain the vulnerability probability of the nexus under different scenarios using forward and backward inference and sensitivity analysis of Bayesian networks, overcoming the drawback that many prediction models cannot achieve diagnostic inference. The results show that: 1) from 2008 to 2019, the overall vulnerability of the water-energy-food-ecology nexus in the Yangtze River Economic Belt is low. 2) The key factors at the indicator level that hinder the vulnerability reduction of the water-energy-food-ecology nexus mainly include the storage capacity of water conservancy projects, wastewater discharge per 10,000-yuan GDP, and water consumption per 10,000-yuan GDP, and the subsystem level is water, food, energy and ecology system, in that order. 3) The reduction in vulnerability within an individual subsystem can have a beneficial impact on reducing vulnerability within the water-energy-food-ecology nexus. However, this reduction may also lead to an increase in vulnerability within other subsystems. Therefore, in the process of developing water, energy, food, and ecology system, high priority should be given to the coordinated development of all four.

Temporal noisy-adder of bayesian network for scalable consecutive-k-out-of-n:F system reliability analysis

The Bayesian network(BN) is a powerful tool for modeling and analyzing reliability, safety, and risk in various engineering systems. Although there have been studies that incorporate it into the modeling of Consecutive-k-out-of-n: F systems(C(k,n:F)), these applications are limited to naive mode and cannot tackle scalable systems (e.g., n>2000,k>40). By introducing temporal Noisy-adder structure, named NaCkoon model, the number of nodes in equivalent C(k,n:F) BN model is significantly reduced, and the Conditional Probability Table (CPT) of temporary nodes can be controlled to linear complexity. The correctness of the proposed algorithm is validated by the comparison of literature case. Through numerical experiments, the novel algorithms proposed in this paper were compared with existing ones, including BN construction and inference computation time, was found to be the most optimal solution. Finally, the computable ranges of novel and existing algorithms are obtained through the capability computation. The test results show that NaCkoon model proposed in this paper has great advantages over existing algorithms and proposed NaA+NaO model based BN.

Assembly and phylogeographical analysis of novel Taenia solium mitochondrial genomes suggest stratification within the African-American genotype

BackgroundTaenia solium is a parasite of public health concern, causing human taeniasis and cysticercosis. Two main genotypes have been identified: Asian and African-American. Although characterizing T. solium genotypes is crucial to understanding the genetic epidemiology of its diseases, not much is known about the differences between T. solium mitochondrial genomes from different genotypes. Also, little is known about whether genotypes are further subdivided. Therefore, this study aimed to identify a set of point mutations distributed throughout the T. solium mitochondrial genome that differentiate the African-American from the Asian genotype. Another objective was to identify whether T. solium main genotypes are further stratified.MethodsOne Mexican and two Peruvian T. solium mitochondrial genomes were assembled using reads available in the NCBI Sequence Read Archive and the reference genome from China as a template. Mutations with respect to the Chinese reference were identified by multiple genome alignment. Jensen–Shannon and Grantham scores were computed for mutations in protein-coding genes to evaluate whether they affected protein function. Phylogenies by Bayesian inference and haplotype networks were constructed using cytochrome c oxidase subunit 1 and cytochrome b from these genomes and other isolates to infer phylogeographical relationships.ResultsA set of 31 novel non-synonymous point mutations present in all genomes of the African-American genotype were identified. These mutations were distributed across the mitochondrial genome, differentiating the African-American from the Asian genotype. All occurred in non-conserved protein positions. Furthermore, the analysis suggested a stratification of the African-American genotypes into an East African and a West African sublineage.ConclusionsA novel set of 31 non-synonymous mutations differentiating the main T. solium genotypes was identified. None of these seem to be causing differences in mitochondrial protein function between parasites of the two genotypes. Furthermore, two sublineages within the African-American genotype are proposed for the first time. The presence of the East African sublineage in the Americas suggests an underestimated connection between East African and Latin American countries that might have arisen in the major slave trade between Portuguese Mozambique and the Americas. The results obtained here help to complete the molecular epidemiology of the parasite.Graphical

A link between mitochondrial damage and the immune microenvironment of delayed onset muscle soreness

BackgroundDelayed onset muscle soreness (DOMS) is a self-healing muscle pain disorder. Inflammatory pain is the main feature of DOMS. More and more researchers have realized that changes in mitochondrial morphology are related to pain. However, the role of mitochondria in the pathogenesis of DOMS and the abnormal immune microenvironment is still unknown.MethodsMitochondria-related genes and gene expression data were obtained from MitoCarta3.0 and NCBI GEO databases. The network of mitochondrial function and the immune microenvironment of DOMS was constructed by computer algorithm. Subsequently, the skeletal muscle of DOMS rats was subjected to qPCR to verify the bioinformatics results. DOMS and non-DOMS histological samples were further studied by staining and transmission electron microscopy.ResultsBioinformatics results showed that expression of mitochondria-related genes was changed in DOMS. The results of qPCR showed that four hub genes (AMPK, PGC1-α, SLC25A25, and ARMCX1) were differentially expressed in DOMS. These hub genes are related to the degree of skeletal muscle immune cell infiltration, mitochondrial respiratory chain complex, DAMPs, the TCA cycle, and mitochondrial metabolism. Bayesian network inference showed that IL-6 and PGC1-α may be the main regulatory genes of mitochondrial damage in DOMS. Transmission electron microscopy revealed swelling of skeletal muscle mitochondria and disorganization of myofilaments.ConclusionsOur study found that skeletal muscle mitochondrial damage is one of the causes of inflammatory factor accumulation in DOMS. According to the screened-out hub genes, this study provides a reference for follow-up clinical application.

Train Brake System Pipe Leakage Detection and Early Warning Method Based on Bayesian Networks

This paper proposes a method for detecting and warning about leaks in train braking system pipelines based on Bayesian networks. Firstly, a detection model for pipeline leaks is established through the learning and inference of Bayesian networks. In the anomaly detection phase, the Bayesian network model is trained using historical data to monitor brake pressure abnormalities in real-time. Secondly, in the parameter regression calibration phase, the location and severity of the pipeline leaks are estimated based on the current brake pressure and relevant parameters. Finally, in the fault inference phase, the Bayesian network model is used to infer the possible causes of the leaks. The effectiveness and reliability of this method are verified through simulation design and actual data analysis. Compared to existing methods, this method can provide accurate leak detection and warning, thereby contributing to the safety of train operation. This research provides an effective method for detecting and warning about leaks in brake system pipelines and has practical application value.

Human PBMC scRNA-seq-based aging clocks reveal ribosome to inflammation balance as a single-cell aging hallmark and super longevity.

Quantifying aging rate is important for evaluating age-associated decline and mortality. A blood single-cell RNA sequencing dataset for seven supercentenarians (SCs) was recently generated. Here, we generate a reference 28-sample aging cohort to compute a single-cell level aging clock and to determine the biological age of SCs. Our clock model placed the SCs at a blood biological age to between 80.43 and 102.67 years. Compared to the model-expected aging trajectory, SCs display increased naive CD8+ T cells, decreased cytotoxic CD8+ T cells, memory CD4+ T cells, and megakaryocytes. As the most prominent molecular hallmarks at the single-cell level, SCs contain more cells and cell types with high ribosome level, which is associated with and, according to Bayesian network inference, contributes to a low inflammation state and slow aging of SCs. Inhibiting ribosomal activity or translation in monocytes validates such translation against inflammation balance revealed by our single-cell aging clock.

A comprehensive method for causation analysis of ship-ice collision risk in Arctic waters.

The influence of harsh Arctic environmental factors has caused ship-ice collision accidents to become the main threat faced by navigating ships. It is necessary to quantify the causation of ship accidents and perform effective risk management and control measures to ensure ship navigation safety. This study proposes a Bayesian network (BN) model-based risk analysis method for ship-ice collision accidents, which is used to quantitatively analyze the key risk factors and primary risk causation paths of ship accidents. First, the fault tree analysis (FTA) method is proposed to constructing the BN structure model, and the BN parameter solution method is further established. Consequently, a triangular fuzzy and defuzzification method is developed to quantify uncertain expert knowledge. Then, the BN inference method is used to analyze the collision risk causation where the North Atlantic and Arctic waters meet. The results indicate that the main risk factors in Arctic waters come from the environment. There are four primary risk causation paths; the proposed management and control measures of the risk causation paths (A), (B), (C), and (D) can effectively reduce the navigation risk by 29.95%, 9.98%, 25.05%, and 3.99%, respectively; the combination of these four measures can reduce the navigation risk by 54.63%.The proposed method has positive guiding significance for ensuring the safety of ship navigation in Arctic waters.

Use of machine learning-based integration to develop a monocyte differentiation-related signature for improving prognosis in patients with sepsis

BackgroundAlthough significant advances have been made in intensive care medicine and antibacterial treatment, sepsis is still a common disease with high mortality. The condition of sepsis patients changes rapidly, and each hour of delay in the administration of appropriate antibiotic treatment can lead to a 4–7% increase in fatality. Therefore, early diagnosis and intervention may help improve the prognosis of patients with sepsis.MethodsWe obtained single-cell sequencing data from 12 patients. This included 14,622 cells from four patients with bacterial infectious sepsis and eight patients with sepsis admitted to the ICU for other various reasons. Monocyte differentiation trajectories were analyzed using the “monocle” software, and differentiation-related genes were identified. Based on the expression of differentiation-related genes, 99 machine-learning combinations of prognostic signatures were obtained, and risk scores were calculated for all patients. The “scissor” software was used to associate high-risk and low-risk patients with individual cells. The “cellchat” software was used to demonstrate the regulatory relationships between high-risk and low-risk cells in a cellular communication network. The diagnostic value and prognostic predictive value of Enah/Vasp-like (EVL) were determined. Clinical validation of the results was performed with 40 samples. The “CBNplot” software based on Bayesian network inference was used to construct EVL regulatory networks.ResultsWe systematically analyzed three cell states during monocyte differentiation. The differential analysis identified 166 monocyte differentiation-related genes. Among the 99 machine-learning combinations of prognostic signatures constructed, the Lasso + CoxBoost signature with 17 genes showed the best prognostic prediction performance. The highest percentage of high-risk cells was found in state one. Cell communication analysis demonstrated regulatory networks between high-risk and low-risk cell subpopulations and other immune cells. We then determined the diagnostic and prognostic value of EVL stabilization in multiple external datasets. Experiments with clinical samples demonstrated the accuracy of this analysis. Finally, Bayesian network inference revealed potential network mechanisms of EVL regulation.ConclusionsMonocyte differentiation-related prognostic signatures based on the Lasso + CoxBoost combination were able to accurately predict the prognostic status of patients with sepsis. In addition, low EVL expression was associated with poor prognosis in sepsis.

Performing Bayesian Network Inference Using Amortized Region Approximation with Graph Factorization

Exact inference for large, directed graphical models, also known as Bayesian networks (BNs), can be intractable as the space complexity grows exponentially in the tree-width of the model. Approximate inference, such as generalized belief propagation (GBP), is used instead. GBP treats inference as the Bethe/Kikuchi energy function optimization problem. The solution is found using iterative message passing, which is inefficient and convergent problematic. Recent progress on amortized technique for GBP is an attractive alternative solution that can optimize the Bethe/Kikuchi energy function using (deep) neural networks, requiring no message passing. Despite being efficient, the amortized technique for GBP is applied to undirected graphical models with specific structures and factors, with no guarantee of the approximation quality. This is because the energy function to be amortized is defined by a region (or factor) graph that is ad hoc and difficult to construct to ensure sensible approximations. This paper proposes a new amortized GBP algorithm applied to BN for efficient inference. The proposed algorithm is composed of the following: (i) a new pairwise conversion (PWC) algorithm that converts all the conditional probability distributions in the BN into pairwise factors to facilitate efficient region graph constructions; (ii) following PWC, an improved loop structured region graph (LSRG) algorithm was derived to generate a valid region graph satisfying desired regional properties; and (iii) the energy function defined by the proposed PWC-LSRG region graph can be directly amortized using (deep) neural networks to ensure sensible approximations. Empirical studies show that the proposed amortized PWC-LSRG algorithm is of practical use and significantly improves convergence and efficiency compared to conventional algorithms.

Reliability Analysis of Failure-Dependent System Based on Bayesian Network and Fuzzy Inference Model

With the rapid development of information and automation technology, the manufacturing system is evolving towards more complexity and integration. The system components will inevitably suffer from degeneration, and the impact of component-level failure on the system reliability is a valuable issue to be studied, especially when failure dependence exists among the components. Thus, it is vital to construct a system reliability evaluation mechanism that helps to characterize the healthy status of the system and facilitate wise decision making. In this paper, a reliability analysis framework for a failure-dependent system is proposed, in which copula functions with optimized parameters are used for the description of different failure correlations, and a fuzzy inference model is constructed to derive the subsystem reliability based on the component-level failure correlation. Finally, a Bayesian network is applied to infer the system reliability based on the system structure combined with the impact of failure correlation inside. Simulation results of the proposed method show that the inference results of system reliability are reasonable and effective in different cases. Compared with the copula Bayesian network method, the proposed method shows better adaptability to failure-dependent systems to varying degrees. This work can provide theoretical guidance for evaluating the reliability of manufacturing systems of different types.

Exploring Spatiotemporal Patterns of Expressway Traffic Accidents Based on Density Clustering and Bayesian Network

Exploring spatiotemporal patterns of traffic accidents from historic crash databases is one essential prerequisite for road safety management and traffic risk prevention. Presently, with the emergence of GIS and data mining technologies, numerous geospatial analysis methods have been successfully adopted for traffic accident analysis. As characterized by high driving speeds, diverse vehicle types, and isolated traffic environments, expressways are confronted with more serious accident risks than urban roads. In this paper, we propose a combined method based on improved density clustering and the Bayesian inference network to explore spatiotemporal patterns of expressway accidents. Firstly, the spatiotemporal accident neighborhood is integrated into the DBSCAN clustering algorithm to discover multi-scale expressway black spots. Secondly, the Bayesian network model is separately employed in both local-scale black spots and regional-scale expressway networks to fully explore spatially heterogenous accident factors in various black spots and expressways. The experimental results show that the proposed method can correctly extract spatiotemporal aggregation patterns of multi-scale expressway black spots and meanwhile efficiently discover diverse causal factors for various black spots and expressways, providing a comprehensive analysis of accident prevention and safety management.

Lightweight Deep Neural Network Embedded with Stochastic Variational Inference Loss Function for Fast Detection of Human Postures.

Fusing object detection techniques and stochastic variational inference, we proposed a new scheme for lightweight neural network models, which could simultaneously reduce model sizes and raise the inference speed. This technique was then applied in fast human posture identification. The integer-arithmetic-only algorithm and the feature pyramid network were adopted to reduce the computational complexity in training and to capture features of small objects, respectively. Features of sequential human motion frames (i.e., the centroid coordinates of bounding boxes) were extracted by the self-attention mechanism. With the techniques of Bayesian neural network and stochastic variational inference, human postures could be promptly classified by fast resolving of the Gaussian mixture model for human posture classification. The model took instant centroid features as inputs and indicated possible human postures in the probabilistic maps. Our model had better overall performance than the baseline model ResNet in mean average precision (32.5 vs. 34.6), inference speed (27 vs. 48 milliseconds), and model size (46.2 vs. 227.8 MB). The model could also alert a suspected human falling event about 0.66 s in advance.

Electrical Fire Dynamic Risk Assessment for High‐Rise Buildings Based on Variable Fuzzy Set Theory and Bayesian Network

High‐rise buildings fires are far more harmful than ordinary fires. In this regard, fire risk assessment is an important way to control fire risk and reduce losses. This study presents a comprehensive model to electrical fire dynamic risk assessment of high‐rise buildings based on a Bayesian network (BN) and a variable fuzzy set theory (VFST). Firstly, electric system, safety management, and other factors were comprehensively analyzed based on three categories: hazard sources identification (HSI), fault tree (FT) analysis, and VFST. A high‐rise building electrical fire dynamic risk assessment model was established based on a BN. Secondly, the prior probability of BN root nodes was determined by VFST, and the conditional probability table (CPT) was determined by the analytic hierarchy process (AHP) and decomposition method. On that basis, the quantitative inference and sensitivity analysis can be performed on the electrical fire risks of high‐rise buildings in combination with the variable fuzzy Bayesian network (VFBN) inference. Finally, a high‐rise building in Wuhan, China, was used as an example for verification. The results show that the proposed method can realize dynamic risk assessment of electrical fires in high‐rise buildings. This study provides a new method for fire risk assessment of high‐rise buildings to reduce the possibility of fire.

Noise Prior Knowledge Informed Bayesian Inference Network for Hyperspectral Super-Resolution.

Well-known deep learning (DL) is widely used in fusion based hyperspectral image super-resolution (HS-SR). However, DL-based HS-SR models have been designed mostly using off-the-shelf components from current deep learning toolkits, which lead to two inherent challenges: i) they have largely ignored the prior information contained in the observed images, which may cause the output of the network to deviate from the general prior configuration; ii) they are not specifically designed for HS-SR, making it hard to intuitively understand its implementation mechanism and therefore uninterpretable. In this paper, we propose a noise prior knowledge informed Bayesian inference network for HS-SR. Instead of designing a "black-box" deep model, our proposed network, termed as BayeSR, reasonably embeds the Bayesian inference with the Gaussian noise prior assumption to the deep neural network. In particular, we first construct a Bayesian inference model with the Gaussian noise prior assumption that can be solved iteratively by the proximal gradient algorithm, and then convert each operator involved in the iterative algorithm into a specific form of network connection to construct an unfolding network. In the process of network unfolding, based on the characteristics of the noise matrix, we ingeniously convert the diagonal noise matrix operation which represents the noise variance of each band into the channel attention. As a result, the proposed BayeSR explicitly encodes the prior knowledge possessed by the observed images and considers the intrinsic generation mechanism of HS-SR through the whole network flow. Qualitative and quantitative experimental results demonstrate the superiority of the proposed BayeSR against some state-of-the-art methods.

The Core Industry Manufacturing Process of Electronics Assembly Based on Smart Manufacturing

This research takes a case study approach to show the development of a diverse adoption and product strategy distinct from the core manufacturing industry process. It explains the development status in all aspects of smart manufacturing, via the example of ceramic circuit board manufacturing and electronic assembly, and outlines future smart manufacturing plans and processes. The research proposed two experiments using artificial intelligence and deep learning to demonstrate the problems and solutions regarding methods in manufacturing and factory facilities, respectively. In the first experiment, a Bayesian network inference is used to find the cause of the problem of metal residues between electronic circuits through key process and quality correlations. In the second experiment, a convolutional neural network is used to identify false defects that were overinspected during automatic optical inspection. This improves the manufacturing process by enhancing the yield rate and reducing cost. The contributions of the study built in circuit board production. Smart manufacturing, with the application of a Bayesian network to an Internet of Things setup, has addressed the problem of residue and redundant conductors on the edge of the ceramic circuit board pattern, and has improved and prevented leakage and high-frequency interference. The convolutional neural network and deep learning were used to improve the accuracy of the automatic optical inspection system, reduce the current manual review ratio, save labor costs, and provide defect classification as a reference for preprocess improvement.

Risk Prediction of Gasifier System Based on Dynamic Cloud Bayesian Network

To address the current problem of insufficient risk level analysis of gasifier systems, an evaluation method combining the cloud model and dynamic Bayesian network is proposed, taking the gasifier system as the evaluation target, using the cloud model to discretize continuous data, and setting each risk factor in the evaluation system as a node in the dynamic Bayesian network to build a dynamic Bayesian network. The entropy weight method is used to calculate the weight of each risk indicator, the maximum likelihood estimation method is used to process the affiliation degree obtained from the cloud model, and the probability obtained by the affiliation-probability conversion method is input into the dynamic Bayesian network as evidence. Finally, the risk prediction assessment for the gasifier system is completed by using the features of forwarding and backward inference of the dynamic Bayesian network, combined with the comprehensive analysis of importance. The study shows that human maintenance efficiency, equipment integrity, gasifier pressure, and oxygen-coal ratio are the weak points that need to be focused on in the operation of the system.