Flame Images Research Articles

Effect and mechanism of combustion enhancement and emission reduction for non-premixed pure ammonia combustion based on fuel preheating

Ammonia (NH3) is a hydrogen-rich carrier with high energy density and is a carbon-free energy source. NH3 is a new carbon-free alternative fuel with great attraction and prospect. NH3 is restricted to large-scale industrial application because of its low reactivity and high nitrogen oxide (NOx) emissions in combustion. This research evaluated non-premixed pure NH3 combustion enhancement method and emission law based on fuel preheating. An NH3 preheating device, non-premixed burner, flue gas analysis, and temperature measurement system were designed and constructed. Combustion characteristics of NH3 preheating were investigated under different preheating temperatures and NH3 flow rates. NH3 flame images were captured by a CCD camera. The temperature distributions of the NH3 flames were reconstructed by deflection tomography technology. In addition, the composition concentrations of flue gas were obtained by a multi-component gas analyzer. The influence of preheating temperature on flame characteristics, combustion temperature field, and pollutant emission were analyzed. A numerical simulation was performed under the same experimental conditions. The effects of NH3 preheating on flame and emission characteristics were analyzed by sensitivity analysis and reaction path analysis. The mechanisms of the combustion enhancement and emission reduction for preheating NH3 non-premixed combustion were explained. Results showed that stable non-premixed pure NH3 flame was generated at different experimental flow rates when the NH3 preheating temperature was higher than 593 K. The brightness, color, and structure of the flame varied significantly with increasing preheating temperature. Furthermore, the concentration of NOx decreased obviously, and the concentration of unburned NH3 was extremely low. The chemical reaction rate was tremendously promoted with increasing preheating temperature, and the reaction pathway of N radical was extremely related to nitric oxide (NO) production and consumption.

Combustion Phenomena and Emissions in a Dual-Fuel Optical Engine Fueled with Diesel and Natural Gas

<div class="section abstract"><div class="htmlview paragraph">The application of dual-fuel combustion in the freight transportation sectors has received considerable attention due to the capability of achieving higher fuel efficiency and less pollutant emissions than the conventional diesel engines. In this study, high-speed flame visualization was used to investigate the phenomena of natural gas/diesel dual-fuel combustion in a single-cylinder heavy-duty engine with optical access. To implement diverse fuel blending conditions, diesel injection timing and natural gas substitution ratio were varied under constant fuel energy input. A novel flame regime separation method was implemented based on color segmentation in HSV color space to characterize the spatial distributions of premixed and non-premixed flame regimes. Flame images for larger natural gas substitution showed a significant reduction in the non-premixed flame regime accompanied by flame propagation along the vaporized diesel sprays. Advanced diesel injection shifted the location of early flames toward the piston bowl wall and created a rapid influx of propagating flame. These changes in fueling parameters resulted in the reduction of soot radiation, which was verified by the particulate matter emissions measured in the performance engine test under identical operating conditions.</div></div>

An ensemble deep learning model for exhaust emissions prediction of heavy oil-fired boiler combustion

Accurate and reliable prediction of exhaust emissions is crucial for combustion optimization control and environmental protection. This study proposes a novel ensemble deep learning model for exhaust emissions (NOx and CO2) prediction. In this ensemble learning model, the stacked denoising autoencoder is established to extract the deep features of flame images. Four forecasting engines include artificial neural network, extreme learning machine, support vector machine and least squares support vector machine are employed for preliminary prediction of NOx and CO2 emissions based on the extracted image deep features. After that, these preliminary predictions are combined by Gaussian process regression in a nonlinear manner to achieve a final prediction of the emissions. The effectiveness of the proposed ensemble deep learning model is evaluated through 4.2 MW heavy oil-fired boiler flame images. Experimental results suggest that the predictions are achieved from the four forecasting engines are inconsistent, however, an accurate prediction accuracy has been achieved through the proposed model. The proposed ensemble deep learning model not only provides accurate point prediction but also generates satisfactory confidence interval.

The Application of Flame Image Velocimetry to After-injection Effects on Flow Fields in a Small-Bore Diesel Engine

This study implements Flame Image Velocimetry (FIV), a diagnostic technique based on post-processing of high-speed soot luminosity images, to show the in-flame flow field development impacted by after-injection in a single-cylinder, small-bore optical diesel engine. Two after-injection cases with different dwell times between the main injection and after-injection, namely, close-coupled and long-dwell, as well as a main-injection-only case are compared regarding flow fields, flow vector magnitude, and turbulence intensity distribution. For each case, high-speed soot luminosity movies from 100 individual combustion cycles are recorded at a high frame rate of 45 kHz for FIV processing. The Reynolds decomposition using a spatial filtering method is applied to the obtained flow vectors so that bulk flow structures and turbulence intensity distributions can be discussed. The results show significant after-injection-induced flow structures as a group of vectors travelling back toward the center of the combustion chamber upon the jet impingement on the piston bowl wall and then jet-to-jet collision. Despite higher cyclic variations caused by the after-injection, increased bulk flow magnitude and turbulence intensity upon the after-injection event suggests locally enhanced mixing. Compared to the long-dwell after-injection, the close-coupled after-injection shows more significant turbulence enhancements.

Early detection of lean blowout in a combustor using symbolic analysis of colour images

Recently, lean combustion has been accepted as one of the preferable modes to run both small and large scale energy sectors as it reduces the level of NOx (oxides of nitrogen) emissions. However, ultra-lean combustion is more susceptible to lean blowout (LBO) which significantly reduces engine life and operating reliability. Therefore, early detection of LBO is important so that operator can get sufficient lead time to initiate precautionary actions. The current investigation proposes a novel technique based on the conversion of flame colour matrix into symbolic strings. The generation of symbolic strings is done based on the partitioning of the hue space obtained from transformation of RGB (Red-Green-Blue) to HSV (Hue-Saturation-Value) using uniform partitioning method. The partitions are determined at the reference state and kept fixed for other equivalence ratios (anomalous or non-reference states). Thus, the hue value at each pixel of the image is assigned a symbol depending on its belonging to a particular partition. Thereafter, an array of symbolic state probability vector at reference combustion state is determined based on the probability of pixel hue value being in a particular symbolic state or symbol (number of symbolic states is considered as 8). Also, for other anomalous or non-reference combustion states, symbolic state probability vectors are estimated using the partition formed at the reference state. The difference in symbolic state probability vectors between reference and anomalous combustion states gives the anomaly measure, which is used here to detect the flame blowout. The early increasing trend of anomaly towards the blowout for both premixed and partially premixed flames indicate that the measure can be helpful to detect the approach towards LBO. The estimation of anomaly is shown for using the flame images of different quantities. The random selection of five images from a flame video shows to be very effective in capturing the trend of anomaly which also ensures low computational effort. Thus, the tool based on symbolic colour image analysis (SCIA) may also be helpful in the mitigation of LBO.

Multiple Kernel Boosting Saliency Detection of Flame Image of Sintering Section

<p indent=0mm>The flame image of the sintering section is usually interfered by smoke and halo, which would cause the blurring of flame edge and material layer in the image. In order to solve the problem that traditional saliency detection method based on two-dimensional image features is difficult to effectively obtain the actual saliency of the cross-sectional flame image, a saliency detection method based on boundary connectivity and multi-kernel Boosting is proposed. Firstly, the color de-correlation is used in the process of image color space conversion. Boundary connectivity and dark channel prior principle are used to obtain the initial saliency map. Secondly, the super-pixel region information, regional variance and regional contrast of the flame image are used to construct region descriptor, the multiple kernel Boosting algorithm based on support vector regression is used to generate the complementary saliency map on 4 scales. Finally, the initial saliency map and the complementary saliency map are fused to obtain the final saliency map. 600 flame images including manual labeling are used to compare the proposed method with other 5 existing methods and each stage of the proposed method is analyzed. <italic>P</italic>-<italic>R</italic> curve, <italic>F</italic>-measure, mean absolute error and running time are taken as evaluation indexes. The experimental results show that the proposed method is superior to the other 5 methods, and the detection performance of each stage is gradually enhanced, which lays a foundation for improving the effective information extraction of sintering flame section image.

Dynamics of stratified swirl flame near lean blow out

This paper studies the flame dynamics near lean blowout (LBO) conditions in a stratified swirl burner. The novel BASIS burner is used and fueled with premixed methane/air mixture at atmospheric conditions. In the experiments, only the pilot flame works with four different equivalent ratios. It is found that as the flame changes from stable combustion to near LBO, the flame macrostructure changes from V-shape to M-shape. Meanwhile, a characteristic frequency of 405 Hz is found in the spectrum of heat release signal. The closer to LBO, the higher amplitude is found at the frequency peak. Post-processing of flame images illustrates that the 405 Hz comes from the circumferential movement of the flame. Large-eddy simulations (LESs) via OpenFOAM reproduce the flame dynamics near LBO, which agrees with the experimental results. LESs provide more insights into the flow fields, showing that the characteristic frequency of 405 Hz comes from the flame-vortex interaction. This study reveals the mechanism of flame and vortex interactions in a stratified swirl burner near LBO. The heat release signal is also found can be used as a candidate precursor for LBO.

Study on the inhibitory property and mechanism of inert powder on dust explosion flame of oil shale

In order to study the inhibitory property and mechanism of inert powder on dust explosion flame in oil shale, five common inert powder and two types of oil shale were selected for the explosion flame inhibition experiment by using the dust explosion flame propagation test system. First, a high-speed camera was used to record the flame image during the inhibition of oil shale dust explosion by inert powder, and the differences in explosion flame length, minimum inerting ratio and flame morphology and structure were analyzed in detail. Then, the thermal decomposition and endothermic characteristics of the inert powder were tested by Thermogravimetric-Differential Scanning Calorimetry (TG-DSC), and the TG-DTG-DSC thermal characteristic curve of the inert powder was analyzed. Combined with the analysis of the inhibition effect of the inert powder in the preheating zone and the combustion flame zone, the inhibitory property and the mechanism of the inert powder on the oil shale dust explosion flame were systematically studied. The research results show that the inhibitory performance of inert powder to the explosion flame of two kinds of oil shale dust is ranked as: ABC dry powder ＞Al(OH)3＞Mg(OH)2＞NaHCO3＞rock powder, and their explosion inhibition performance against Huadian oil shale (HDOS) is better than that of Longkou oil shale (LKOS). In this paper, the physical model of the inhibition mechanism of inert powder on the explosion flame of oil shale dust is established and the inhibition mechanism is analyzed. Through the mechanism analysis, it is shown that the high-efficiency explosive inhibition powder should have good thermal stability (decomposition temperature around 200−400 ℃), high heat absorption, and the decomposition of intermediate products can combine with the combustion reactive radical to play a chemical inhibition effect. The research results can provide theoretical basis and basis for the design of explosion inhibition of oil shale dust and the development of explosive inhibition powder.

Development of Conventional Controller Based on Image Processing for Monitoring and Controlling Burning Zone Temperature in a Cement Plant in Rotary Kiln Process Through IOT

In the cement factories, a rotary kiln is a pyro-processing device that is used to raise the temperature of the materials in a continuous process. Temperature monitoring is an essential process in the rotary kiln to yield high quality clinker and it has been implemented using various image processing techniques. In this paper we are measuring and controlling the temperature of rotational kiln in cement industry to get proper clinker ouput. Burning zone flame images are captured using CCD(Charge Coupled Device) camera and are processed using image processing with PID(Proportion Integration and Derivative) controller and which are programmed on raspberry pi card with the help of python language, also the captured images and attributes are transferred to authorized mobile/pc through Raspberry PI by selecting the IP address of mobile or PC. All the attributes received in the mobile in the form of web page the according to the object following data temperature controlled and object is ceaselessly followed to get the proper clinker output. Picture handling calculation with Open cv, as indicated by the calculation the edge estimation of the camera is settled. The frame value of the camera is set. Conversion from RGB color space to HSV color space is achieved and the reference color threshold value is determined. The range esteem estimated by the camera is contrasted and the reference esteem. In this study temp of rotational kiln is measured effectively using PID controller, this controller continuously control the temperature of revolving kiln by varying the i/p images of burning zone at finally fix one flame which is giving 1400degc.

Reconstruction of Temperature Field in Coal-fired Boiler Based on Limited Flame Image Information

ABSTRACT The internal combustion system of coal-fired boiler is a complex nonlinear system with large delay and strong coupling. The most basic requirement for the boiler combustion system is that the combustion flame inside the boiler is stable and uniform, and the flame temperature is the most important combustion characteristic of the combustion flame, so the method of using flame temperature as an intermediate variable to control the boiler system has been proposed by many scholars. Compared with the traditional control method that takes the main steam pressure and other parameters as the control variable, this method greatly reduces the delay of the system, and at the same time improves the stability and economy of the system. In this paper, a temperature field reconstruction method based on the improved two-color temperature measurement method is proposed. In this method, a mathematical model between the gray value of flame image and the parameters to be solved in the improved two-color temperature measurement method is established by using BP neural network. Then, the temperature field in furnace is reconstructed according to the gray value of flame image. This method solves the problem that traditional furnace temperature field reconstruction algorithm can not obtain accurate radiation energy data, which makes the measurement and calculation of this method simplified, and the accuracy is improved. The simulation results show that the maximum temperature error of this method is 4% and the average error is around 0.68%, which can meet the requirements of industry.

A novel method for the identification of flame front’s position on thermoacoustic coupling combustion oscillations

AbstractThe flame front area is an important parameter to quantify the heat release rate. However, the limitations imposed by the measuring instruments and the measurement methodologies make it difficult to determine the flame front position in an image. This work introduces a method to detect and optimize the flame front boundary. A high‐speed camera was employed to continuously capture the flame images. By setting appropriate intensity thresholds, the impact noise can be eliminated from the image and the flame front boundary can be determined. The binary diagram of the image was morphologically processed to obtain the normalized area fluctuations of the flame front in a temporally resolved manner. Two flame types and combustion regimes were studied. A LABVIEW‐based program was used for the synchronous triggering of the camera, the photomultiplier, and the microphones. The normalized area and photomultiplier output signals of a multiple flame burner obtained within the same period were compared and combined with the spectrum information from the microphone in the cavity. The trend charts were investigated in terms of the Pearson correlation coefficient. The results showed a strong correlation, thereby verifying the feasibility of this method.

Experimental investigation of flow field features and spark ignition process in a multi-swirl airblast injector

Lean combustion has been applied in aero gas turbine engines in response to rising concerns regarding the effects of pollutant emissions on the environment. As one of low emissions combustion technologies, stratified partially premixed combustion is capable of reducing NOx emissions while still providing robust pilot flames. In this work, the processes of liquid fuel atomization, fuel–air mixing and spark ignition in a multi-swirl staged combustor with airblast atomizer are experimentally investigated to further illustrate the effects of flow field features on flame propagation and the underlying reasons leading to successful and failed ignitions (misfires). A trajectory processing algorithm is developed to identify the detailed flame propagation behaviors in successful and misfire scenarios. PIV, PMie and High-speed imaging technologies are applied to provide the velocity distribution, droplets distribution and time-resolved flame images, respectively. Evolution of projected flames shows that the mechanism of flame motions in the early phase following spark can be explained by non-reacting flow structures and spray distributions. It is demonstrated that the early phase propagating trajectories play an important role in determining overall successful ignition. The variation regarding the flow field, droplets distribution, and flame evolution under pressure drops varying from 1% to 6% are analyzed. The ignition and combustion reach a relatively stable state when the pressure drop of injector exceeds 4%, which reflects in the minimum LLO FAR varies around 0.021 ± 0.001 and the time of flame development stabilizes at 40 ± 5 ms.

Investigation of the OH∗ chemiluminescence characteristics in CH4/O2 lifted flames

Premixed and non-premixed flames of CH4/O2 are widely used in industrial fields, including the non-catalytic partial oxidation (NC-POX) of natural gas to produce hydrogen-rich syngas. As a major spontaneous emission present in flames, OH∗ chemiluminescence is often applied in combustion diagnostics to indicate flame structure, reaction mechanism and equivalence ratio. In this study, the flame imaging technique was applied in order to investigate the flame structure, lift-off height, peak OH∗ intensity and core reaction zone for a CH4/O2 lifted flame impinging a plate. The results demonstrated that both the increase in the global equivalence ratio (λ) as well as the reduction in the plate-to-nozzle distance (L/De) can contribute the generation of the lifted flame. In regard to the fixed L/De, the lifted flame exhibited a narrower flame front and lower chemical reaction intensity in the case of a small λ. As L/De decreased, the position of the flame core reaction zone shifted to the flame base. Meanwhile, the core reaction zone of the flame and peak OH∗ intensity was found to increase progressively. In addition, with the increase in λ, the flame lift-off height (H) initially rose gradually, after which it then declined sharply and thereafter increased. As indicated by the flame lift-off characteristic, as λ rose, a circulation zone developed at the flame base, which was located between the flame inner shear layer and the jet of oxidizer. Accordingly, the circulation zone can promote the generation of a lifted flame due to the enhanced mixing of the reactants.

Combustion Instability of Swirl Premixed Flame with Dielectric Barrier Discharge Plasma

The effects of plasma on the combustion instability of a methane swirling premixed flame under acoustic excitation were investigated. The flame image of OH planar laser-induced fluorescence and the fluctuation of flame transfer function showed the mechanism of plasma in combustion instability. The results show that when the acoustic frequency is less than 100 Hz, the gain in flame transfer function gradually increases with the frequency; when the acoustic frequency is 100~220 Hz, the flame transfer function shows a trend of first decreasing and then increasing with acoustic frequency. When the acoustic frequency is greater than 220 Hz, the flame transfer function gradually decreases with acoustic frequency. When the voltage exceeds the critical discharge value of 5.3 kV, the premixed gas is ionized and the heat release rate increases significantly, thereby reducing the gain in flame transfer function and enhancing flame stability. Plasma causes changes in the internal recirculation zone, compression, and curling degree of the flame, and thereby accelerates the rate of chemical reaction and leads to an increase in flame heat release rate. Eventually, the concentration of OH radicals changes, and the heat release rate changes accordingly, which ultimately changes the combustion instability of the swirling flame.

A Statistical Image Feature-Based Deep Belief Network for Fire Detection

Detecting fires is of significance to guarantee the security of buildings and forests. However, it is difficult to fast and accurately detect fire stages in complex environment because of the large variations of the fire features of color, texture, and shapes for flame and smoke images. In this paper, a statistic image feature-based deep belief network (DBN) is proposed for fire detections. Firstly, for each individual image, all the statistic image features extracted from a flame and smoke image in time domain, frequency domain, and time-frequency domain are calculated to construct training and testing samples. Then, the constructed samples are fed into DBN to classify the multiple fire stages in complex environment. DBN can automatically learn fault features layer by layer using restricted Boltzmann machine (RBM). Experiments using the benchmark data of three groups of fire and fire-like images are classified by the present method, and the classification results are also compared with those commonly used support vector machine (SVM) and convolutional deep belief networks (CDBNs) to manifest the superiority of the classification accuracy.

The influence of inter-jet spacing and jet-swirl interaction on flame image velocimetry (FIV) derived flow fields in a small-bore diesel engine

This study applies Flame Image Velocimetry (FIV) to show the in-flame flow field development with an emphasis on the jet-jet interaction and jet-swirl interaction phenomena in a single-cylinder small-bore optically accessible diesel engine. Two-hole nozzle injectors with three different inter-jet spacing angles of 45°, 90° and 180° are prepared to cause different levels of jet-jet interaction. The engine has a swirl ratio of 1.7, which is used to evaluate jet-swirl interaction of the selected 180° inter-jet spacing nozzle. High-speed soot luminosity imaging was performed at a high frame rate of 45 kHz for the FIV processing. For each inter-jet spacing angle, a total of 100 individual combustion cycles were recorded to address the cyclic variations. The ensemble averaged flow fields are shown to illustrate detailed flow structures while the Reynolds decomposition using spatial filtering is applied to analyse turbulence intensity. The results showed reduced bulk flow magnitude and turbulence intensity at smaller inter-jet spacing, suggesting the two opposed wall-jet heads colliding immediately after the jet impingement on the wall can cause flow suppression effects. This raised a concern on the mixing as lower inter-jet spacing creates more fuel-rich mixtures in the jet-jet interaction region. Despite lower flow magnitude, the cyclic variation was also estimated higher for narrower inter-jet spacing, which is another drawback of the significant jet-jet interaction. Regarding the jet-swirl interaction, the wall-jet head penetrating on the up-swirl side showed lower bulk flow magnitude as the counter-flow arrangement suppressed the flow, similar with the narrower interact-jet spacing results. However, the turbulence intensity was measured higher on the up-swirl side, suggesting the relatively weaker swirl flow vectors opposed to the penetrating wall-jet head could in fact enhance the mixing.

Flow directed spark stretch and flame propagation in a high-tumble production engine

This study performs endoscopic high-speed imaging to enhance the fundamental knowledge of in-cylinder flow structure and flame development process in a selected high-tumble production engine. The endoscopic high-speed particle image velocimetry (eHS-PIV) was performed for varied engine speeds and intake valve closing (IVC) timings to evaluate their impact on the in-cylinder flow structure in a motored engine condition. On another endoscope engine sharing the same hardware, high-speed flame imaging was conducted to visualise spark stretch and flame propagation. The flow and flame measurements were repeated for over 100 cycles and the ensemble-averaged results are compared. The eHS-PIV showed that a strong tumble vortex is generated during the piston compression with the flow directed towards the exhaust side. As the piston reaches top dead centre (TDC), however, a complex flow breakup involving multiple flow components occurs. This is followed by lateral flow vectors travelling back towards the intake side, which is termed as the bounce-back flow. For a tested engine speed range of 1700–2700 revolutions per minute (rpm), 2500 rpm shows the most significant bounce-back flow as a result of competition between the remaining exhaust-ward tumble flow strength and the newly formed bounce-back flow strength. At a retarded IVC timing, the flow loss leads to a weakened tumble flow and subsequently no bounce-back flow formation to maintain the exhaust-ward TDC flow direction. From the comparison between the flow results and spark/flame high-speed images, a strong positive correlation is found between the TDC flow direction and spark plasma stretch, and subsequently the flame propagation direction. The findings indicate that the TDC flow direction should be considered as a key parameter in the engine design and operating condition settings.

Experimental study on self-excited thermoacoustic instabilities and intermittent switching of azimuthal and longitudinal modes in an annular combustor

For annular combustors, the interactions between different modal structures (longitudinal, azimuthal spinning, standing, and slanted) remain a challenge in ongoing studies of thermoacoustic instabilities. The present work develops an annular combustor named TurboCombo, similar to the multiple-injector annular combustor named MICCA, but with a convergent turbine section connected to the combustor outlet. A Helmholtz solver provides a proper estimate of the modal frequencies and reveals the spatial structures for both two rigs. The experimental results present the consistencies of frequency domains between the acoustic pressure signals and the hydroxyl (OH*) radicals emission signal. The high-speed flame images are processed by the dynamic mode decomposition method, statistically showing longitudinal, azimuthal spinning, or standing nature in different cases. The stability maps of the two combustors show slight differences under atmospheric conditions. In the TurboCombo rig, there are two types of interaction behaviors between the 1/4 wave longitudinal mode of the chamber and the 1st order azimuthal mode of the plenum. Unlike the prevalent “dual-mode coexistence” characteristics, an interesting behavior of “intermittent switching” is revealed, which manifests as conspicuous two-way jumps between the two modes with different frequencies. Such intermittent switching occurs randomly between two stable limit cycles in a long-time observation without any external excitation, but is driven by the turbulence-induced stochastic process. Moreover, there still exist some weak azimuthal oscillations while the longitudinal mode is dominated. The standing or spinning bias of the azimuthal mode also features intermittent transition, corresponding to the occasions dominated by longitudinal mode or azimuthal mode, respectively.

IR thermography study of flow structure and parameters in diffusion flames

The flame structure and flow modes have been investigated in the process of diffusion combustion of a number of liquid fuels, such as ethanol, benzine, kerosene and diesel fuel, by using the technique of infrared (IR) thermography in the narrow spectral range (2.5–2.7 μm). IR thermography provides useful information on the thermal behavior of turbulent diffusion flames, which is the thermal representation of turbulent processes in the hydrodynamic structure of flames. By analyzing IR images and spectra of flame temperature fluctuations, seven characteristic zones with different flow modes have been identified in the flames. To estimate flame flow parameters, the turbulent Reynolds number has been used thus taking account the amplitude and frequency of the temperature fluctuations determined by processing IR images of flames. The Reynolds number strongly increases during the diffusion combustion of kerosene, in contrast to benzine, diesel fuel and ethanol. This can be explained by a more complex chemical composition of kerosene and a multi-stage character of chemical reactions.

A hybrid deep neural network based prediction of 300 MW coal-fired boiler combustion operation condition

In power generation industries, boilers are required to be operated under a range of different conditions to accommodate demands for fuel randomness and energy fluctuation. Reliable prediction of the combustion operation condition is crucial for an in-depth understanding of boiler performance and maintaining high combustion efficiency. However, it is difficult to establish an accurate prediction model based on traditional data-driven methods, which requires prior expert knowledge and a large number of labeled data. To overcome these limitations, a novel prediction method for the combustion operation condition based on flame imaging and a hybrid deep neural network is proposed. The proposed hybrid model is a combination of convolutional sparse autoencoder (CSAE) and least support vector machine (LSSVM), i.e., CSAE-LSSVM, where the convolutional sparse autoencoder with deep architectures is utilized to extract the essential features of flame image, and then essential features are input into the least support vector machine for operation condition prediction. A comprehensive investigation of optimal hyper-parameter and dropout technique is carried out to improve the performance of the CSAE-LSSVM. The effectiveness of the proposed model is evaluated by 300 MW tangential coal-fired boiler flame images. The prediction accuracy of the proposed hybrid model reaches 98.06%, and its prediction time is 3.06 ms/image. It is observed that the proposed model could present a superior performance in comparison to other existing neural network models.