Search

We present a data-driven method for monitoring machine status in manufacturing processes. Audio and vibration data from precision machining are used for inference in two operating scenarios: (a) variable machine health states (anomaly detection); and (b) settings of machine operation (state estimation). Audio and vibration signals are first processed through Fast Fourier Transform and Principal Component Analysis to extract transformed and informative features. These features are then used in the training of classification and regression models for machine state monitoring. Specifically, three classifiers (K-nearest neighbors, convolutional neural networks and support vector machines) and two regressors (support vector regression and neural network regression) were explored, in terms of their accuracy in machine state prediction. It is shown that the audio and vibration signals are sufficiently rich in information about the machine that 100% state classification accuracy could be accomplished. Data fusion was also explored, showing overall superior accuracy of data-driven regression models.

A methodology is presented for simultaneous sensor selection and design of fault diagnosis tests in complex systems, for which steady-state or dynamic models are available. The method assesses all possible sensor combinations for their information with respect to system faults in the presence of uncertainty and chooses sensors based on their contribution to information gain. In particular, sensors are cast as binary variables in the normalized Fisher information matrix, which is used to calculate the optimal fault detection and isolation (FDI) test designs by manipulating admissible system inputs. Test design optimization is always naturally embedded in the proposed method. Then, the Kullback–Leibler divergence and the Hellinger distance are used to explore the isolation capability of an FDI test when uncertainty is considered for the system inputs and parameters. The FDI tests are deployed and verified using a principal component analysis and the $k$ -nearest neighbor classification algorithm. The capability of the method to detect and isolate faults at high correct classification rates and low false alarm rates is demonstrated. The tool-chain proposed is tested on a benchmark three-tank system at various levels of measurement noise and uncertainty. The benefits and drawbacks of each step of the proposed method are assessed and discussed along with their computational footprint.

A model-based method is presented for the simultaneous selection of tests and sensors in fault detection and isolation (FDI) of systems subject to uncertainty. Tests and sensors are selected out of a continuous or discrete set of options based on their contribution to information gain with respect to fault identifiability. The objective of the optimization of the tests designed is to maximize the sensitivity of sensed outputs with respect to faults and minimize the joint confidence between faults and sources of uncertainty. The methodology is intended for active FDI in systems that can be limited to a finite number of input design scenarios, with a set of sensors that may or may not be valuable for the purpose of fault detection. The optimization of discrete sensors and input designs is formulated as a constrained mixed integer non-linear program that maximizes a measure of Fisher information, which calculates output sensitivities with respect to faults and uncertainty, by treating those as parameters in the system model. Kullback–Leibler divergence is used to determine the isolation capacity of a FDI test when there is uncertainty in inputs and parameters. FDI tests are executed using k -nearest neighbor classification, which is used as a verification method for test designs and sensor networks that result in high correct classification rates. The proposed design framework is tested on a virtual benchmark three-tank system, subject to multiple faults and sources of uncertainty.

Chemical-looping combustion is explored as a chemical reactor design problem. The continuous operation of fixed bed reactors using gaseous fuels for the purpose of power generation through integrat...

An algorithm is presented for the design of inferential sensors for fault diagnostics in thermal management systems. The algorithm uses input and output sensed system information to improve the detection and isolation of a fault by generating inferential sensors that augment the measured information to: (i) reduce the evidence of uncertainty in the inferred variables, and thus decrease false alarm and nondetection rates; and (ii) provide distinguishable responses to faults, and thus reduce reduce the rate of misdiagnoses. The novelty of the algorithm is its use of genetic programming to evolve explainable inferential sensors that maximize information criteria specific to fault diagnostics. The chosen criteria: (i) least squares regression; and (ii) Ds -optimality (calculated from the Fisher Information Matrix), leverage symbolic mathematics and automatic differentiation to obtain parametric sensitivities of the measured outputs and inferential sensors. The algorithm is included in a standard work for fault diagnostics, where its effectiveness is assessed through k-NN classification and illustrated in an application to an aircraft cross-flow plate-fin heat exchanger.

Fault detection and isolation (FDI) is becoming increasingly difficult due to the complexity and uncertainty of modern systems. For industrial systems with explicit models available, model-based active FDI tests can improve the capability for fault diagnosis. These tests should be determined and evaluated prior to implementation to minimize on-site computational costs. In this paper, a methodology is presented for the design optimization and assessment of tests for active fault diagnosis. The objective is to maximize the information from system outputs with respect to faults while minimizing the correlation between faults and uncertainty. After a test is designed, it is deployed with a k-nearest neighbor algorithm combined with principal component analysis.Two case studies are used to verify the proposed methodology, a three-tank system and a diesel engine.

The increasing uncertainty due to advances in modern systems has negatively impacted system health and safety. System health and safety problems associated with uncertainty can be attributed to (1) inadequate system design, (2) ill-suited controller and operating envelope, and (3) lack of robustness in system diagnostics. The latter issue (3) is the main focus of this paper. This work presents an algorithm for the design of active fault detection and isolation (FDI) tests that provide rigorous guarantees of robustness in safety-critical systems. A semi-infinite program with implicit functions embedded is formulated with the objective of maximizing FDI effectiveness at the worst-case realization of uncertainty by manipulating admissible system inputs, while taking into account system safety constraints. This problem is solved locally and globally illustrating deficiencies in the performance of FDI tests designed for the mean values of anticipated uncertainty. The resulting solution is an optimally performing fault diagnostic test that guarantees system safety over the entire domain of uncertainty. This is illustrated using a benchmark three-tank system and further analyzed through Monte Carlo simulation and k-NN classification.

A method for the optimal selection of sensors in active fault detection and isolation tests is presented. The usefulness of test information is improved using parametric sensitivities derived by a digital twin of the system. In this model-based approach, inputs are manipulated to maximize output sensitivities with respect to faults and minimize the joint confidence region between faults and other uncertain parameters. Two criteria are used to select optimal sensors based on information theory. The optimal selection is dependent on measurement precision, as shown in a case study of a virtual three-tank system subject to multiple faults.

The need for energy independence and security has refocused research on Fischer-Tropsch Synthesis (FTS) and its capability to selectively produce short and branched hydrocarbons. In this work, FTS selectivity to gasoline-range products was explored as a catalysis and process engineering problem. Novel structured catalysts for in situ upgrading of FTS products to gasoline-range hydrocarbons were successfully synthesized by coating controlled-thickness ZSM-5 films on the surface of Co-Al2O3/monolith substrates. These catalysts were synthesized on the premise that the monolith serves as an efficient mass and heat transfer structure, while the ZSM-5 layer functions as a hydrocarbon cracking and isomerization agent. Compared to traditional diesel-selective FTS, considerably higher quantity and quality of gasoline-range products were achieved with these bi-layered/bifunctional catalysts. Oil products with gasoline yield higher than 65 wt.% and selectivity to isomers and olefins greater than 70 wt.% were formed at relatively high temperature (230 °C) and intermediate pressure (12 bar), without compromising CO conversion, which remained as high as 79%.

Ni was impregnated on various mesoporous USY zeolites, which were characterized and tested for catalytic hydrodeoxygenation (HDO) of anisole. Parent USY zeolite was modified using the mild surfactant-assisted method, and more severe desilication and desilication with pore directed agent techniques. Ni impregnation on the zeolites with higher micropore volume produced small, easily reduced, internal Ni crystallites, whereas the larger mesopores promoted the formation of bulky external Ni particles. Catalysts with larger external Ni particles were more effective for HDO of anisole. HDO with the optimal mesoporous USY supported catalyst proceeded at more than double the reaction rate, with greater yield to fully deoxygenated products, compared to HDO using Ni supported on the parent USY. Kinetic modeling using a simple lumped mechanism demonstrated that this improvement was a result of greater rates of initial Ni-catalyzed hydrogenation reactions, most likely due to greater access to the Ni surface. The most successful technique was further modified to produce a final mesoporous USY supported Ni zeolite with superior catalytic activity for anisole HDO.

The major obstacles for the production of drop-in fuel from traditional fast pyrolysis are low bio-oil yield, poor fuel heating value and high solid yield. To overcome these obstacles, we studied the introduction of high pressure hydrogen into the pyrolysis reactor, in a process known as catalytic fast hydropyrolysis (CFHP). CFHP of anisole, a bio-oil model compound, was performed at different temperatures and pressures, to determine the effects of those variables on the CFHP reaction network. Alkanes were observed at low temperature and high pressure, whereas aromatics were dominant at low pressure and high temperature. Aspen modelling and calculation of reaction equilibrium constants showed that there is a thermodynamic barrier for simultaneous pyrolysis and hydrogenation at 600 °C. Reduced heating rates and CFHP followed by second-stage hydroprocessing (SH) were proposed as a means to control reaction temperature and efficiently deoxygenate CFHP vapors. The two-stage CFHP followed by hydroprocessing produced nearly 18 mol% carbon yield to alkanes, and a total bio-oil carbon yield of 25 mol%. The approximate bio-oil higher heating value (HHV) and aromaticity were 43.4 MJ/kg and 28 wt%, respectively, both within the range of gasoline.

Integration of combined cycle (CC) with chemical-looping combustion power plants (CLC) is studied for its potential for high power plant efficiency and low-cost CO2 capture. Dynamic modeling of the integrated process is used as a tool to analyze the extrema of CLC-CC power plant efficiency. Specifically, this work proposes control architectures for the CLC reactor and the integrated power plant. The time-averaged optimal CLC-CC power plant efficiency is estimated at 51.84% with CO2 capture efficiency at 96%. The main factor that limits the CLC-CC power plant efficiency is the reactor temperature, which is constrained by the oxygen carrier material. Plant-level sensitivity analysis shows that the inlet air temperature at the heat removal stage and gas compressor/gas turbine pressure ratio are the most important operating variables and if properly tuned the CLC-CC power plant can reach efficiencies sufficiently high for its economic deployment as a carbon neutral power generation option.

Useful simulation of complex systems for real-time or otherwise computationally constrained applications often requires the creation of simplified system models from existing analytical models and data. In this work, the creation of candidate models through machine learning and heuristic-informed model simplification methods are explored, and the resulting candidate models are evaluated and compared through an optimal experiment design process. The dynamic system of interest is a counterflow air-to-water heat exchanger.

Stranded Natural Gas (SNG) flaring has increased over the last decade resulting in additional Green House Gas (GHG) emissions and underutilization of our finite fossil fuel resources. However, utilization of SNG is difficult due to its low volume, inconsistent composition, and intermittent flow. In this paper, intensified Chemical looping for Ammonia Synthesis (CLAS) is explored as a process suitable for remote deployment at SNG sites. CLAS mediate ammonia synthesis by using compounds that carry nitrogen and hydrogen in stepwise reactions, allowing ammonia synthesis under milder conditions and smaller, more flexible processes. The challenge in CLAS is to find a feasible combination of materials and process conditions that yield ammonia at economically acceptable rates. We propose a framework that systematically evaluates CLAS to determine their suitability for SNG deployment. CLAS reaction schemes are reviewed and organized based on the mediating compounds used to synthesize ammonia. Different compounds are evaluated for each CLAS in search of a thermodynamically spontaneous pair. Spontaneous CLAS flowsheets are, then, optimized to maximize ammonia output at equilibrium. The results are ranked against performance metrics of energy intensity, efficiency, and cost. Chemical loops with Ca3N2/CaH2, SrH2/Sr3N2, MnO2/Mn5N2 and MoO2/Mo2N are found to have favorable performance, along with literature-reported evidence and are recommended for further evaluation toward a distributed ammonia synthesis solution.

System-level dynamic models of power plants are valuable tools for the assessment and prediction of plant performance, decisions on the design configuration, and the tuning of operating procedures and control strategies. In this work, the development of an integrated power plant model is presented. This model is validated against steady-state data from a subcritical power plant with reheat and regenerative cycles. The coal-fired power plant model studied has nominal power generation of 605 MW and efficiency of 38.3%. Traditional, regulatory control architectures are incorporated into and tuned with the dynamic power plant model. Dynamic simulation shows that the plant model is stable for sudden changes in coal load, and the controllers are able to maintain the controlled variables at their set points. In this two-part publication, we present the complete workflow of data collection, model development and validation, control tuning, dynamic optimization formulation and solution, and supervisory control architecture for a coal-fired subcritical power plant. Part I focuses on elements of model development and analysis, illustrating the advantages of acausal, object-oriented modeling in power plant simulation. Part II illustrates the use of this model for efficiency optimization under transient part-load operation.

This paper presents a supervisory control strategy for resilient chiller plants in the presence of condenser fouling. Fouling results in off-nominal performance of chiller parameters, such as increased refrigerant mass flow rate, compressor motor speed, discharge pressure, and discharge temperature. These effects further lead to faster deterioration of condenser pipes and tubes, and increase the risk of early motor failures. Thus, the main objective of this paper is to provide resilience, i.e., to bring the system parameters back to normalcy, and thereby protect the system from the adverse effects of fouling and improve its life expectancy while ensuring energy efficiency and meeting the desired cooling load. The supervisory control strategy presented here incorporates fault detection and diagnosis (FDD) and resilient control for mitigating the effects of condenser fouling. A computationally efficient and robust FDD scheme enables the estimation of the condenser fouling level using optimal sensor selection and statistical classifiers, thus facilitating condition-based maintenance. On the other hand, the resilient control scheme enables redistribution of load between chillers in order to reduce the load on faulty equipment in an energy-efficient manner, while still providing the required overall cooling load. The performance of this method is tested and validated using a high-fidelity chiller plant model and the proposed strategy is shown to diagnose condenser fouling with a high accuracy and effectively mitigate the effects of fouling at low computational cost. It is shown that the supervisory controller is able to meet the desired building load requirements at lower energy consumption, as compared with no supervisory control.

Effective simulation models of complex systems must achieve an appropriate balance between computational complexity and model fidelity. This work assesses that compromise, in the context of chiller plant modeling, simulation, and optimization. A high-fidelity, equation-based model of a water-cooled chiller is developed and evaluated in comparison to the DOE-2 semi-empirical model, as well as operating data from a real plant. The developed model is found to be realistic in a broader operating range, but computationally expensive for use in real-time optimization, taking approximately 10× computational time compared to the empirical model. The value of this high-fidelity model is demonstrated in simulation case studies that expand beyond the scope of the empirical model and through its suitability as a basis for the development of advanced control architectures. This work identifies the value proposition of the high-fidelity model and illustrates the advantages and disadvantages of each modeling approach, concluding with an outlook of the need for custom, reduced-order models for advanced analysis of chiller plants, targeted to specific uses, such as efficiency optimization and fault detection and accommodation.

A model-based method is presented for the simultaneous active diagnosis of faults and sensor selection for complex systems subject to uncertainty. Sensors are selected based on their contribution to information gain with respect to FDI. Admissible input trajectories are selected that maximize the sensitivity of sensed outputs with respect to faults and minimize the joint confidence between faults and sources of uncertainty. The methodology is intended for active FDI in systems that are limited to a finite number of input design scenarios. The optimization of discrete sensors and input designs is formulated as a constrained mixed non-linear integer problem that maximizes a measure of Fisher Information, which is based on output sensitivities with respect to faults and uncertainty represented as parameters in the system model. FDI tests are executed using k-nearest neighbor classification, which is used as a verification method for test designs and sensor networks that result in high correct classification rates. The proposed design framework is tested on a virtual benchmark three-tank system, subject to multiple faults and sources of uncertainty.

A chemical looping combustion (CLC) system with a multi-stage fuel reactor (FR) and a two-stage air reactor (AR), was constructed and operated in an experimental study and simulated via the computational particle fluid dynamics (CPFD) method. Hydrodynamic results with a bubble-based EMMS drag model agreed well with the experimental measurements. The wall effect in the small-scale AR led to an uneven particle distribution and irregular fluidization at a small NAR. The bubble size and velocity were limited in the FR, thereby significantly boosting the gas-solid contact. The particle residence time in the FR was prolonged, resulting in a high carbon capture rate. With the increase in fluidization number in FR and AR, the mass flux increased, but the residence time decreased. Adding a secondary gas flow in the FR has a positive impact on the solid circulation rate. The results in the thermal test indicated that a high CO2 capture rate and a low oxygen demand were achieved; thus the proposed reactor design improved the fuel conversion in the CLC process.

Fischer-Tropsch synthesis (FTS) is explored as a candidate one-step process for the selective conversion of synthesis gas to short and branched hydrocarbons. Novel structured catalysts for in-situ upgrading of FTS products to gasoline-range hydrocarbons are synthesized by coating controlled-thickness ZSM-5 films on the surface of Co-Al2O3/monolith substrates. FTS catalysts with coating of microporous and mesoporous ZSM-5 are synthesized, characterized and tested. These catalysts leverage the benefits of monolith reactors as efficient mass and heat transfer structures, with the ZSM-5 layer as a hydrocarbon cracking and isomerization agent. Hierarchical ZSM-5 co-catalysts are shown to relax mass transfer limitations through the zeolite film, thus improving conversion and selectivity. The proposed co-catalysts exhibit high CO conversion and high gasoline selectivity with high octane number. Loss of catalyst activity was observed and attributed to ZSM-5 pore plugging by C13+ saturated hydrocarbons. In-situ regeneration of the catalysts enabled stability in FTS performance.

Supported ZSM-5 is commonly used for in-situ biomass pyrolysis vapor upgrading in fluidized bed reactors. In this work, we investigate the effects of catalyst support on the pyrolysis product distribution by studying pyrolysis of woody biomass and bio-oil derived model compounds. Specifically, miscanthus, toluene and furan are catalytically pyrolyzed over bulk ZSM-5, supported ZSM-5 and a commercial Al 2 O 3 support. The pyrolysis product distributions of biomass and each model compound are presented and discussed. The coke formed on the catalysts and support, which is also responsible for catalyst deactivation, is analyzed using temperature programmed oxidation (TPO). It is shown that pyrolysis of miscanthus, toluene and furan over supported ZSM-5 produces less liquid products, particularly mono-aromatic hydrocarbons, than over bulk ZSM-5, mainly due to the dilution of Bronsted acid sites by the catalyst support. The Lewis acidity of the Al 2 O 3 support promotes alkylation and cyclization of propylene leading to enhanced benzene production. The Lewis acidity of the support also enhances coke production with a less condensed structure, while selectivity to poly-aromatics is deteriorated.

The feasibility of incorporating chemical-looping combustion (CLC) into a combined cycle power plant is explored. We propose a model-based framework to optimally design the cycle strategy for a CLC system, composed of multiple fixed bed reactors. The CLC reactor system is optimally integrated with the downstream gas turbine of a combined cycle power plant, by solving a dynamic program to optimize the temperature profile of the heat removal step out of each reactor, with constraints on the system performance during reduction and oxidation. We demonstrate the feasibility of fixed bed CLC systems to achieve different exhaust temperatures, using synthetic Cu- and Ni-based oxygen carriers. The optimal operating strategy is scaled-up to a system of fixed bed reactors operating in parallel. We show that through process optimization, the batch-operation of fixed bed CLC reactors is void of significant exhaust gas temperature fluctuations and can reach high thermal efficiencies with in-situ CO2 capture.