Chemical processes can bene t tremendously from fast and accurate ef uent composition prediction for plant design, control, and optimization. The Industry 4.0 revolution claims that by introducing machine learning int...Chemical processes can bene t tremendously from fast and accurate ef uent composition prediction for plant design, control, and optimization. The Industry 4.0 revolution claims that by introducing machine learning into these elds, substantial economic and environmental gains can be achieved. The bottleneck for high-frequency optimization and process control is often the time necessary to perform the required detailed analyses of, for example, feed and product. To resolve these issues, a framework of four deep learning arti cial neural networks (DL ANNs) has been developed for the largest chemicals production process steam cracking. The proposed methodology allows both a detailed characterization of a naphtha feedstock and a detailed composition of the steam cracker ef uent to be determined, based on a limited number of commercial naphtha indices and rapidly accessible process characteristics. The detailed char- acterization of a naphtha is predicted from three points on the boiling curve and paraf ns, iso-paraf ns, ole ns, naphthenes, and aronatics (PIONA) characterization. If unavailable, the boiling points are also estimated. Even with estimated boiling points, the developed DL ANN outperforms several established methods such as maximization of Shannon entropy and traditional ANNs. For feedstock reconstruction, a mean absolute error (MAE) of 0.3 wt% is achieved on the test set, while the MAE of the ef uent predic- tion is 0.1 wt%. When combining all networks using the output of the previous as input to the next the ef uent MAE increases to 0.19 wt%. In addition to the high accuracy of the networks, a major bene t is the negligible computational cost required to obtain the predictions. On a standard Intel i7 processor, predictions are made in the order of milliseconds. Commercial software such as COILSIM1D performs slightly better in terms of accuracy, but the required central processing unit time per reaction is in the order of seconds. This tremendous speed-up and minimal accuracy loss make the presented framework highly suitable for the continuous monitoring of dif cult-to-access process parameters and for the envi- sioned, high-frequency real-time optimization (RTO) strategy or process control. Nevertheless, the lack of a fundamental basis implies that fundamental understanding is almost completely lost, which is not always well-accepted by the engineering community. In addition, the performance of the developed net- works drops signi cantly for naphthas that are highly dissimilar to those in the training set.展开更多
By combining machine learning with the design of experiments,thereby achieving so-called active machine learning,more efficient and cheaper research can be conducted.Machine learning algorithms are more flexible and a...By combining machine learning with the design of experiments,thereby achieving so-called active machine learning,more efficient and cheaper research can be conducted.Machine learning algorithms are more flexible and are better than traditional design of experiment algorithms at investigating processes spanning all length scales of chemical engineering.While active machine learning algorithms are maturing,their applications are falling behind.In this article,three types of challenges presented by active machine learning—namely,convincing the experimental researcher,the flexibility of data creation,and the robustness of active machine learning algorithms—are identified,and ways to overcome them are discussed.A bright future lies ahead for active machine learning in chemical engineering,thanks to increasing automation and more efficient algorithms that can drive novel discoveries.展开更多
The developed SEMK model is used to provide an insight into the contribution of individual reactions in the cracking of methylcyclohexane as well as the site coverage by various carbenium ions. The preferred reaction ...The developed SEMK model is used to provide an insight into the contribution of individual reactions in the cracking of methylcyclohexane as well as the site coverage by various carbenium ions. The preferred reaction pathways for the conversion of methylcyclohexane are hydride transfer reactions followed by PCP-isomerizations, deprotonation and endocyclic β-scission, accounting for 61%, 22% and 12% of its disappearance, respectively, at 693 K and 30% conversion of methylcyclohexane. Protolysis plays a minor role in the cracking of methylcyclohexane. Once cyclic diolefins are formed, all of them can be instantaneously transformed to aromatics, which are easily interconverted via disproportionation. Judging from the carbenium ion concentrations it is evident that, at the investigated operating conditions, less than 5% of the acid sites are covered by carbenium ions, less than 2% of which corresponds to cyclic type species including allylic ones.展开更多
Up to 9% of the global CO_(2) emissions come from the iron and steel industry. Here, a combined chemical looping process to produce CO, a building block for the chemical industry, from the CO_(2) -rich blast furnace g...Up to 9% of the global CO_(2) emissions come from the iron and steel industry. Here, a combined chemical looping process to produce CO, a building block for the chemical industry, from the CO_(2) -rich blast furnace gas of a steel mill is proposed. This cyclic process can make use of abundant Fe_(3)O_(4) and CaO as solid oxygen and CO_(2) carriers at atmospheric pressure. A proof of concept was obtained in a laboratory-scale fixed bed reactor with synthetic blast furnace gas and Fe_(3)O_(4) /CaO = 0.6 kg/kg. CO production from the proposed process was investigated at both isothermal conditions(1023 K) and upon imposing a temperature program from 1023 to 1148 K. The experimental results were compared using performance indicators such as CO yield, CO space time yield, carbon recovery of the process, fuel utilisation, and solids’ utilisation.The temperature-programmed CO production resulted in a CO yield of 0.056 ± 0.002 mol per mol of synthetic blast furnace gas at an average CO space time yield of 7.6 mmol kgFe^(-1) s^(-1) over 10 cycles, carbon recovery of 48% ± 1%, fuel utilisation of 23% ± 2%, and an average calcium oxide and iron oxide utilisation of 22% ± 1% and 11% ± 1%. These experimental performance indicators for the temperature-programmed CO production were consistently better than those of the isothermal implementation mode by 20% to 35%. Over 10 consecutive process cycles, no significant losses in CO yield were observed in either implementation mode. Process simulation was carried out for 1 million metric tonnes per year of equivalent CO_(2) emissions from the blast furnace gas of a steel mill to analyse the exergy losses in both modes of operation. Comparison of the exergy efficiency of the temperature-programmed process to the isothermal process showed that the former is more efficient because of the higher CO concentration achievable,despite 20% higher exergy losses caused by heat transfer required to change temperature.展开更多
文摘Chemical processes can bene t tremendously from fast and accurate ef uent composition prediction for plant design, control, and optimization. The Industry 4.0 revolution claims that by introducing machine learning into these elds, substantial economic and environmental gains can be achieved. The bottleneck for high-frequency optimization and process control is often the time necessary to perform the required detailed analyses of, for example, feed and product. To resolve these issues, a framework of four deep learning arti cial neural networks (DL ANNs) has been developed for the largest chemicals production process steam cracking. The proposed methodology allows both a detailed characterization of a naphtha feedstock and a detailed composition of the steam cracker ef uent to be determined, based on a limited number of commercial naphtha indices and rapidly accessible process characteristics. The detailed char- acterization of a naphtha is predicted from three points on the boiling curve and paraf ns, iso-paraf ns, ole ns, naphthenes, and aronatics (PIONA) characterization. If unavailable, the boiling points are also estimated. Even with estimated boiling points, the developed DL ANN outperforms several established methods such as maximization of Shannon entropy and traditional ANNs. For feedstock reconstruction, a mean absolute error (MAE) of 0.3 wt% is achieved on the test set, while the MAE of the ef uent predic- tion is 0.1 wt%. When combining all networks using the output of the previous as input to the next the ef uent MAE increases to 0.19 wt%. In addition to the high accuracy of the networks, a major bene t is the negligible computational cost required to obtain the predictions. On a standard Intel i7 processor, predictions are made in the order of milliseconds. Commercial software such as COILSIM1D performs slightly better in terms of accuracy, but the required central processing unit time per reaction is in the order of seconds. This tremendous speed-up and minimal accuracy loss make the presented framework highly suitable for the continuous monitoring of dif cult-to-access process parameters and for the envi- sioned, high-frequency real-time optimization (RTO) strategy or process control. Nevertheless, the lack of a fundamental basis implies that fundamental understanding is almost completely lost, which is not always well-accepted by the engineering community. In addition, the performance of the developed net- works drops signi cantly for naphthas that are highly dissimilar to those in the training set.
基金financial support from the Fund for Scientific Research Flanders(FWO Flanders)through the doctoral fellowship grants(1185822N,1S45522N,and 3F018119)funding from the European Research Council(ERC)under the European Union’s Horizon 2020 research and innovation programme(818607)。
文摘By combining machine learning with the design of experiments,thereby achieving so-called active machine learning,more efficient and cheaper research can be conducted.Machine learning algorithms are more flexible and are better than traditional design of experiment algorithms at investigating processes spanning all length scales of chemical engineering.While active machine learning algorithms are maturing,their applications are falling behind.In this article,three types of challenges presented by active machine learning—namely,convincing the experimental researcher,the flexibility of data creation,and the robustness of active machine learning algorithms—are identified,and ways to overcome them are discussed.A bright future lies ahead for active machine learning in chemical engineering,thanks to increasing automation and more efficient algorithms that can drive novel discoveries.
基金the financial support from the China Scholarship Councilthe Long Term Structural Methusalem Funding by the Flemish Government
文摘The developed SEMK model is used to provide an insight into the contribution of individual reactions in the cracking of methylcyclohexane as well as the site coverage by various carbenium ions. The preferred reaction pathways for the conversion of methylcyclohexane are hydride transfer reactions followed by PCP-isomerizations, deprotonation and endocyclic β-scission, accounting for 61%, 22% and 12% of its disappearance, respectively, at 693 K and 30% conversion of methylcyclohexane. Protolysis plays a minor role in the cracking of methylcyclohexane. Once cyclic diolefins are formed, all of them can be instantaneously transformed to aromatics, which are easily interconverted via disproportionation. Judging from the carbenium ion concentrations it is evident that, at the investigated operating conditions, less than 5% of the acid sites are covered by carbenium ions, less than 2% of which corresponds to cyclic type species including allylic ones.
基金financial support from the project Cabon4PUR which has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 768919support of Dr. Alessandro Longo for Rietveld refinement of XRDsupport of the Wim Rogiers and Micha?l Lottin at the LCT for the fixed bed reactor setup used for experimental validation of the process concept。
文摘Up to 9% of the global CO_(2) emissions come from the iron and steel industry. Here, a combined chemical looping process to produce CO, a building block for the chemical industry, from the CO_(2) -rich blast furnace gas of a steel mill is proposed. This cyclic process can make use of abundant Fe_(3)O_(4) and CaO as solid oxygen and CO_(2) carriers at atmospheric pressure. A proof of concept was obtained in a laboratory-scale fixed bed reactor with synthetic blast furnace gas and Fe_(3)O_(4) /CaO = 0.6 kg/kg. CO production from the proposed process was investigated at both isothermal conditions(1023 K) and upon imposing a temperature program from 1023 to 1148 K. The experimental results were compared using performance indicators such as CO yield, CO space time yield, carbon recovery of the process, fuel utilisation, and solids’ utilisation.The temperature-programmed CO production resulted in a CO yield of 0.056 ± 0.002 mol per mol of synthetic blast furnace gas at an average CO space time yield of 7.6 mmol kgFe^(-1) s^(-1) over 10 cycles, carbon recovery of 48% ± 1%, fuel utilisation of 23% ± 2%, and an average calcium oxide and iron oxide utilisation of 22% ± 1% and 11% ± 1%. These experimental performance indicators for the temperature-programmed CO production were consistently better than those of the isothermal implementation mode by 20% to 35%. Over 10 consecutive process cycles, no significant losses in CO yield were observed in either implementation mode. Process simulation was carried out for 1 million metric tonnes per year of equivalent CO_(2) emissions from the blast furnace gas of a steel mill to analyse the exergy losses in both modes of operation. Comparison of the exergy efficiency of the temperature-programmed process to the isothermal process showed that the former is more efficient because of the higher CO concentration achievable,despite 20% higher exergy losses caused by heat transfer required to change temperature.
