Journal of Food Engineering最新文献

Spectral–spatial fusion attention transformer for brewing wheat variety identification and food-processing raw-material quality evaluation 酿酒小麦品种鉴定和食品加工原料质量评价的光谱-空间融合注意力转换器

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-20 DOI: 10.1016/j.jfoodeng.2026.112993

Liangliang Xie , Hao Zhang , Juan Wang , Yuansong Peng , Xiang Wan , Haili Yang , Xinjun Hu , Manjiao Chen , Jianping Tian , Dan Huang , Anying Cai , Rongzhi Wang , Jianping Yang , Kaiyang Yuan , Haonan Yi , Shunbo Zhang

Current spectral reconstruction methods exhibit limitations in jointly modeling spatial and spectral features, making it difficult to fully capture complex spectral information and thereby constraining reconstruction fidelity. To address this challenge, this paper proposes a novel Multi-stage Spectral–Spatial Fusion Attention Transformer (MSFAT), designed to enable efficient fusion and collaborative modeling of spatial-spectral data, ultimately improving the accuracy of spectral reconstruction. The core innovation lies in the development of the Spectral–Spatial Fusion Attention Module (SFAM), which integrates a dual frequency-spatial attention mechanism. By combining fast-Fourier transform with channel attention, SFAM effectively extracts global spectral dependencies and local spatial features synergistically. This mechanism allows SFAM to capture long-range spectral correlations in the frequency domain while emphasizing critical spatial details, significantly boosting the model's capacity to handle complex hyperspectral data. Experimental results validate the efficacy of MSFAT, showing a 0.37 dB improvement in PSNR over MST++ for hyperspectral image reconstruction, along with a 5.6 % reduction in MRAE and a 4.8 % decrease in RMSE. Furthermore, a one-dimensional (1D)-CNN-based brewing wheat variety classification model was developed using the reconstructed hyperspectral imaging data from MSFAT. This model achieved a test accuracy, recall, and F1 score of 96.66 %, 96.66 %, and 96.63 %, respectively—representing a 2.08 % accuracy improvement over MST++ and closely approximating the performance obtained using original spectral data (97.08 %). These findings demonstrate that integrating MSFAT with 1D-CNN offers a high-precision and cost-effective solution for the identification of brewing wheat raw materials, underscoring its broad potential for applications in smart agriculture and food composition analysis.In addition, the spectral data reconstructed by the proposed model can be further applied to analyze physicochemical characteristics relevant to food processing, such as starch, protein, and moisture content. This RGB-based spectral reconstruction approach provides a new avenue for raw-material sorting, quality assessment, and process monitoring in brewing and other food-processing operations.

目前的光谱重建方法在空间和光谱特征联合建模方面存在局限性，难以完全捕获复杂的光谱信息，从而限制了重建的保真度。为了解决这一挑战，本文提出了一种新型的多级光谱-空间融合注意力转换器（MSFAT），旨在实现空间-光谱数据的高效融合和协同建模，最终提高光谱重建的精度。其核心创新在于开发了频谱-空间融合注意模块（sfm），该模块集成了双频-空间注意机制。该方法将快速傅里叶变换与信道注意相结合，有效地协同提取全局频谱依赖关系和局部空间特征。该机制允许sfm在强调关键空间细节的同时捕获频域的远程光谱相关性，从而显著提高模型处理复杂高光谱数据的能力。实验结果验证了MSFAT的有效性，显示高光谱图像重建的PSNR比MST++提高了0.37 dB， MRAE降低了5.6%，RMSE降低了4.8%。此外，利用MSFAT重建的高光谱成像数据，建立了一维(1D)- cnn的酿造小麦品种分类模型。该模型的测试准确率、召回率和F1分数分别为96.66%、96.66%和96.63%，比MST++提高了2.08%，与使用原始光谱数据获得的性能（97.08%）非常接近。这些研究结果表明，将MSFAT与1D-CNN相结合，为酿造小麦原料的鉴定提供了高精度和经济高效的解决方案，强调了其在智能农业和食品成分分析方面的广泛应用潜力。此外，该模型重建的光谱数据可以进一步用于分析与食品加工相关的理化特征，如淀粉、蛋白质和水分含量。这种基于rgb的光谱重建方法为酿造和其他食品加工操作中的原料分类、质量评估和过程监控提供了新的途径。

{"title":"Spectral–spatial fusion attention transformer for brewing wheat variety identification and food-processing raw-material quality evaluation","authors":"Liangliang Xie , Hao Zhang , Juan Wang , Yuansong Peng , Xiang Wan , Haili Yang , Xinjun Hu , Manjiao Chen , Jianping Tian , Dan Huang , Anying Cai , Rongzhi Wang , Jianping Yang , Kaiyang Yuan , Haonan Yi , Shunbo Zhang","doi":"10.1016/j.jfoodeng.2026.112993","DOIUrl":"10.1016/j.jfoodeng.2026.112993","url":null,"abstract":"<div><div>Current spectral reconstruction methods exhibit limitations in jointly modeling spatial and spectral features, making it difficult to fully capture complex spectral information and thereby constraining reconstruction fidelity. To address this challenge, this paper proposes a novel Multi-stage Spectral–Spatial Fusion Attention Transformer (MSFAT), designed to enable efficient fusion and collaborative modeling of spatial-spectral data, ultimately improving the accuracy of spectral reconstruction. The core innovation lies in the development of the Spectral–Spatial Fusion Attention Module (SFAM), which integrates a dual frequency-spatial attention mechanism. By combining fast-Fourier transform with channel attention, SFAM effectively extracts global spectral dependencies and local spatial features synergistically. This mechanism allows SFAM to capture long-range spectral correlations in the frequency domain while emphasizing critical spatial details, significantly boosting the model's capacity to handle complex hyperspectral data. Experimental results validate the efficacy of MSFAT, showing a 0.37 dB improvement in PSNR over MST++ for hyperspectral image reconstruction, along with a 5.6 % reduction in MRAE and a 4.8 % decrease in RMSE. Furthermore, a one-dimensional (1D)-CNN-based brewing wheat variety classification model was developed using the reconstructed hyperspectral imaging data from MSFAT. This model achieved a test accuracy, recall, and F1 score of 96.66 %, 96.66 %, and 96.63 %, respectively—representing a 2.08 % accuracy improvement over MST++ and closely approximating the performance obtained using original spectral data (97.08 %). These findings demonstrate that integrating MSFAT with 1D-CNN offers a high-precision and cost-effective solution for the identification of brewing wheat raw materials, underscoring its broad potential for applications in smart agriculture and food composition analysis.In addition, the spectral data reconstructed by the proposed model can be further applied to analyze physicochemical characteristics relevant to food processing, such as starch, protein, and moisture content. This RGB-based spectral reconstruction approach provides a new avenue for raw-material sorting, quality assessment, and process monitoring in brewing and other food-processing operations.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"413 ","pages":"Article 112993"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146185769","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep visual measurement of corn starch particles size via RIS-UNet 基于RIS-UNet的玉米淀粉粒度的深度视觉测量

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-12 DOI: 10.1016/j.jfoodeng.2026.112982

Xuyan Zhao , Wenlu Zhu , Xiangyang Li , Hanwen Niu , Qiao Wang , Qunfeng Niu , Li Wang , Yuan Zhang

As a key ingredient in the food industry, corn starch has a particle size distribution that directly governs product processing performance and final product quality. Although image-based analysis offers an effective approach for precise particle characterization, particle adhesion still limits the accuracy of existing segmentation algorithms. To address the small size, adhesion, and diverse morphologies of starch particles, this study developed a high-resolution imaging system and constructed a dataset covering dispersed, partially agglomerated, and densely agglomerated states. A segmentation model termed RIS-UNet was proposed, integrating ResNet50 for deep feature extraction, Inception modules, and squeeze-and-excitation (SE) attention for improved boundary detection. Experiments showed RIS-UNet achieved a mean intersection over union (mIoU) of 94.92 % and accuracy of 98.71 %, outperforming UNet, Deeplab_v3, FCN, and Lraspp. Compared with a BT-9300H laser particle size analyzer, relative errors for median diameter (D50) and mean diameters were below 5 %, with highly consistent distribution curves. Overall, the proposed method provides a robust and cost-effective solution for intelligent and quantitative starch particle analysis.

作为食品工业的关键原料，玉米淀粉的粒度分布直接决定了产品的加工性能和最终产品的质量。尽管基于图像的分析为精确的颗粒表征提供了有效的方法，但颗粒粘附性仍然限制了现有分割算法的准确性。为了解决淀粉颗粒的小尺寸、粘附性和多种形态，本研究开发了一个高分辨率成像系统，并构建了一个涵盖分散、部分凝聚和密集凝聚状态的数据集。提出了一种称为RIS-UNet的分割模型，该模型集成了用于深度特征提取的ResNet50、Inception模块和用于改进边界检测的挤压和激励（SE）注意。实验表明，RIS-UNet的平均交联率（mIoU）为94.92%，准确率为98.71%，优于UNet、Deeplab_v3、FCN和Lraspp。与BT-9300H激光粒度分析仪相比，中位粒径（D50）与平均粒径的相对误差均在5%以下，分布曲线高度一致。总的来说，该方法为智能和定量淀粉颗粒分析提供了一种鲁棒性和成本效益高的解决方案。

{"title":"Deep visual measurement of corn starch particles size via RIS-UNet","authors":"Xuyan Zhao , Wenlu Zhu , Xiangyang Li , Hanwen Niu , Qiao Wang , Qunfeng Niu , Li Wang , Yuan Zhang","doi":"10.1016/j.jfoodeng.2026.112982","DOIUrl":"10.1016/j.jfoodeng.2026.112982","url":null,"abstract":"<div><div>As a key ingredient in the food industry, corn starch has a particle size distribution that directly governs product processing performance and final product quality. Although image-based analysis offers an effective approach for precise particle characterization, particle adhesion still limits the accuracy of existing segmentation algorithms. To address the small size, adhesion, and diverse morphologies of starch particles, this study developed a high-resolution imaging system and constructed a dataset covering dispersed, partially agglomerated, and densely agglomerated states. A segmentation model termed RIS-UNet was proposed, integrating ResNet50 for deep feature extraction, Inception modules, and squeeze-and-excitation (SE) attention for improved boundary detection. Experiments showed RIS-UNet achieved a mean intersection over union (mIoU) of 94.92 % and accuracy of 98.71 %, outperforming UNet, Deeplab_v3, FCN, and Lraspp. Compared with a BT-9300H laser particle size analyzer, relative errors for median diameter (D50) and mean diameters were below 5 %, with highly consistent distribution curves. Overall, the proposed method provides a robust and cost-effective solution for intelligent and quantitative starch particle analysis.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"412 ","pages":"Article 112982"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975543","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

High internal phase emulsions stabilized by milk protein-arabinoxylan-quercetin ternary complex: Enhanced environmental stability and fat digestive property (Milk protein-AX-Q complex-stabilized HIPEs: Stability and digestion) 乳蛋白-阿拉伯木聚糖-槲皮素三元配合物稳定的高内相乳剂：增强的环境稳定性和脂肪消化性能（乳蛋白- ax -q配合物稳定的HIPEs：稳定性和消化）

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-23 DOI: 10.1016/j.jfoodeng.2026.112997

Xuenan Gao , Peizhu Liu , Sirong Li , Olayemi Eyituoyo Dudu , Yanjun Li , Yan Li , Hongfang Yang , Jinju Cheng , Tingsheng Yan

In this study, ternary complexes were constructed using whey protein concentrate (WPC), sodium caseinate (SC), arabinoxylan (AX), and quercetin (Q), named WPC-AX-Q and SC-AX-Q, respectively, and their structural and functional properties were systematically characterized and compared with individual protein, protein/AX mixture and protein-AX binary conjugate. High internal phase emulsions (HIPEs) were also prepared to investigate their effects on the stability and fat digestion in the HIPEs. Results from fluorescence spectroscopy and Fourier transform infrared spectroscopy (FTIR) revealed notable alterations in the structure of milk proteins, suggesting interactions among milk proteins, AX and Q. Dynamic investigation into interfacial characteristics revealed that, in comparison to proteins alone, the ternary complexes exhibited lower interfacial tension and greater adsorption ability at the oil-water interface. Additionally, the ternary complexes showed a marked decrease in surface hydrophobicity and a notable increase in solubility, along with the most powerful foaming, emulsifying, and antioxidant capabilities compared with protein, protein/AX mixture and protein-AX binary conjugate. The stability of HIPEs stabilized by the ternary complexes was confirmed through rheological and microstructure studies. HIPEs stabilized by the ternary complexes exhibit superior environmental and in vitro digestive stability, and can mitigate the extent of lipid oxidation, delay and reduce the release of free fatty acids. This research offers theoretical backing for how the combined effect of polysaccharides and polyphenols impacts protein structure and properties, as well as the enhancement of fat digestion traits in HIPEs.

本研究以乳清浓缩蛋白（WPC）、酪蛋白酸钠（SC）、阿拉伯木聚糖（AX）和槲皮素(Q)为原料构建了三元配合物，分别命名为WPC-AX-Q和SC-AX-Q，系统表征了它们的结构和功能特性，并与单个蛋白、蛋白/AX混合物和蛋白-AX二元偶联物进行了比较。制备了高内相乳剂（HIPEs），考察了其对高内相乳剂稳定性和脂肪消化的影响。荧光光谱和傅里叶变换红外光谱（FTIR）结果显示，乳蛋白的结构发生了明显的变化，表明乳蛋白、AX和q之间存在相互作用。对界面特性的动态研究表明，与单独的蛋白质相比，三元配合物在油水界面具有更低的界面张力和更强的吸附能力。此外，与蛋白质、蛋白质/AX混合物和蛋白质-AX二元共轭物相比，三元配合物的表面疏水性显著降低，溶解度显著提高，同时具有最强的发泡、乳化和抗氧化能力。通过流变学和微观结构研究证实了三元配合物稳定HIPEs的稳定性。三元配合物稳定的HIPEs表现出优越的环境和体外消化稳定性，可以减轻脂质氧化的程度，延缓和减少游离脂肪酸的释放。本研究为多糖和多酚的联合作用如何影响HIPEs的蛋白质结构和性质，以及增强脂肪消化特性提供了理论支持。

{"title":"High internal phase emulsions stabilized by milk protein-arabinoxylan-quercetin ternary complex: Enhanced environmental stability and fat digestive property (Milk protein-AX-Q complex-stabilized HIPEs: Stability and digestion)","authors":"Xuenan Gao , Peizhu Liu , Sirong Li , Olayemi Eyituoyo Dudu , Yanjun Li , Yan Li , Hongfang Yang , Jinju Cheng , Tingsheng Yan","doi":"10.1016/j.jfoodeng.2026.112997","DOIUrl":"10.1016/j.jfoodeng.2026.112997","url":null,"abstract":"<div><div>In this study, ternary complexes were constructed using whey protein concentrate (WPC), sodium caseinate (SC), arabinoxylan (AX), and quercetin (Q), named WPC-AX-Q and SC-AX-Q, respectively, and their structural and functional properties were systematically characterized and compared with individual protein, protein/AX mixture and protein-AX binary conjugate. High internal phase emulsions (HIPEs) were also prepared to investigate their effects on the stability and fat digestion in the HIPEs. Results from fluorescence spectroscopy and Fourier transform infrared spectroscopy (FTIR) revealed notable alterations in the structure of milk proteins, suggesting interactions among milk proteins, AX and Q. Dynamic investigation into interfacial characteristics revealed that, in comparison to proteins alone, the ternary complexes exhibited lower interfacial tension and greater adsorption ability at the oil-water interface. Additionally, the ternary complexes showed a marked decrease in surface hydrophobicity and a notable increase in solubility, along with the most powerful foaming, emulsifying, and antioxidant capabilities compared with protein, protein/AX mixture and protein-AX binary conjugate. The stability of HIPEs stabilized by the ternary complexes was confirmed through rheological and microstructure studies. HIPEs stabilized by the ternary complexes exhibit superior environmental and in vitro digestive stability, and can mitigate the extent of lipid oxidation, delay and reduce the release of free fatty acids. This research offers theoretical backing for how the combined effect of polysaccharides and polyphenols impacts protein structure and properties, as well as the enhancement of fat digestion traits in HIPEs.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"412 ","pages":"Article 112997"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146074716","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Particle image velocimetry analysis of indica rice motion in moderate pressure polishing 中压抛光中籼米运动的粒子图像测速分析

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-21 DOI: 10.1016/j.jfoodeng.2026.112994

Linfeng Zhu, Xiaopeng Liu, Yonglin Zhang, Zhigang Hu, Beihai Wang, Qiang Yin

Conventional iron-roller polishing and over-processing often cause excessive breakage of indica rice. To overcome this limitation, we developed a spherical ribbed, flexible friction–collision polishing method. We applied particle image velocimetry (PIV) to explore rice polishing, representing the first attempt of this technology in this field. Specifically, PIV was applied to analyze rice grain motion in the polishing chamber, focusing on the effects of the roller speed and pressure gate opening on breakage and polishing quality. The results showed that grain trajectories followed regular global patterns, with local variations under different operating conditions. Alternating high- and low-speed flows reduced inter-grain collisions and significantly lowered the breakage compared with traditional polishing. At 4-mm gate opening, rice movement was constrained, leading to extended residence times and high breakage (4.53–6.21 %). At 8–12 mm openings, moderate pressure and optimal grain deflection angles were achieved. Within this range, speeds of 1000–1200 r min⁻¹ yielded spiral grain trajectories along the chamber wall, ensuring uniform polishing with low breakage (2.48–3.31 %). In contrast, excessive speeds (1200–1400 r min⁻¹) caused over-polishing, while a 16-mm gate opening resulted in insufficient pressure, high chalkiness and poor surface quality. This study provides a theoretical basis for optimising polishing machine parameters to minimise breakage and improve the processing quality of indica rice.

常规的铁辊抛光和过度加工往往会导致籼米过度破碎。为了克服这一限制，我们开发了一种球形肋、柔性摩擦碰撞抛光方法。我们将粒子图像测速（PIV）技术应用于大米抛光，这是该技术在该领域的首次尝试。具体而言，利用PIV分析了抛光室中米粒的运动，重点研究了辊速和压门开度对破碎率和抛光质量的影响。结果表明，粮食轨迹在全球范围内具有规律性，但在不同的操作条件下存在局部差异。与传统抛光相比，高低速交替流动减少了颗粒间碰撞，显著降低了破碎率。在4 mm开闸处，水稻运动受限，滞留时间延长，破损率高（4.53 ~ 6.21%）。在8 ~ 12 mm开孔处，压力适中，颗粒偏转角度最佳。在此范围内，1000-1200 r / min - 1的速度会产生沿腔壁的螺旋状颗粒轨迹，确保抛光均匀，破碎率低（2.48 - 3.31%）。相比之下，过高的速度（1200-1400 r min - 1）会导致过度抛光，而16毫米的浇口开口会导致压力不足，白垩度高，表面质量差。该研究为优化抛光机参数，减少破损，提高籼米加工质量提供了理论依据。

{"title":"Particle image velocimetry analysis of indica rice motion in moderate pressure polishing","authors":"Linfeng Zhu, Xiaopeng Liu, Yonglin Zhang, Zhigang Hu, Beihai Wang, Qiang Yin","doi":"10.1016/j.jfoodeng.2026.112994","DOIUrl":"10.1016/j.jfoodeng.2026.112994","url":null,"abstract":"<div><div>Conventional iron-roller polishing and over-processing often cause excessive breakage of indica rice. To overcome this limitation, we developed a spherical ribbed, flexible friction–collision polishing method. We applied particle image velocimetry (PIV) to explore rice polishing, representing the first attempt of this technology in this field. Specifically, PIV was applied to analyze rice grain motion in the polishing chamber, focusing on the effects of the roller speed and pressure gate opening on breakage and polishing quality. The results showed that grain trajectories followed regular global patterns, with local variations under different operating conditions. Alternating high- and low-speed flows reduced inter-grain collisions and significantly lowered the breakage compared with traditional polishing. At 4-mm gate opening, rice movement was constrained, leading to extended residence times and high breakage (4.53–6.21 %). At 8–12 mm openings, moderate pressure and optimal grain deflection angles were achieved. Within this range, speeds of 1000–1200 r min<sup>−1</sup> yielded spiral grain trajectories along the chamber wall, ensuring uniform polishing with low breakage (2.48–3.31 %). In contrast, excessive speeds (1200–1400 r min<sup>−1</sup>) caused over-polishing, while a 16-mm gate opening resulted in insufficient pressure, high chalkiness and poor surface quality. This study provides a theoretical basis for optimising polishing machine parameters to minimise breakage and improve the processing quality of indica rice.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"413 ","pages":"Article 112994"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077119","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Moisture adsorption characteristics of freeze-dried royal jelly: thermodynamic properties, modeling, and assessing microstructural changes 冻干蜂王浆的吸湿特性：热力学性质、建模和微观结构变化评估

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-20 DOI: 10.1016/j.jfoodeng.2026.112989

Rachida Ouaabou , Lahcen Hssaini , Abderrahim Alahyane , Abdessamad Lahrach , Nadia Massaoudi , Jalal Isaad , EL amine Ajal , Aadil Bajoub , Said Ennahli

This study aimed to investigate the sorption characteristics of royal jelly (RJ) to better understand how it interacts with moisture under varying conditions. We employed gravimetric static techniques to generate sorption isotherms, using chemical salts to precisely control water activity (a_w) over a range of 0.05–0.90. The experiments were conducted at three different temperatures 30 °C, 40 °C, and 50 °C. To gain further insights, we applied the Clausius-Clapeyron equation to calculate the integral thermodynamic properties of the RJ water system under constant spreading pressure. Our goal was to determine the best storage conditions to maintain the quality and extend the shelf life of royal jelly powder.

The results showed that royal jelly's sorption isotherms followed a Type II sigmoidal pattern, meaning that moisture content increased as water activity rose. Among the models we tested, the Enderby model provided the most accurate fit to our experimental data. Furthermore, the analysis of integral entropy revealed a minimum value when the moisture content ranged between 4 and 5 % (dry basis) and the relative humidity was between 0.15 and 0.23. These specific conditions were identified as ideal for prolonging the shelf life and preserving the quality of royal jelly. Scanning Electron Microscopy revealed an agglomerated structure, and thermal analysis showed a multi-stage degradation process, with the major mass loss occurring above 246 °C. Overall, this comprehensive investigation into RJ's sorption properties, including a detailed assessment of its thermodynamic behavior, offers valuable insights for optimizing its storage and preservation, ensuring the integrity of this highly nutritious food product.

本研究旨在研究蜂王浆（RJ）的吸附特性，以更好地了解它在不同条件下与水分的相互作用。我们采用重力静态技术生成吸附等温线，使用化学盐在0.05-0.90范围内精确控制水活度（aw）。实验在30°C、40°C和50°C三种不同的温度下进行。为了进一步深入了解，我们应用Clausius-Clapeyron方程计算了恒定扩散压力下RJ水体系的积分热力学性质。我们的目的是确定最佳的储存条件，以保持蜂王浆粉的质量和延长保质期。结果表明，蜂王浆的吸附等温线呈II型s型，即水分含量随水分活度的升高而增加。在我们测试的模型中，Enderby模型对我们的实验数据提供了最准确的拟合。此外，积分熵分析表明，当含水率在4% ~ 5%（干基）之间，相对湿度在0.15 ~ 0.23之间时，积分熵值最小。这些条件是延长蜂王浆保质期和保持蜂王浆质量的理想条件。扫描电镜显示其呈团块状结构，热分析显示其存在多阶段降解过程，主要质量损失发生在246℃以上。总的来说，对RJ吸附特性的全面调查，包括对其热力学行为的详细评估，为优化其储存和保存提供了有价值的见解，确保了这种高营养食品的完整性。

{"title":"Moisture adsorption characteristics of freeze-dried royal jelly: thermodynamic properties, modeling, and assessing microstructural changes","authors":"Rachida Ouaabou , Lahcen Hssaini , Abderrahim Alahyane , Abdessamad Lahrach , Nadia Massaoudi , Jalal Isaad , EL amine Ajal , Aadil Bajoub , Said Ennahli","doi":"10.1016/j.jfoodeng.2026.112989","DOIUrl":"10.1016/j.jfoodeng.2026.112989","url":null,"abstract":"<div><div>This study aimed to investigate the sorption characteristics of royal jelly (RJ) to better understand how it interacts with moisture under varying conditions. We employed gravimetric static techniques to generate sorption isotherms, using chemical salts to precisely control water activity (a<sub>w</sub>) over a range of 0.05–0.90. The experiments were conducted at three different temperatures 30 °C, 40 °C, and 50 °C. To gain further insights, we applied the Clausius-Clapeyron equation to calculate the integral thermodynamic properties of the RJ water system under constant spreading pressure. Our goal was to determine the best storage conditions to maintain the quality and extend the shelf life of royal jelly powder.</div><div>The results showed that royal jelly's sorption isotherms followed a Type II sigmoidal pattern, meaning that moisture content increased as water activity rose. Among the models we tested, the Enderby model provided the most accurate fit to our experimental data. Furthermore, the analysis of integral entropy revealed a minimum value when the moisture content ranged between 4 and 5 % (dry basis) and the relative humidity was between 0.15 and 0.23. These specific conditions were identified as ideal for prolonging the shelf life and preserving the quality of royal jelly. Scanning Electron Microscopy revealed an agglomerated structure, and thermal analysis showed a multi-stage degradation process, with the major mass loss occurring above 246 °C. Overall, this comprehensive investigation into RJ's sorption properties, including a detailed assessment of its thermodynamic behavior, offers valuable insights for optimizing its storage and preservation, ensuring the integrity of this highly nutritious food product.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"413 ","pages":"Article 112989"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146185765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Surplus bread fermentation by Pleurotus ostreatus monitored by magnetic resonance imaging 利用磁共振成像技术监测平菇发酵剩余面包

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-05 DOI: 10.1016/j.jfoodeng.2026.112967

Maja Musse , Isabelle Gimbert , Yves Diascorn , Stéphane Quellec , Sylvain Challois , Craig B Faulds , Tiphaine Lucas

This study addresses circular approaches to reuse food surpluses and losses, focusing on starchy products like bread that are fermented by fungi to produce a nutrient-rich biomass. Traditional methods have evaluated fungal growth at a macro scale, but improving the processing of heterogeneous substrates, such as highly porous ones, requires detailed insight into local fungal activity and growth. Magnetic Resonance Imaging (MRI) offers a unique advantage by simultaneously capturing morphological details and water distribution. The present work aimed to assess MRI as a non-destructive, non-invasive tool for monitoring spatial and temporal changes in biomass during solid-state fermentation (SSF) of bread slices with Pleurotus ostreatus. MRI protocols were developed to measure morphological changes and map transverse NMR relaxation times. MRI data were validated against conventional measurements of water content and dry matter, with further evaluation through fibre and protein analyses to determine added nutritional value.

Bread was shown to be a suitable substrate for cultivating fungi, producing a nutritionally-valuable biomass rich in soluble fibre and changes in some amino acid composition. T₂ relaxation was an effective focus to monitor mycelial growth through association to mycelial hydrolysis and water content distribution. T₂ mapping revealed significant heterogeneities in the bread, which persisted long under experimental conditions employed (approximatively one month). While the observed changes were at the macroscopic level, future microscale investigations using multi-exponential T₂ could help isolate fungal signals and enable quantitative biomass assessment.

这项研究解决了循环利用食物剩余和损失的方法，重点关注面包等淀粉类产品，这些产品由真菌发酵以产生营养丰富的生物质。传统的方法已经在宏观尺度上评估了真菌的生长，但是改进对异质基质（如高多孔基质）的处理，需要详细了解当地真菌的活性和生长。磁共振成像（MRI）提供了一个独特的优势，同时捕捉形态细节和水的分布。本研究旨在评估MRI作为一种非破坏性、非侵入性的工具，用于监测面包片与平菇固态发酵（SSF）过程中生物量的时空变化。MRI方案被开发用来测量形态学变化和绘制横向核磁共振弛豫时间。MRI数据与传统的含水量和干物质测量相对照，通过纤维和蛋白质分析进一步评估，以确定增加的营养价值。面包被证明是培养真菌的合适基质，可以产生富含可溶性纤维和改变某些氨基酸组成的有营养价值的生物质。T2弛豫是通过与菌丝水解和含水量分布的关联来监测菌丝生长的有效焦点。T2图谱揭示了面包的显著异质性，这种异质性在实验条件下持续了很长时间（大约一个月）。虽然观察到的变化是宏观水平的，但未来使用多指数T2的微观研究可以帮助分离真菌信号并进行定量生物量评估。

{"title":"Surplus bread fermentation by Pleurotus ostreatus monitored by magnetic resonance imaging","authors":"Maja Musse , Isabelle Gimbert , Yves Diascorn , Stéphane Quellec , Sylvain Challois , Craig B Faulds , Tiphaine Lucas","doi":"10.1016/j.jfoodeng.2026.112967","DOIUrl":"10.1016/j.jfoodeng.2026.112967","url":null,"abstract":"<div><div>This study addresses circular approaches to reuse food surpluses and losses, focusing on starchy products like bread that are fermented by fungi to produce a nutrient-rich biomass. Traditional methods have evaluated fungal growth at a macro scale, but improving the processing of heterogeneous substrates, such as highly porous ones, requires detailed insight into local fungal activity and growth. Magnetic Resonance Imaging (MRI) offers a unique advantage by simultaneously capturing morphological details and water distribution. The present work aimed to assess MRI as a non-destructive, non-invasive tool for monitoring spatial and temporal changes in biomass during solid-state fermentation (SSF) of bread slices with <em>Pleurotus ostreatus</em>. MRI protocols were developed to measure morphological changes and map transverse NMR relaxation times. MRI data were validated against conventional measurements of water content and dry matter, with further evaluation through fibre and protein analyses to determine added nutritional value.</div><div>Bread was shown to be a suitable substrate for cultivating fungi, producing a nutritionally-valuable biomass rich in soluble fibre and changes in some amino acid composition. T<sub>2</sub> relaxation was an effective focus to monitor mycelial growth through association to mycelial hydrolysis and water content distribution. T<sub>2</sub> mapping revealed significant heterogeneities in the bread, which persisted long under experimental conditions employed (approximatively one month). While the observed changes were at the macroscopic level, future microscale investigations using multi-exponential T<sub>2</sub> could help isolate fungal signals and enable quantitative biomass assessment.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"412 ","pages":"Article 112967"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975542","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Dual-encapsulated γ-cyclodextrin liposomal hydrogel beads crosslinked with Ca2+-SHP for pH-responsive and targeted lutein delivery 双封装γ-环糊精脂质体水凝胶珠与Ca2+-SHP交联，用于ph响应和靶向叶黄素递送

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-21 DOI: 10.1016/j.jfoodeng.2026.112995

Lanxin Zhang , Shengnan Wang , Xuejing Tian , Rui Wang , He Liu

In this study, core-shell hydrogel microspheres were synthesized through cross-linking of soybean hull polysaccharides (SHP) with Ca²⁺ and subsequently integrating them with γ-CD lutein liposomes (γ-CD Lu-lip) to form γ-CD Lu-lip@SHP beads. FTIR confirmed hydrogen bonding as the primary interaction between γ-CD Lu-lip and SHP. The successful encapsulation of lutein was further confirmed by XRD, DSC, and FTIR analyses in both γ-CD Lu-lip and γ-CD Lu-lip@SHP beads, with an encapsulation efficiency of 84.14 %. The resulting microspheres exhibited pH-responsive release behavior, characterized by Fick diffusion at pH 2.0, 4.0 and 6.8 values and non-Fick diffusion at pH 7.4. This pH-dependent behavior enabled controlled and sustained lutein release under simulated gastrointestinal conditions, ultimately enhancing lutein's oral bioaccessibility to 29.21 %. This study reveals that γ-CD Lu-lip@SHP bead systems can effectively deliver and protect environmentally labile bioactive compounds through controlled release mechanisms.

本研究将大豆壳多糖（SHP）与Ca2+交联合成核壳水凝胶微球，并将其与γ-CD叶黄素脂质体（γ-CD Lu-lip）结合形成γ-CD Lu-lip@SHP微球。FTIR证实了γ-CD Lu-lip与SHP之间的主要相互作用为氢键。通过对γ-CD Lu-lip和γ-CD Lu-lip@SHP微球的XRD、DSC和FTIR分析，进一步证实了叶黄素的包封成功，包封效率为84.14%。所得微球表现出pH响应释放行为，在pH值为2.0、4.0和6.8时表现为菲克扩散，而在pH值为7.4时表现为非菲克扩散。这种ph依赖性行为使叶黄素在模拟胃肠道条件下的控制和持续释放成为可能，最终将叶黄素的口服生物可及性提高到29.21%。本研究揭示了γ-CD Lu-lip@SHP微球体系可以通过控释机制有效地传递和保护环境不稳定的生物活性化合物。

{"title":"Dual-encapsulated γ-cyclodextrin liposomal hydrogel beads crosslinked with Ca2+-SHP for pH-responsive and targeted lutein delivery","authors":"Lanxin Zhang , Shengnan Wang , Xuejing Tian , Rui Wang , He Liu","doi":"10.1016/j.jfoodeng.2026.112995","DOIUrl":"10.1016/j.jfoodeng.2026.112995","url":null,"abstract":"<div><div>In this study, core-shell hydrogel microspheres were synthesized through cross-linking of soybean hull polysaccharides (SHP) with Ca<sup>2+</sup> and subsequently integrating them with γ-CD lutein liposomes (γ-CD Lu-lip) to form γ-CD Lu-lip@SHP beads. FTIR confirmed hydrogen bonding as the primary interaction between γ-CD Lu-lip and SHP. The successful encapsulation of lutein was further confirmed by XRD, DSC, and FTIR analyses in both γ-CD Lu-lip and γ-CD Lu-lip@SHP beads, with an encapsulation efficiency of 84.14 %. The resulting microspheres exhibited pH-responsive release behavior, characterized by Fick diffusion at pH 2.0, 4.0 and 6.8 values and non-Fick diffusion at pH 7.4. This pH-dependent behavior enabled controlled and sustained lutein release under simulated gastrointestinal conditions, ultimately enhancing lutein's oral bioaccessibility to 29.21 %. This study reveals that γ-CD Lu-lip@SHP bead systems can effectively deliver and protect environmentally labile bioactive compounds through controlled release mechanisms.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"412 ","pages":"Article 112995"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146074718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Enhancing the oral processing and swallowing performance of cod soft gels 提高鳕鱼软凝胶的口腔加工和吞咽性能

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-09 DOI: 10.1016/j.jfoodeng.2025.112937

Xiliang Yu , Yue Wang , Lin Han , Shiyi Liu , Michelle Ji Yeon Yoo , Xiuping Dong

The decline in body functions caused by aging often leads to dysphagia in older adults, thereby making a comfortable and safe eating process particularly important. The study investigated the effects of oral processing, swallowing and in vitro digestive behavior of cod protein soft gels (CGs) with different amounts of HIPEs added (also with different textural properties). In oral environment (37 °C), the CGs gradually liquefied. The modulus and yield strain were significantly reduced; the viscosity was increased compared to the room temperature (25 °C). During chewing, the CGs with HIPEs required more force to chew, but the number of chewing cycles was reduced. The oil droplet in HIPEs significantly decreased the friction coefficient of CGs, which is beneficial for safety and comfort in swallowing. CGs with added HIPEs were preferred by sensory evaluators. Multivariate analysis successfully differentiated the samples and identified that sensory attributes like swallowing residue' and key lubrication parameters (VIP>1) were the primary drivers of sample classification. The addition of HIPEs significantly affected the digestibility of proteins in CGs, with a gradual decrease in the particle size and number of peptides in the digests. The results of this study can be used as a reference for the development of food products for older people with different levels of dysphagia.

衰老引起的身体机能下降往往会导致老年人吞咽困难，因此，舒适和安全的饮食过程尤为重要。本研究考察了添加不同量的HIPEs（也具有不同的质地性质）对鳕鱼蛋白软凝胶（CGs）的口腔加工、吞咽和体外消化行为的影响。在口腔环境（37℃）下，CGs逐渐液化。模量和屈服应变显著降低；与室温（25℃）相比，粘度有所增加。在咀嚼过程中，HIPEs的CGs需要更大的咀嚼力，但咀嚼次数减少。HIPEs中的油滴显著降低了CGs的摩擦系数，有利于吞咽的安全性和舒适性。感官评估者更喜欢添加HIPEs的cg。多变量分析成功地对样品进行了分类，并发现吞咽残留物和关键润滑参数（VIP>1）等感官属性是样品分类的主要驱动因素。HIPEs的添加显著影响了猪对蛋白质的消化率，消化产物的颗粒大小和多肽数量逐渐减少。本研究结果可为开发针对不同程度吞咽困难老年人的食品提供参考。

{"title":"Enhancing the oral processing and swallowing performance of cod soft gels","authors":"Xiliang Yu , Yue Wang , Lin Han , Shiyi Liu , Michelle Ji Yeon Yoo , Xiuping Dong","doi":"10.1016/j.jfoodeng.2025.112937","DOIUrl":"10.1016/j.jfoodeng.2025.112937","url":null,"abstract":"<div><div>The decline in body functions caused by aging often leads to dysphagia in older adults, thereby making a comfortable and safe eating process particularly important. The study investigated the effects of oral processing, swallowing and <em>in vitro</em> digestive behavior of cod protein soft gels (CGs) with different amounts of HIPEs added (also with different textural properties). In oral environment (37 °C), the CGs gradually liquefied. The modulus and yield strain were significantly reduced; the viscosity was increased compared to the room temperature (25 °C). During chewing, the CGs with HIPEs required more force to chew, but the number of chewing cycles was reduced. The oil droplet in HIPEs significantly decreased the friction coefficient of CGs, which is beneficial for safety and comfort in swallowing. CGs with added HIPEs were preferred by sensory evaluators. Multivariate analysis successfully differentiated the samples and identified that sensory attributes like swallowing residue' and key lubrication parameters (VIP>1) were the primary drivers of sample classification. The addition of HIPEs significantly affected the digestibility of proteins in CGs, with a gradual decrease in the particle size and number of peptides in the digests. The results of this study can be used as a reference for the development of food products for older people with different levels of dysphagia.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"412 ","pages":"Article 112937"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975541","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Rapid traceability of coffee bean by integrating an electronic tongue with knowledge distillation and transfer learning 结合知识蒸馏和迁移学习的电子舌，实现咖啡豆的快速追溯

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-28 DOI: 10.1016/j.jfoodeng.2026.113001

Suchao Xu , Zhiqiang Wang , Tianrui Han , Jinyang Zhang , Zhen Li , Yubin Lan

The origin of coffee beans plays a crucial role in determining their flavor, quality, and market value. Conventional traceability methods possess several disadvantages, such as high costs, complex procedures, and time-consuming operations. This study proposes a rapid coffee beans traceability method by using an electronic tongue (ET) integrated with knowledge distillation and transfer learning. First, the taste fingerprint signals of coffee beans sourced from diverse regions were collected by a self-developed ET system. The continuous wavelet transform (CWT) was applied to convert the 1D signal into 2D spectrograms for effectively characterizing the time-frequency features within the ET signals. The resulting spectrograms were subsequently fed into the heterogeneous knowledge distillation framework to perform pattern recognition. Specifically, this framework employs a Vision Transformer (ViT) as the teacher model, which is optimized through transfer learning to strengthen its generalization performance under limited ET samples. A lightweight EfficientNet is adopted as the student model and trained under the guidance of the optimized teacher to achieve accurate discrimination among different coffee beans. Experimental results demonstrate that the distilled EfficientNet not only maintains a low computational cost but also achieves superior origin recognition performance compared with conventional deep learning models, with an accuracy of 98.33 % and a parameter count of only 11.3 M. This work provides a rapid, low-cost, and portable analytical approach for coffee bean origin traceability, making it easily deployed on edge or portable devices for on-site detection.

咖啡豆的产地在决定其风味、质量和市场价值方面起着至关重要的作用。传统的可追溯性方法有几个缺点，如高成本、复杂的程序和耗时的操作。本研究提出了一种结合知识蒸馏和迁移学习的电子舌（ET）咖啡豆快速溯源方法。首先，利用自主研发的ET系统采集不同产地咖啡豆的味道指纹信号。利用连续小波变换（CWT）将一维信号转换为二维频谱图，有效表征ET信号的时频特征。然后将得到的谱图输入到异构知识蒸馏框架中进行模式识别。具体而言，该框架采用视觉转换器（Vision Transformer, ViT）作为教师模型，并通过迁移学习对其进行优化，以增强其在有限ET样本下的泛化性能。采用轻量级的EfficientNet作为学生模型，在优化后的老师指导下进行训练，实现对不同咖啡豆的准确区分。实验结果表明，与传统的深度学习模型相比，蒸馏后的effentnet不仅保持了较低的计算成本，而且具有优越的原产地识别性能，准确率达到98.33%，参数数仅为11.3 M.这项工作为咖啡豆原产地溯源提供了一种快速、低成本、便携式的分析方法，使其易于部署在边缘或便携式设备上进行现场检测。

{"title":"Rapid traceability of coffee bean by integrating an electronic tongue with knowledge distillation and transfer learning","authors":"Suchao Xu , Zhiqiang Wang , Tianrui Han , Jinyang Zhang , Zhen Li , Yubin Lan","doi":"10.1016/j.jfoodeng.2026.113001","DOIUrl":"10.1016/j.jfoodeng.2026.113001","url":null,"abstract":"<div><div>The origin of coffee beans plays a crucial role in determining their flavor, quality, and market value. Conventional traceability methods possess several disadvantages, such as high costs, complex procedures, and time-consuming operations. This study proposes a rapid coffee beans traceability method by using an electronic tongue (ET) integrated with knowledge distillation and transfer learning. First, the taste fingerprint signals of coffee beans sourced from diverse regions were collected by a self-developed ET system. The continuous wavelet transform (CWT) was applied to convert the 1D signal into 2D spectrograms for effectively characterizing the time-frequency features within the ET signals. The resulting spectrograms were subsequently fed into the heterogeneous knowledge distillation framework to perform pattern recognition. Specifically, this framework employs a Vision Transformer (ViT) as the teacher model, which is optimized through transfer learning to strengthen its generalization performance under limited ET samples. A lightweight EfficientNet is adopted as the student model and trained under the guidance of the optimized teacher to achieve accurate discrimination among different coffee beans. Experimental results demonstrate that the distilled EfficientNet not only maintains a low computational cost but also achieves superior origin recognition performance compared with conventional deep learning models, with an accuracy of 98.33 % and a parameter count of only 11.3 M. This work provides a rapid, low-cost, and portable analytical approach for coffee bean origin traceability, making it easily deployed on edge or portable devices for on-site detection.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"413 ","pages":"Article 113001"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146185785","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A computational study for batch-level diagnostics in industrial extrusion using interpretable machine learning and counterfactual analysis 使用可解释机器学习和反事实分析的批量级工业挤出诊断的计算研究

IF 5.8 2区农林科学 Q1 ENGINEERING, CHEMICAL

Journal of Food Engineering

Pub Date : 2026-06-01 Epub Date: 2026-01-29 DOI: 10.1016/j.jfoodeng.2026.112999

Mads Kjærgaard Nielsen , Jacob Mikkelsen

Industrial feed extrusion involves complex interactions between raw materials, processing conditions, and product outcomes. This study develops machine learning models to predict critical process and product attributes across key stages of extrusion, conditioning, extrusion, and drying, based on a large-scale, two-year dataset with high batch and recipe variability. Ensemble methods and neural networks are applied to model outcomes such as specific mechanical energy, die pressure, bulk density, hardness, and the chemical composition of final products. Model performance was evaluated using cross-validation and test data, with XGBoost achieving the highest predictive accuracy. R² values exceeded 0.90 for specific mechanical energy and compositional attributes and were around 0.80 for physical pellet characteristics and die pressure. Beyond prediction, model interpretability was addressed using Shapley Additive Explanations to uncover key variable interactions and support process transparency. These insights enabled simulation of counterfactual scenarios to assess how adjusting upstream parameters could potentially influence product outcomes. The approach demonstrates how machine learning can support targeted batch optimization and reverse engineering in industrial settings, offering a scalable framework for integrating data-driven modelling into complex manufacturing workflows.

工业饲料挤压涉及原料、加工条件和产品结果之间复杂的相互作用。本研究开发了机器学习模型，以基于具有高批次和配方可变性的大规模两年数据集，预测挤出、调理、挤出和干燥关键阶段的关键工艺和产品属性。集成方法和神经网络应用于模型结果，如比机械能，模具压力，体积密度，硬度和最终产品的化学成分。使用交叉验证和测试数据评估模型性能，XGBoost实现了最高的预测准确性。比机械能和成分属性的R2值超过0.90，颗粒物理特性和模具压力的R2值在0.80左右。除了预测之外，还使用Shapley加性解释解决了模型的可解释性，以揭示关键变量的相互作用并支持过程透明度。这些见解使模拟反事实情景能够评估调整上游参数可能如何影响产品结果。该方法展示了机器学习如何在工业环境中支持有针对性的批量优化和逆向工程，为将数据驱动的建模集成到复杂的制造工作流程中提供了可扩展的框架。

{"title":"A computational study for batch-level diagnostics in industrial extrusion using interpretable machine learning and counterfactual analysis","authors":"Mads Kjærgaard Nielsen , Jacob Mikkelsen","doi":"10.1016/j.jfoodeng.2026.112999","DOIUrl":"10.1016/j.jfoodeng.2026.112999","url":null,"abstract":"<div><div>Industrial feed extrusion involves complex interactions between raw materials, processing conditions, and product outcomes. This study develops machine learning models to predict critical process and product attributes across key stages of extrusion, conditioning, extrusion, and drying, based on a large-scale, two-year dataset with high batch and recipe variability. Ensemble methods and neural networks are applied to model outcomes such as specific mechanical energy, die pressure, bulk density, hardness, and the chemical composition of final products. Model performance was evaluated using cross-validation and test data, with XGBoost achieving the highest predictive accuracy. R<sup>2</sup> values exceeded 0.90 for specific mechanical energy and compositional attributes and were around 0.80 for physical pellet characteristics and die pressure. Beyond prediction, model interpretability was addressed using Shapley Additive Explanations to uncover key variable interactions and support process transparency. These insights enabled simulation of counterfactual scenarios to assess how adjusting upstream parameters could potentially influence product outcomes. The approach demonstrates how machine learning can support targeted batch optimization and reverse engineering in industrial settings, offering a scalable framework for integrating data-driven modelling into complex manufacturing workflows.</div></div>","PeriodicalId":359,"journal":{"name":"Journal of Food Engineering","volume":"413 ","pages":"Article 112999"},"PeriodicalIF":5.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077118","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0