Abstract: In the era of big data, data-driven technologies are increasingly leveraged by industry to facilitate autonomous learning and intelligent decision-making. However, the challenge of “small samples in big data” emerges when datasets lack the comprehensive information necessary for addressing complex scenarios, which hampers adaptability. Thus, enhancing data completeness is essential. Knowledge-guided virtual sample generation transforms domain knowledge into extensive virtual datasets, thereby reducing dependence on limited real samples and enabling zero-sample fault diagnosis. This study used building air conditioning systems as a case study. We innovatively used the large language model (LLM) to acquire domain knowledge for sample generation, significantly lowering knowledge acquisition costs and establishing a generalized framework for knowledge acquisition in engineering applications. This acquired knowledge guided the design of diffusion boundaries in mega-trend diffusion (MTD), while the Monte Carlo method was used to sample within the diffusion function to create information-rich virtual samples. Additionally, a noise-adding technique was introduced to enhance the information entropy of these samples, thereby improving the robustness of neural networks trained with them. Experimental results showed that training the diagnostic model exclusively with virtual samples achieved an accuracy of 72.80%, significantly surpassing traditional small-sample supervised learning in terms of generalization. This underscores the quality and completeness of the generated virtual samples.

面向信息完备性增强和零样本故障诊断的建筑热力系统知识引导虚拟样本扩散生成方法

[1]BishopCM, 1995. Training with noise is equivalent to Tikhonov regularization. Neural Computation, 7(1):108-116.

[2]ChenKL, WangZW, GuXW, et al., 2021. Multicondition operation fault detection for chillers based on global density-weighted support vector data description. Applied Soft Computing, 112:107795.

[3]ChenZL, O’NeillZ, WenJ, et al., 2023. A review of data-driven fault detection and diagnostics for building HVAC systems. Applied Energy, 339:121030.

[4]ComstockMC, BraunJE, GrollEA, 2001. The sensitivity of chiller performance to common faults. HVAC&R Research, 7(3):263-279.

[5]DuM, RenFF, MinR, et al., 2024. Detecting non-uniform structures in oil-in-water bubbly flow experiments. Physica A: Statistical Mechanics and Its Applications, 637:129602.

[6]ForthK, BorrmannA, 2024. Semantic enrichment for BIM-based building energy performance simulations using semantic textual similarity and fine-tuning multilingual LLM. Journal of Building Engineering, 95:110312.

[7]FrankSJ, FrankAM, 2020. Salient slices: improved neural network training and performance with image entropy. Neural Computation, 32(6):1222-1237.

[8]GappmairW, 1999. Claude E. Shannon: the 50th anniversary of information theory. IEEE Communications Magazine, 37(4):102-105.

[9]GuoYB, LiuYX, ZhangZ, et al., 2024. Research on fault detection and diagnosis of carbon dioxide heat pump systems in buildings based on transfer learning. Journal of Building Engineering, 85:108774.

[10]HolmstromL, KoistinenP, 1992. Using additive noise in back-propagation training. IEEE Transactions on Neural Networks, 3(1):24-38.

[11]HuangCF, 1997. Principle of information diffusion. Fuzzy Sets and Systems, 91(1):69-90.

[12]HuangCF, MoragaC, 2004. A diffusion-neural-network for learning from small samples. International Journal of Approximate Reasoning, 35(2):137-161.

[13]JiangG, MaZH, ZhangL, et al., 2024. EPlus-LLM: a large language model-based computing platform for automated building energy modeling. Applied Energy, 367:123431.

[14]KhamisN, SelamatH, IsmailFS, 2022. Improved optimization parameters prediction using the modified mega trend diffusion function for a small dataset problem. Knowledge and Information Systems, 64(11):3129-3149.

[15]LiDC, WuCS, TsaiTI, et al., 2007. Using mega-trend-diffusion and artificial samples in small data set learning for early flexible manufacturing system scheduling knowledge. Computers & Operations Research, 34(4):966-982.

[16]LiGN, YaoQ, FanC, et al., 2021. An explainable one-dimensional convolutional neural networks based fault diagnosis method for building heating, ventilation and air conditioning systems. Building and Environment, 203:108057.

[17]LiGN, ChenL, LiuJY, et al., 2023a. Comparative study on deep transfer learning strategies for cross-system and cross-operation-condition building energy systems fault diagnosis. Energy, 263:125943.

[18]LiGN, XiongJH, TangR, et al., 2023b. In-situ sensor calibration for building HVAC systems with limited information using general regression improved Bayesian inference. Building and Environment, 234:110161.

[19]LiL, LiQH, NiYS, et al., 2024a. Critical penetrating vibration evolution behaviors of the gas-liquid coupled vortex flow. Energy, 292:130236.

[20]LiL, XuP, XuWX, et al., 2024b. Multi-field coupling vibration patterns of the multiphase sink vortex and distortion recognition method. Mechanical Systems and Signal Processing, 219:111624.

[21]LiL, XuP, LiQH, et al., 2025a. A coupled LBM-LES-DEM particle flow modeling for microfluidic chip and ultrasonic-based particle aggregation control method. Applied Mathematical Modelling, 143:116025.

[22]LiL, XuP, LiQH, et al., 2025b. Multi-field coupling particle flow dynamic behaviors of the microreactor and ultrasonic control method. Powder Technology, 454:120731.

[23]LiTT, ZhouYZ, ZhaoY, et al., 2022a. A hierarchical object oriented Bayesian network-based fault diagnosis method for building energy systems. Applied Energy, 306:118088.

[24]LiTT, ZhaoY, ZhangCB, et al., 2022b. A semantic model-based fault detection approach for building energy systems. Building and Environment, 207:108548.

[25]LiZ, WangCY, LiL, et al., 2024. Numerical investigation of mesoscale multiphase mass transport mechanism in fibrous porous media. Engineering Applications of Computational Fluid Mechanics, 18(1):2363246.

[26]LiuJY, LiX, ZhangQ, et al., 2023. An efficient sensor and thermal coupling fault diagnosis methodology for building energy systems. Energy and Buildings, 296:113367.

[27]LuJ, ZhangCB, LiJY, et al., 2022. Graph convolutional networks-based method for estimating design loads of complex buildings in the preliminary design stage. Applied Energy, 322:119478.

[28]LuJ, TianXN, ZhangCB, et al., 2025. Evaluation of large language models (LLMs) on the mastery of knowledge and skills in the heating, ventilation and air conditioning (HVAC) industry. Energy and Built Environment, 6(5):875-892.

[29]LuJG, LiDD, 2013. Bias correction in a small sample from big data. IEEE Transactions on Knowledge and Data Engineering, 25(11):2658-2663.

[30]NiyogiP, GirosiF, PoggioT, 1998. Incorporating prior information in machine learning by creating virtual examples. Proceedings of the IEEE, 86(11):2196-2209.

[31]QiGJ, LuoJB, 2022. Small data challenges in big data era: a survey of recent progress on unsupervised and semi-supervised methods. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(4):2168-2187.

[32]SanchezFAS, KhambampatiAK, KimKY, 2023. Generative adversarial network model for two-phase flow imaging by electrical impedance tomography. IEEE Transactions on Instrumentation and Measurement, 72:4507212.

[33]SivakumarJ, RamamurthyK, RadhakrishnanM, et al., 2022. Synthetic sampling from small datasets: a modified mega-trend diffusion approach using k-nearest neighbors. Knowledge-Based Systems, 236:107687.

[34]SunZ, YaoQW, 2024. Self-correction method for sensor faulty heat pump system based on machine learning. Results in Engineering, 22:102170.

[35]SunZ, JinHQ, GuJP, et al., 2019. Gradual fault early stage diagnosis for air source heat pump system using deep learning techniques. International Journal of Refrigeration, 107:63-72.

[36]SunZ, JinHQ, GuJP, et al., 2020. Studies on the online intelligent diagnosis method of undercharging sub-health air source heat pump water heater. Applied Thermal Engineering, 169:114957.

[37]SunZ, JinHQ, XuYJ, et al., 2022. Severity-insensitive fault diagnosis method for heat pump systems based on improved benchmark model and data scaling strategy. Energy and Buildings, 256:111733.

[38]SunZ, YaoQW, JinHQ, et al., 2024. A novel in-situ sensor calibration method for building thermal systems based on virtual samples and autoencoder. Energy, 297:131314.

[39]TanYF, NiYS, XuWX, et al., 2023. Key technologies and development trends of the soft abrasive flow finishing method. Journal of Zhejiang University-SCIENCE A, 24(12):1043-1064.

[40]TianY, DongQY, TianJD, et al., 2023. Capacity estimation of lithium-ion batteries based on optimized charging voltage section and virtual sample generation. Applied Energy, 332:120516.

[41]WangZW, WangZW, HeSW, et al., 2017. Fault detection and diagnosis of chillers using Bayesian network merged distance rejection and multi-source non-sensor information. Applied Energy, 188:200-214.

[42]WangZW, WangL, TanYY, et al., 2021. Fault diagnosis using fused reference model and Bayesian network for building energy systems. Journal of Building Engineering, 34:101957.

[43]WeiCH, OokaR, 2023. Indoor airflow field reconstruction using physics-informed neural network. Building and Environment, 242:110563.

[44]WhittakerT, JanikRA, OzY, 2024. Turbulence scaling from deep learning diffusion generative models. Journal of Computational Physics, 514:113239.

[45]WuJF, LiL, YinZC, et al., 2024. Mass transfer mechanism of multiphase shear flows and interphase optimization solving method. Energy, 292:130475.

[46]XuML, YoonS, FuentesA, et al., 2023. A comprehensive survey of image augmentation techniques for deep learning. Pattern Recognition, 137:109347.

[47]XuP, LiQH, WangCY, et al., 2025. Interlayer healing mechanism of multipath deposition 3D printing models and interlayer strength regulation method. Journal of Manufacturing Processes, 141:1031-1047.

[48]XuYJ, WangJF, ShenX, et al., 2023. Thermodynamic analyses and performance improvement on a novel cascade-coupling-heating heat pump system for high efficiency hot water production. Energy Conversion and Management, 293:117448.

[49]YuFW, LiGN, ChenHX, et al., 2018. A VRF charge fault diagnosis method based on expert modification C5.0 decision tree. International Journal of Refrigeration, 92:106-112.

[50]YuXR, HeYL, XuY, et al., 2019. A mega-trend-diffusion and monte carlo based virtual sample generation method for small sample size problem. Journal of Physics: Conference Series, 1325:012079.

[51]ZhangCB, TianXN, ZhaoY, et al., 2022. Causal discovery-based external attention in neural networks for accurate and reliable fault detection and diagnosis of building energy systems. Building and Environment, 222:109357.

[52]ZhangCB, ZhangJ, ZhaoY, et al., 2024a. Automated data mining framework for building energy conservation aided by generative pre-trained transformers (GPT). Energy and Buildings, 305:113877.

[53]ZhangCB, LuJ, HuangJH, et al., 2024b. End-to-end data-driven modeling framework for automated and trustworthy short-term building energy load forecasting. Building Simulation, 17(8):1419-1437.

[54]ZhangJ, ZhangCB, LuJ, et al., 2025. Domain-specific large language models for fault diagnosis of heating, ventilation, and air conditioning systems by labeled-data-supervised fine-tuning. Applied Energy, 377:124378.

[55]ZhaoY, LiTT, ZhangXJ, et al., 2019. Artificial intelligence-based fault detection and diagnosis methods for building energy systems: advantages, challenges and the future. Renewable and Sustainable Energy Reviews, 109:85-101.

[56]ZhouM, LiuTY, LiY, et al., 2019. Toward understanding the importance of noise in training neural networks. Proceedings of the 36th International Conference on Machine Learning, p.7594-7602.