The underground concrete structure is in a complex and changeable service environment for a long time, and it is often affected by the combined effects of load, corrosive ions and temperature changes. There is a significant nonlinear coupling relationship among these three factors, which makes the concrete show a completely different damage evolution law from the single effect during the service process. Especially in the underground environment, chloride ions and sulfate ions will accelerate pore dissolution and micro-crack propagation, and temperature changes will vary the hydration rate and ion migration ability. Under external loads, the generation and propagation of microcracks further promote ion penetration and chemical erosion, resulting in a rapid degradation of material properties. Conventional durability tests are often difficult to quantitatively describe the damage law under the combined action of multiple factors, and it is difficult to accurately predict the service life of underground concrete. It is thus of great significance to reveal the deterioration mechanism and service life evaluation of underground concrete via constructing a method that can comprehensively reflect the damage degree of concrete under multi-factor coupling and realize a life prediction.

In the experiment, ordinary Portland cement was used to prepare concrete specimens, and the mix ratio of each group was kept consistent. The test was carried out under the accelerated conditions of simulated underground environment, and the service environment of different load, ion and temperature combinations was set up to investigate the damage evolution law under the coupling of three factors. During the test, the five indexes of concrete failure load, ultrasonic wave velocity, dynamic elastic modulus, rebound value and quality change were tested regularly. These indexes could reflect the mechanical properties, internal compactness, surface hardness and mass loss of concrete, respectively, as the important indexes for comprehensive evaluation of damage degree. In addition, the evaluation system of concrete damage degree was also established based on the entropy weight method, and the weight was automatically assigned by the dispersion of index changes, so that the damage degree evaluation could be more objective and comparable.

For the life prediction, four models were selected for comparison, including Even Grey Model (EGM), Even Discrete Grey Model (EDGM), Optimized Discrete Grey Model (ODGM) and Discrete Grey Model (DGM), based on the grey system theory. The grey model has a good prediction performance under the condition of small sample and weak information, and is suitable for durability research with limited test period and small data volume. The accuracy of the model could be improved by cumulative generation and parameter analysis, and the prediction performance of each model was evaluated by various error indexes, and the optimal model was selected for the life prediction.

The damage degree of concrete under four different conditions firstly increases and then decreases. The failure load, ultrasonic wave velocity, dynamic elastic modulus, rebound value and mass change trend under the same conditions are similar. However, the turning points of damage attenuation under the four conditions are not consistent due to the difference of hydration reaction and internal pore structure. At 50 ℃, the faster early reaction rate can result in the faster pore filling, thus improving the compactness.

Based on the entropy weight method, the weight coefficient of each index to the damage degree of concrete is obtained. The influencing degree from large to small is failure load, dynamic elastic modulus, rebound value, ultrasonic wave velocity, and quality change. Also, the accuracy ranking of the four prediction models is obtained based on the comprehensive score ranking method. The prediction accuracy from high to low is ODGM > EGM = EDGM > DGM.

Under the combined action of load-ion-temperature, the failure load, ultrasonic wave velocity, dynamic elastic modulus, rebound value and mass of concrete all firstly increased and then decreased, but the turning points of damage under various environmental effects were inconsistent. In addition, the hydration products under chloride erosion environment were mainly filled with micropores (i.e., ≤ 0.1 mm³) and small pores (i.e., 0.1-1.0 mm³), and the pores of concrete at 50 ℃ in the early stage were more dense than those at 20 ℃. The damage degree of each index was quantified by the entropy method, i.e., failure load (i.e., 80%-90%), dynamic elastic modulus (i.e., 5%-8%), rebound value (i.e., 1%-6%), ultrasonic wave velocity (i.e., 1%-4%), and mass change (i.e., < 1%). The ODGM model had the highest accuracy in predicting the service life of concrete under the combined action of load-ion-temperature. The service life of concrete was 82 months, 29 months, 24 months and 20 months under the four environments (i.e., 0.2Fmax, 3%NaCl, 20 ℃ and 0.2Fmax, 3%NaCl, 50 ℃ and 0.2Fmax, 5%Na₂SO₄, 20 ℃ and 0.2Fmax, 5%Na₂SO₄, 50 ℃), respectively.

As a type of energy-saving and environmental protection building system, steel structure building is widely used in engineering construction in recent years, and its fire protection measures are particularly important. At present, the commonly used fire retardant coatings are mainly divided into two categories, i.e., organic intumescent and inorganic non-intumescent coatings. Although organic intumescent coatings exhibit a good fire resistance in the initial stage, their duration of protection is relatively short, and they may release toxic and corrosive gases during combustion, posing certain safety risks. In contrast, inorganic non-intumescent fireproof coatings have advantages such as high thermal resistance, strong durability, good wear resistance, and low cost, making them considered as an ideal fire protection material for steel structures. However, some issues such as insufficient adhesion and poor overall durability seriously limit its widespread application in practical engineering. Magnesium phosphate cement (MPC) is regarded as a fireproof material with an application potential due to its high strength, high-temperature resistance and excellent interfacial bonding performance with steel. However, the conventional raw material for MPC like dead-burned magnesia has a high density and carbon emissions, which is unfavorable for its green and sustainable development. To address the above issues, this study was to select a low-grade natural brucite powder as a complete substitute for dead-burned magnesia to prepare magnesium phosphate cement, and optimize the physical, mechanical, and fire insulation properties of brucite-based magnesium phosphate steel structure fireproof coatings (BMPC) via introducing modified expanded perlite.

The raw materials for the preparation of BMPC included natural brucite powder, ammonium dihydrogen phosphate, analytical grade borax, and expanded perlite. Expanded perlite was divided into two types, i.e., untreated (EP) and water absorption pretreatment (HEP) according to different treatment methods. Under the condition of fixed process parameters (i.e., M/P mass ratio of 2.5, borax content of 15%, water-cement ratio of 0.17, and expanded perlite content of 0–20% of the total mass of brucite powder and ammonium dihydrogen phosphate), the BMPC slurry was prepared.

The dry density and surface drying time of the BMPC coating under natural curing conditions were tested in accordance with the standard GB 14907—2018. The compressive strength and adhesive strength tests were conducted on the specimens after 3-d, 7-d and 28-d curing, respectively. The thermal conductivity of BMPC was determined by a transient plane source method. The fire resistance limit was tested according to the standard GB/T 9978.1—2008. The internal pore structure of BMPC samples was analyzed by an industrial computed tomography system, and its microstructure was analyzed by field emission scanning electron microscopy. The phase change of BMPC coating before and after fire test was analyzed by X-ray diffractometry, and its thermal stability in the range of 30–1000 ℃ was analyzed by thermogravimetric analysis.

The expanded perlite (HEP) after water absorption treatment has a more open pore structure and a higher surface roughness rather than the untreated expanded perlite (EP). In the BMPC system, the HEP has a relatively small effect on the mechanical properties, especially under low density conditions, it can still maintain high compressive strength and adhesive strength. At the HEP content of 20%, the microporous structure inside the material significantly reduces the thermal conductivity and thermal diffusion coefficient. This is mainly since the HEP acts as a low thermal conductivity filler, and the air layer retained in its pores (i.e., thermal conductivity of 0.026 W/(m·K)) effectively blocks heat transfer. Also, the introduction of HEP optimizes the pore structure, reduces the proportion of large pores, makes the heat conduction path more complex, delays the diffusion process of heat into the substrate, and significantly slows down the flame propagation rate and thermal decomposition process. The results of fire resistance test and pore structure show that the porosity of BMPC-HEP sample is higher than that of the control group, and the standard deviation of pore fluctuation increases from 4.0% to 5.1%, indicating that HEP forms a lightweight porous structure at a high temperature, further improving the thermal insulation performance and thermal stability of the material. The addition of HEP has a significant effect on the microstructure of BMPC at a high temperature. Some struvite and unreacted raw materials absorb a large amount of heat during pyrolysis, which plays a role in delaying temperature rise and enhancing thermal buffering, and provides an important support for the thermal insulation protection mechanism of fire retardant coatings.

The expanded perlite after water absorption treatment had little effect on the mechanical properties of BMPC system. When the density reduced from 1062 kg/m³ to 647 kg/m³, the 28-d compressive strength and adhesive strength were only reduced by 26.5% and 17.7%, respectively. At the content of HEP of 20%, the thermal conductivity and thermal diffusivity of BMPC were 49.3% and 44.0% lower than those of BMPC without HEP, respectively. The significant improvement of this thermal performance could be mainly attributed to the fact that HEP itself had low density and low thermal diffusion characteristics, effectively slowing down the heat transfer rate in the process of temperature change, and thereby improving the thermal insulation performance of the material. The fire resistance limit of BMPC-HEP system reached 206 min, which was significantly better than that of the control sample. Its excellent fire resistance was mainly attributed to the synergistic effect of porous structure, low thermal conductivity filler and microstructure stability at a high temperature, indicating that it could have a promising application potential in the field of fire retardant coatings for steel structures.