A magnesium potassium phosphate hydrate-based flame-retardant coating (MKPC) is formulated by dead-burnt magnesium oxide (magnesia) and potassium dihydrogen phosphate (KH2PO4), behaving as a matrix. Constituents of the MKPC include wollastonite, vermiculite, aluminum fluoride, aluminum trihydroxide, and calcium carbonate. Some of the ingredients inter-react to produce mullite whiskers at high temperatures, despite an acid-base hydration induced reaction between magnesia and KH2PO4. The MKPC’s thermal, corrosion-resistant, mechanical, and flame-resistant properties were analyzed using scanning electron microscopy, electrochemical corrosion testing, compression testing, thermogravimetric analysis, and freeze/thaw tests. The results show that with the molar ratio = 4 of magnesia to KH2PO4, MKPC demonstrates lower thermal conductivity (0.19 W/m K), along with better corrosion resistance, stronger compressive strength (10.5 MPa), and higher bonding strength (6.62 kgf/cm2) to the steel substrate. Furthermore, acceptable additives to the formulation could enhance its flame-retardancy and increase its mechanical strength as well. Mullite whisker formed from the interaction of wollastonite, aluminum trihydroxide, and aluminum fluoride acts as an outer ceramic shield that enhances mechanical strength and compactness. In addition, Mg-containing minerals with calcium carbonate treated at high temperatures, transform into magnesium calcium carbonate after releasing CO2. At the optimum composition of MKPC (magnesia/KH2PO4 molar ratio = 4; wollastonite:vermiculite = 20:10 wt.%; aluminum trihydroxide = 10 wt.%; and calcium carbonate = 5 wt.%), coated on a steel substrate, the flame-resistance limit results exhibit below 200 °C on the back surface of the steel substrate after one hour of flaming (ca. 1000 °C) on the other surface, and the flame-resistance rating results demonstrate only 420 °C on the back surface of the steel substrate after three hours of flaming (>1000 °C) on the other surface. Both requirements for the flame-resistance limit and three-hour flame-resistance rating are met with the optimum compositions, indicating that MKPC plays an effective role in establishing flame-retardancy.

No abstract

No abstract

No abstract

One of the key challenges to deep reinforcement learning (deep RL) is to ensure safety at both training and testing phases. In this work, we propose a novel technique of unsupervised action planning to improve the safety of onpolicy reinforcement learning algorithms, such as trust region policy optimization (TRPO) or proximal policy optimization (PPO). We design our safety-aware reinforcement learning by storing all the history of "recovery" actions that rescue the agent from dangerous situations into a separate "safety" buffer and finding the best recovery action when the agent encounters similar states. Because this functionality requires the algorithm to query similar states, we implement the proposed safety mechanism using an unsupervised learning algorithm, k-means clustering. We evaluate the proposed algorithm on six robotic control tasks that cover navigation and manipulation. Our results show that the proposed safety RL algorithm can achieve higher rewards compared with multiple baselines in both discrete and continuous control problems. The supplemental video can be found at: https://youtu.be/AFTeWSohILo.