In many cyber-physical systems, imaging can be an important but expensive or `difficult to deploy' sensing modality. One such example is detecting combustion instability using flame images where deep learning frameworks have demonstrated high performance. The proposed frameworks are shown to be quite trustworthy such that the domain experts can have sufficient confidence to use these models in real systems to prevent unwanted incidents. However, flame imaging is not a common sensing modality in combustion engines today. Therefore, the current roadblock exists on the hardware side regarding the acquisition and processing of high-volume flame images. On the other hand, the acoustic pressure time series is a more feasible modality for data collection in real combustors. Therefore, the optimal solution can be to utilize acoustic time series as a sensing modality and, at the same time, to implement the image-based deep learning frameworks for more efficient detection of combustion instability. To achieve this, we propose a novel model that can reconstruct cross-modal visual features from acoustic pressure time series in combustion systems. The results demonstrate that the detection accuracy can be enhanced using only time series data by using our approach. Our proposed approach provides the unique benefit of generating synthetic visual features corresponding to time series information. With our proposed model, we anticipate that deep learning based instability detection frameworks will become more feasible by avoiding the use of imaging which is an expensive sensing modality. By deploying such frameworks on a real system, instability can be detected effectively, preventing revenue loss from unwanted incidents (e.g., flame blowout and structural damage to the engines). By providing the benefit of cross-modal reconstruction, this model can prove useful in different domains well beyond the power generation and transportation industries.