Although current advanced generative models exhibit strong adherence to complex textual descriptions, industrial applications often require numerical conditions for the target object, such as the drag coefficient of an aircraft or the maximum thickness of a wing. In some complex scenarios, the target object must satisfy a large number of numerical conditions. To address this challenge, this paper proposes a novel latent diffusion model framework conditioned on multidisciplinary numerical characteristics for the inverse design of airfoils. The framework first employs an airfoil autoencoder to perform dimensionality reduction on the airfoil point cloud data, extracting latent space representation of the airfoil. Subsequently, the diffusion model performs conditional generation within this latent space. Experimental results demonstrate the proposed model's excellent adherence to conditions, enabling it to generate airfoils that meet complex requirements directly. The model's strong performance serves as a valuable reference for other similar applications.