I do not know of any article or study that "proves" that multiple intelligences exist for math (or for what is worth, any other subject). Such a study will be difficult to design and its results will just show some causation between teaching and learning methods, rather than something physiological within the brain. On top of that, research in math education is focused on the elementary grades and most of the articles that cover middle, high school, or college are very rare.
Gardner (1983) has been the biggest proponent of multiple intelligences and his work seems reasonable. The basic idea is that people prefer to learn and do different things based on their dominant intelligence. For that reason, I am good at math (logical-mathematical intelligence), while my sister is good at making friends (intrapersonal intelligence). John Hopkins University has a good page with references on this subject if you wish to research this in more depth.
As a math teacher, it is your goal to leverage each student's preferred intelligence to have him or her learn math. What I see every day is students that don't have a strong logical intelligence give up solving a math problem but they could explain you word-by-word what they think about a problem or students that are able to solve flawlessly a numerical problem grind to a halt when they need to sketch a net of a solid. This might show some correlation between their strongest intelligence and task performance.
To give you another example, let's consider polynomial multiplication. To multiply two binomial, you can use either the distributive property, a rectangle (or area) model, or even stack them and multiply "the long way". The math is the same but its representation on paper is different. I have students preferring one of these three methods. I do not know why, but it could be related to their dominant intelligence (distributive property for logical-mathematical and rectangle model for visual-spatial intelligence).
At the university level, I would expect students to pretty much be sorted out of math if they do not have a strong logical-mathematical intelligence. After all, you will not go very far in algebra if you are not able to use math symbols or follow algorithms. You can, however, see the different learning styles/intelligences in action in an elementary class where students use different methods to perform basic operations. One could argue that the selection of math discipline could be an example of different intelligences in math (analysis vs. topology for example).