Python - Pandas - Remove only splits that only numeric but maintain if it have alphabetic

You can do something like:

s = df['Col1'].str.split('_',expand=True).stack()
s.mask(s.str.isdigit(), '').groupby(level=0).agg('_'.join)

Output:

0    Table_A112
1      Table_A_
dtype: object

Here's one way using str.replace:

df = pd.DataFrame({'Col1': ['Table_A112', 'Table_A_112', 'Table_112_avs']})

print(df)

        Col1
0     Table_A112
1    Table_A_112
2  Table_112_avs

---

df.Col1.str.replace(r'(?:^|_)(\d+)(?:$|_)', '_', regex=True)

0    Table_A112
1      Table_A_
2     Table_avs
Name: Col1, dtype: object

See demo