If the value is a number or string then you can convert it to a date with an appropriate mask - which is what you are missing, and what is causing the error you are getting (as it's using your session's NLS_DATE_FORMAT
setting, which apparently does not match the format of the data; but which you should not rely on anyway, as @MTO said in comments):
to_date(ID_BB_SECURITY, 'YYYYMMDD')
and then extract the month number from that:
select extract(month from to_date(ID_BB_SECURITY, 'YYYYMMDD')) from BT_EXPORT
Or you could just use a substring:
select to_number(substr(ID_BB_SECURITY, 5, 2)) from BT_EXPORT;
Those assume a fixed consistent format, which is always a risky assumption when using the wrong data type. Ans if it's a number they are doing an implicit conversion from number to string, which you could turn into an explicit conversion for greater clarity.
If it's already a date - as it should be, of course - then you don't need the conversion:
select extract(month from ID_BB_SECURITY) from BT_EXPORT
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…